All

The short answer The SOMI Partner Network is a vetted group of specialised manufacturing factories in China, coordinated through one point of contact. Buyers get CNC machining, sheet metal fabrication, metal stamping, die casting, injection molding and 3D printing from one programme, with one engineering review, one quality route and one quotation, instead of managing a separate supplier for every process. What the network actually is It is not a single factory and it is not a marketplace. It is a managed supplier group: a set of factories that each specialise in one or two processes, sitting behind a single commercial and engineering interface. SOMI states that it has built this group over more than ten years of manufacturing and service work and has supplied precision parts to more than 1,200 customers. The point of the structure is that the factory doing the work is chosen for the job rather than for convenience. A thin-wall injection moulding part and a five-axis machined bracket are genuinely different businesses, and a buyer who wants both is usually better served by one programme that can route each part to a capable plant than by one plant that claims to be excellent at everything. The six processes and six industries it covers The network is organised around six production capabilities and six end markets, which is what makes it possible to keep a multi-part programme in one place. CapabilityWhat it coversTypical output CNC machiningTurning and 3, 4 and 5 axis millingPrecision metal and plastic parts, prototypes to production Sheet metal fabricationLaser cutting, bending, welding, finishingEnclosures, brackets, panels and frames Metal stampingProgressive and single-die stampingHigh-volume formed parts and contacts Die castingAluminium and zinc alloysHousings, covers and structural castings Injection moldingPlastic, two-shot and insert moldingHousings, enclosures and consumer parts 3D printingFDM, SLA, SLS and MJFPrototypes, jigs and low-volume functional parts The industries served are consumer electronics, automotive, medical, aerospace and aviation, robotics and automation, and industrial machinery. Those are demanding sectors on tolerance, traceability and documentation, and they are the reason the network puts engineering review in front of the quotation rather than behind it. Four production families, one programme. The part goes to the process it needs. How a factory gets onto the network The value of a supplier group is entirely determined by what it refuses to accept. SOMI publishes a six-stage selection and onboarding process, and the stages are worth reading as a buyer because they tell you what has already been checked before your drawing arrives. Preliminary assessment. The factory environment, production equipment, capacity and quality standards are established on site. In-depth evaluation. The quality management system and environmental and social compliance are reviewed in detail against the network's eligibility criteria. Legal due diligence. The supplier has to hold the correct registrations and licences, and the company behind it is background-checked. Technical requirement. Process capability and engineering depth are assessed, including whether the factory can optimise a design and produce CAD or 3D drawings rather than only cut to a drawing. Trial order. A sample order and then a small batch confirm that the factory can hold the quality standard in production, not just on the first part. Ongoing assurance. Partners are audited periodically and reviewed on delivery, quality and communication, with continuous improvement expected. Alongside those stages, partners have to meet standing requirements: at least one production capability, at least five years of production experience, after-sales technical support, documented quality-management evidence and the registrations that apply to their process, in-house design capability, contract compliance and adherence to the network's intellectual property policy. Where a customer programme requires a specific quality-system certification, that requirement is verified against the actual factory doing the work before the order is released rather than assumed from a network-level statement. Selection, then a trial order, then continued auditing. The trial order is the stage that matters most. How a project runs once it is on the network The buyer-side flow is deliberately short. You upload the CAD and 3D files, and they are treated as confidential. Engineering checks and confirms the sample specification, dimensions and surface treatment. If you have no drawings, the technical team produces them; if you have them, they are reviewed for manufacturability before quoting. A quotation follows, then a sample order so you can approve the physical part, then mass production, then quality control against the agreed specification, then delivery with tracking through the logistics partner. The single most important step in that sequence is the design review before the quotation, because most of the cost of a part is fixed by the wall thickness, material and tolerance decisions that are already on the drawing when it arrives. Catching a moulded rib that will sink, or a machined internal corner that cannot be reached, before the tool is cut is worth more than any discount negotiated afterwards. Why one coordinated network beats five direct suppliers Buying each process from the cheapest specialist looks like the low-cost route and usually is not. The quoted price is only part of the cost: every additional supplier adds purchase orders and invoices, a separate quality visit, inbound inspection, rework when two vendors interpret a tolerance differently, and freight between vendors. Manufacturer case data published by integrated suppliers puts the hidden cost of a fragmented supplier base at roughly 20 percent or more above the quoted price, and reports total landed cost premiums of around 18 to 28 percent on a three-supplier arrangement against 0 to 7 percent when the work is consolidated, annual defect rates of 3.2 to 6.8 percent across separate vendors against under 0.4 percent consolidated, and lead times of 45 to 62 days against 21 to 35 days. The lead time effect is largely queue time: five vendors can add three to four weeks of pure waiting, with parts sitting between the cutter, the welder and the coater rather than being worked on. Treat those figures as planning ranges from supplier case studies rather than universal constants, but the direction is consistent and it is the reason consolidated sourcing keeps growing. The savings come from removing touchpoints, not from a bigger discount. What the network does not do It is not a single factory. Coordination has to be paid for. If a program is one process, one material and very high volume, a dedicated plant that owns the whole line can still be the better answer, and we will say so. It does not transfer design responsibility. Engineering review reduces risk; it does not make the buyer's functional requirements our problem. The specification, the load case and the acceptance criteria stay with you, and the first article approval is yours. Certification is program-specific, not network-wide. Regulated markets ask for particular quality systems and particular records. Those have to be confirmed for the factory and the program in question before production, not inferred from a website. Tooling and intellectual property need written terms. Who owns the mould, where it is stored, what happens if the program ends and how a design change is priced all belong in the agreement. The network's intellectual property policy is a starting point, not a contract. Capacity priority is commercial. A network can route work to whichever partner has the right capability and the free capacity, but a rush job competes with other customers' work and should be planned rather than assumed. Logistics and terms are still your project. Incoterms, shipping method, customs documentation and lead time across the destination border sit outside the manufacturing steps, and they have to be agreed up front. How to start Send the CAD or 3D files with the drawing, or the physical sample if no drawing exists, together with the annual volume, the target lead time, the critical tolerances and the surfaces you care about. If the program spans several processes, say so at the start: that is exactly the case the network is built for, and routing the parts to the right factories before quoting is what keeps the total cost visible. See plastic injection molding for the moulded side of the network, metal stamping for high-volume formed parts and sheet metal fabrication for enclosures and frames. Send your drawings for one coordinated quotation Scope and sources. Network structure, capability list, partner requirements and the six onboarding stages were taken in 2026 from the SOMI Custom Parts partner programme page and from the company service and industry overview, which states six production capabilities, six served industries, more than ten years of manufacturing and service experience and more than 1,200 customers. Consolidated-sourcing effects are drawn from supplier case data published by a full-service sheet metal fabricator, a review of vendor-consolidation economics citing procurement studies and consolidation case data for injection molding orders. Those consolidation percentages and lead times are commercial case results, not audited benchmarks, and should be treated as planning ranges rather than promises. Quality-system, certification, tooling-ownership and intellectual property terms are agreed per program and per factory; confirm them in writing before releasing production.

The short answer Nothing, in either direction. A manufacturing partner pays no joining fee, no listing fee and no annual subscription, and a buyer pays nothing to use the network - only for the parts actually ordered. What does cost money is the work: samples, tooling, third-party testing, freight, and any audit you ask us to run. What free to join covers, and what it does not Free means access, not manufacturing. There is no membership tier to buy, no paid placement inside the network and no charge for the first engineering pass on a drawing. Every item that costs real money is either a physical thing, such as a die or a first-off sample, or an outside service, such as a laboratory test or a freight booking. Each of those appears on the quotation as its own line, which is the point: a cost you can see can be challenged, and a cost buried in a membership cannot. Access is free. Material, machine time and outside services are not. ItemWho bears the costTypical range Membership, listing and subscriptionNobodyZero, no annual fee DFM review and quotationSOMIIncluded at the enquiry stage Samples and first article inspectionBuyerCharged at the quoted sample price Tooling, dies and mouldsBuyerUSD 400 to 20,000 and up, by complexity Third-party testingBuyerAt the laboratory or agency rate Freight, duty and customs clearanceBuyerAt the carrier rate On-site factory audit you requestBuyerOne day on site plus travel What a factory actually pays to be in the network The entry price for a manufacturing partner is time and paperwork rather than money, and the size of that bill depends on how ready the factory's records already are. Qualification covers the legal entity and its scope, the equipment list and capacity, gauge calibration, material traceability and the quality system that is actually in use rather than the one on the wall. Published onboarding timelines for custom parts put supplier qualification and quality-system review at about two to four weeks, with the first-article and production-part approval stage adding another two to six weeks after tooling is ready. For a new factory building a new die or mould, that is eight to fourteen weeks between first contact and an approved production part. The mix of that work changes with complexity. Straightforward stamped or moulded parts can be documented in three to four weeks once the tooling exists. Medium-complexity housings and assemblies that need multi-stage processing, third-party performance testing and measurement-system analysis run six to eight weeks. Safety-critical parts with full dimensional reports, capability studies and a capacity verification run ten to sixteen weeks, and the wait is usually for evidence rather than for metal. The expensive outcome is not a fee, it is a repeat audit. Reports from automotive supplier onboarding put the share of Chinese factories that fail their first on-site process audit at around 40 percent, and the findings repeat because they are administrative rather than technical: gauges without current calibration certificates, a quarantine or material review area that is not controlled, incomplete training records for key operators, and sub-tier suppliers such as plating or heat treatment sitting outside the factory's own quality system. Fixing those costs weeks. Not having them costs nothing. What a buyer pays, and what sits outside the part price A quoted unit price for a machined, stamped, moulded or cast part normally carries material, machine time, the standard surface finish and the standard inspection. It does not carry tooling, which is a one-off charge, and it does not carry anything the part has to be told to do later. The lines that sit outside it are predictable: samples and first-article inspection, the die or mould itself, any third-party test that a regulator or your own customer requires, freight, duty and customs clearance, and expedite premiums when a date moves. A buyer who compares two quotations without checking which of those lines are included is comparing two different scopes, not two prices. Where a free route can cost more than a fee Most sourcing routes in this market are not free and the fee models are published, so they can be compared line by line. A commission agent takes a percentage of order value, a flat-fee agent takes a fixed amount per order, a retainer buys monthly availability, and a full-service provider charges a higher percentage for handling compliance and logistics as well. A route that advertises itself as free is normally earning somewhere else in the transaction. Published market fee models. With no separate fee line, the comparison becomes part price against part price. ModelHow it is chargedTypical market range Commission agentPercent of order value3 to 10 percent of FOB value Flat fee per orderFixed amount each orderUSD 200 to 1,500 per order Monthly retainerFixed monthly paymentUSD 500 to 3,000 per month Hourly consultingPer hour workedUSD 30 to 80 per hour Full-service providerPercent plus project work8 to 12 percent of order value SOMI Partner NetworkInside the quoted part priceNo separate fee line The fee that looks cheapest is the one to examine first. A commission below three percent is usually recovered somewhere else, most often in freight or inspection margin, and the example that circulates in this market is an agent quoting four percent while quoting USD 1,180 for freight that the carrier priced at USD 760 - the six-point discount on commission erased several times over. The discipline is simple: ask for the carrier invoice, ask in writing whether the agent receives any rebate from the factory, and normalise every offer to a landed cost per part before comparing it. Where the money goes on a fifty thousand dollar order Arithmetic on the published ranges makes the difference concrete. An order of USD 50,000 carried by a five percent commission agent adds about USD 2,500 in fees before freight is counted, and the fee scales with the order rather than with the work done. A flat fee of USD 200 to 1,500 is cheap on that order and expensive on a repeat order of USD 5,000. A retainer of USD 500 to 3,000 per month only amortises if you place enough work through it. And a free listing that earns its money in freight margin is not free at all, it is a different invoice. Two of the four stages carry a charge, and both are itemised rather than packaged as a fee. Where the service cost sits inside the quoted part price, the comparison becomes part price against part price, which is the only comparison a buyer can actually act on. Ask for the quotation broken into material, machine time, finishing and inspection, and the same arithmetic can be run on real numbers instead of on a market range. Limits: what no fee does not mean It is not a price guarantee. Free to join says nothing about the price of a part. A free network with a slow quotation process can cost more in engineering time than a paid one with a fast one. It does not include tooling. Dies, moulds and fixtures are paid for by the buyer, from roughly USD 400 to 1,000 for a simple solid extrusion or stamping die up to USD 2,000 to 20,000 and beyond for complex hollow profiles and multi-station progressive dies. It does not include regulation. Flame, food-contact, medical and automotive approvals are grade-level and record-level work. Sampling, documentation and third-party testing are charged and they are not optional. It does not buy priority. Rush work competes with other customers' schedules. Expedition is a commercial decision with a price attached, not a membership benefit. It does not remove the commercial terms. Minimum order quantity, payment terms, tooling ownership and storage, and the exit arrangements when a programme ends all still have to be agreed in writing before production. It does not make our margin invisible. Our cost sits inside the part price. If you want to see it, ask, and the material, machine time, finishing and inspection lines can be separated on the quotation. It says nothing about certification. Whether a particular factory holds a particular quality system certificate is confirmed per programme and per factory, in writing, before an order is released. How to get a costed answer instead of a range Send the 3D model, the 2D drawing with tolerances, the annual volume and the delivery destination, and the first pass - geometry review, process route and quotation - is at no charge and no commitment. If the programme spans more than one process, say so at the start, because routing the parts before quoting them is what keeps the total cost visible. See how the partner network works, plastic injection molding for moulded parts and metal stamping for high-volume formed parts. Send a drawing for a no-fee quotation Scope and sources. Fee models, commission bands, flat fees, retainers and the freight-margin example were compiled in 2026 from published sourcing-fee guides by HiSourcing, Statrys, WooSourcing and Feisourcing, which report commission at 3 to 10 percent of FOB value, flat fees of USD 200 to 1,500 per order and retainers of USD 500 to 3,000 per month. Qualification and approval durations come from a published stamping supplier onboarding timeline (qualification two to four weeks, production-part approval two to six weeks after tooling, eight to fourteen weeks end to end) and from a PPAP cycle-time breakdown giving three to four weeks of preparation for simple stamped or moulded parts, six to eight weeks for medium complexity and ten to sixteen weeks for safety-critical parts. The first-audit failure rate and the recurring findings come from an automotive supply chain certification guide. These are market ranges and planning figures, not quotations for any specific part. Nothing on this page states or implies a certification held by any factory; quality systems are confirmed per programme and per factory in writing before production.

The short answer Two tracks. A buyer joins in one step: send a drawing and the quotation workflow starts, within 24 to 48 hours. A factory joins through qualification - document review, an audit, then sample validation - which takes about two to four weeks for the review and audit and six to fourteen weeks end to end when custom tooling is involved. Two tracks, two timelines One question decides which process applies: are you buying parts, or making them? A buyer's entry step is a file transfer. A factory's entry step is evidence. Neither costs money, and neither is a formality, but the two timelines differ by an order of magnitude, so it is worth knowing which one you are on before a launch date is fixed. If you are buying, joining means you are inside the quotation workflow, and nothing else changes: one project manager and one point of contact from the first enquiry through to the shipment. If you are making, joining means your records, your equipment and your output have been checked against a specification that a third party can audit, and the first order is deliberately small for exactly that reason. What we do not do is charge for the door. There is no application fee, no listing fee and no subscription on either side, which is why the honest answer to how long it takes is about paperwork and sample parts rather than about money. The buyer track: four steps from enquiry to released order The buyer path is short because the checks happen on the part rather than on you. Share the files. A STEP or IGES model plus a 2D PDF drawing with tolerances is enough to quote. A physical sample can stand in for a model where no CAD exists, but a drawing with tolerances is what makes two quotations comparable. DFM review. Geometry, tolerances, wall thickness, bend radius, draft and material are reviewed against the process route before price is discussed. This is where cost is usually found: a wall that is too thick for the cycle time, a tolerance no process holds at that size, or a bend that needs a relief cut. Quotation. Unit price, tooling, inspection, lead time and the assumptions behind each of them, in writing. Anything the price depends on that has not been stated is a change order waiting to happen. Release. Sample or first article, approval against the drawing, then production. Repeat orders skip the first two steps. Four steps, one of which needs something from you. What to send with the first message Most of the delay in a first quotation is a missing file rather than a slow engineer. This is the short list, and each item changes a number that appears on the quote. Send thisWhy it moves the price or the date 3D model: STEP, IGES, X_T or STLCycle time, tooling route and DFM comments 2D drawing with tolerances and finish calloutsInspection plan, gauge choice and whether a process can hold it Material and surface finishCost, lead time and whether a secondary supplier is needed Annual volume and first-order quantityProcess choice: tooling-free work and tooled work cross over at different points Target date and delivery destinationProduction slot and freight mode Compliance requirement, if anyWhether a specific grade, record set or third-party test applies NDA or quality agreement, if you need oneSign it before drawings are shared, not after The same list, seen from the factory side of an application. The factory track: what has to exist at each gate For a factory, joining is a sequence of six gates, and each one asks for evidence rather than a promise. Application. What the factory makes, in which processes, on which equipment, and at what tonnage or envelope limits. Entity and scope check. Legal entity, registered scope, facility location and ownership of the machines, so that capacity claimed is capacity controlled. Document review. Quality manual and records, gauge calibration status, material certificates, inspection capability and the list of sub-tier suppliers. Audit. One day on site against a checklist shared in advance, or a remote process review with live video where travel is not practical. For orders above roughly USD 50,000, published guidance treats the on-site audit as mandatory rather than optional. Sample and first article. Trial parts produced against a frozen drawing and reported dimensionally, which is the only step that tests the claims made in the first four. Approved status. Entry on the routing list, an agreed escalation path for quality issues and the first production order. How long it takes, and what makes it slip The published numbers for custom parts are consistent. Qualification and quality-system review take about two to four weeks. The first-article and production-part approval stage adds two to six weeks after tooling is ready, which puts eight to fourteen weeks between first contact and an approved production part when a new die or mould has to be built. Measured on its own, an approval package takes three to four weeks to assemble for simple stamped or moulded parts, six to eight weeks for medium-complexity parts that need multi-stage processing and third-party testing, and ten to sixteen weeks for safety-critical parts. Stages overlap; the total is set by the slowest evidence, not by the sum of the rows. StageTypical durationWhat makes it slip Document review1 to 3 weeksGauge calibration or material certificates missing On-site or remote audit1 day, plus schedulingChecklist and scope not shared in advance Corrective actions2 to 6 weeksFindings in quarantine control or traceability Sample and first article1 to 4 weeksDrawing not frozen, or a change made mid-sample Production release1 to 2 weeksApproval level not agreed before samples ship Why applications stall, and what stops the clock The findings that delay a first approval are almost never about machining capability. They repeat, and they are administrative: Gauges without current calibration. A measurement without a calibration certificate is an opinion, and an inspector will treat it as one. An uncontrolled quarantine or material review area. If non-conforming parts are not physically separated, nothing downstream can be trusted. Incomplete training records. Key operators and inspectors need traceable qualification, not verbal seniority. Sub-tier suppliers outside the system. Plating, heat treatment and coating are where most stamping and machining defects are actually created, and they are often not in the factory's own quality system on paper. Broken material traceability. A mill certificate that takes three weeks to appear is a warning sign rather than a formality. Two practical habits cut most of the delay. Send the checklist two weeks before the visit so the factory prepares the right evidence instead of guessing, and agree the approval submission level before tooling is released rather than after samples ship. Neither is expensive, and both remove the most common cause of a slipped date, which is surprise. Limits, and how to start Not every applicant is accepted. A network is only as good as its worst member, so capability, capacity and record discipline are all part of the decision. Joining does not guarantee volume. Work is routed on capability, capacity, price and past performance, in that order. The audit does not replace your customer's audit. If your own customer requires their own visit, their own checklist or their own approved vendor list, that runs on top of this one. Regulated and export-controlled work needs more. Medical, aerospace and defence programmes carry documentation and control requirements that are assessed case by case before anything is quoted. No certification claims are made here. Which quality system a factory holds, and to what scope, is confirmed per programme and per factory in writing rather than assumed from a website. To start on either track, send what you already have. A buyer needs a model and a drawing. A factory needs a capability list, a licence and the process it wants to be considered for. From there the next step is a free design review on a live part or a scheduled audit, and both are quoted in weeks rather than in fees. See the partner programme for the factory side, CNC machining for the most commonly routed process and sheet metal fabrication for formed and welded work. Start an application or send a drawing Scope and sources. Onboarding durations and the sequence of gates were compiled in 2026 from a published supplier onboarding timeline for stamping parts (qualification two to four weeks, approval two to six weeks after tooling, eight to fourteen weeks end to end, on-site audit mandatory above roughly USD 50,000) and from a PPAP cycle-time breakdown giving three to four weeks for simple stamped or moulded parts, six to eight weeks for medium complexity and ten to sixteen weeks for safety-critical parts. The recurring audit findings, the corrective-action window and the reported first-audit failure rate come from an automotive supply chain certification guide and from an overview of the six-stage automotive supplier admission process. These are planning ranges drawn from published experience, not commitments, and they move with the completeness of the documentation submitted. Nothing here states or implies that any factory holds a particular certificate; quality systems are confirmed per programme, per factory and in writing before production.

The short answer For a buyer the benefit is subtraction: one engineering review, one quotation, one inspection standard, one invoice and one point of accountability across six processes, instead of five separate vendor relationships. For a factory the benefit is inbound work with no listing fee. The hidden surcharge of fragmented sourcing is around 20 percent of the quoted base price. The same network, two different sets of benefits It helps to separate the two sides, because the words benefit means different things depending on which end of the order you sit. A buyer is buying removed work: fewer interfaces, fewer invoices, fewer places for a defect to hide. A factory is buying demand: a route to programmes it could not quote alone, without paying for a listing or a subscription. Same network, two different ledgers. The rest of this answer deals with the buyer side, because that is where the numbers are measurable, and then returns to what a factory actually receives. Fewer interfaces, and the cost they carry A multi-vendor chain does not fail loudly. It leaks. Each vendor adds a purchase order, an inbound inspection, an invoice, a transit leg and a chance for a coating to be chipped or a bracket to arrive bored to the wrong datum. Published breakdowns of the surcharge put inter-vendor logistics at around 6 percent of base cost, rework and transit damage at around 5 percent, per-vendor quality audits at around 4 percent, purchase order and administration overhead at around 3 percent, and expedited freight at around 2 percent, for a total hidden layer of about 20 percent on top of the lowest quoted price. None of these lines appear on a vendor quotation. All of them appear on the total cost of acquisition. Consolidation cases point the same way. Documented fabrication-and-finishing consolidation results report a cost per part around 32 percent lower after the process chain was brought under one supplier, with the saving coming from removed logistics, unified inspection and eliminated rework rather than from a larger markup. That is a commercial case result rather than an audited benchmark, and it should be treated as a direction of travel rather than a promise. It is worth setting against the alternative: keeping two or three qualified suppliers is not free either, and the multi-source premium - the extra management cost of running them - is typically reported at 3 to 5 percent of procurement spend, with a practical threshold of roughly USD 200,000 to 500,000 of annual volume per category before a split pays for itself. Lead time is lost in queues, not in machines Machining a bracket takes minutes. Waiting for a bracket takes weeks. In a five-vendor chain, parts finish cutting on Monday, ship on Tuesday, and then wait for the welder's queue to open on Thursday. Welding finishes, parts ship again, and the coater's queue starts the following week. Five vendors can easily add three to four weeks of pure queuing in which no value is added and no machine is running. Where the steps sit under one roof, the same hand-offs happen the same day, with no transit and no queue reset. This is the mechanism behind the second published case: a fabricator delivering 35 complex information kiosks from design to finished assembly in six weeks, where the timeline held because every step happened in-house with no external hand-off. It is also visible in the multi-source data, which cuts the other way: moving from a single supplier to two or three qualified ones raised reported on-time delivery from 83 to 94 percent and cut disruptions from 2.7 to 1.1 a year, at the price of added coordination. A coordinated network is an attempt to keep the second number without paying the whole first number. Quality you can attribute The practical problem with fragmented sourcing is not that any one shop is bad. It is that when a welded assembly fails, no single vendor owns it. The bracket was in tolerance, the weld was in tolerance, and the powder coat hid the problem until the customer found it. Defect rates across different factories building to identical specifications are reported to vary by 5 to 15 percent, driven by operator training and internal quality culture rather than by the specification. With one route, one drawing set and one inspection standard, a non-conformance has a location and a cause instead of a debate. What that means in day-to-day terms: one drawing revision status across all processes, so the machinist and the welder are not working to different issues of the same print. One first article that covers the finished assembly rather than four first articles that each pass and do not fit together. One report with the shipment. And one place to send a corrective-action request that has to come back with an answer. Engineering that happens before tooling The largest savings in these programmes are usually designed in rather than negotiated later. When bending, welding and finishing sit in the same review, the tolerance stack across the three operations is calculated once, at the quotation stage, instead of being discovered at the trial assembly. A bend relief that a welder needs, a hole that a coater will fill with powder, a datum that a machinist needs to hold flatness: each of these is a small design change before a die is cut and an expensive rework after. The other engineering benefit is a fair comparison. When the same team quotes a stamped bracket and a moulded one, the metal-versus-plastic decision is made on real unit prices at real volumes, including tooling amortisation, rather than on a rule of thumb. What the network does not do It is not a marketplace with instant pricing. Routing a multi-process programme takes engineering judgement, which is why the first step is a design review rather than a checkout page. It is not always the lowest unit price. For one simple part in one process, a specialist shop bidding on its own may quote lower than a coordinated programme. Consolidation pays through the total cost of acquisition, not through the unit price line. It does not remove your due diligence. If you are regulated, you still need your own approved supplier list, your own audit rights and your own records. It does not make capacity infinite. Capacity is process-specific and finite, and peak seasons and the Lunar New Year shutdown are real constraints that have to be planned around rather than argued with. It does not decide your commercial terms for you. Tooling ownership, storage, IP, payment terms and the exit arrangements when a programme ends belong in writing. It does not suit every programme. Below a few thousand parts a year in a single process, tooling and setup dominate and there is little coordination cost to remove in the first place. It makes no certification claims. Quality systems are confirmed per programme and per factory in writing before production rather than asserted on a web page. How to test it with one project The sensible way to evaluate any consolidation is to run one programme through it and compare the two totals, not two quotations. Pick a job that spans at least two processes - a stamped bracket that needs finishing, a machined housing that needs a moulded insert, a sheet metal enclosure with a welded frame and a coated panel. Send the whole drawing set at once rather than one part at a time, and ask for the quotation with the process route shown against each operation. Then count what you did not have to do: the purchase orders you did not raise, the inbound inspections you did not run, the transit legs you did not pay for and the emails you did not send. Four hand-offs replaced by one route. See metal stamping and plastic injection molding for the two highest-volume processes in the network, and surface finishing for the step that most often creates the hidden hand-off. Send a multi-process drawing set for one quotation Scope and sources. The hidden-cost breakdown (inter-vendor logistics 6 percent, rework and transit damage 5 percent, per-vendor audits 4 percent, administration 3 percent, expedited freight 2 percent, about 20 percent in total), the 32 percent per-part consolidation result, the 35-kiosk lead-time case and the queue-time mechanism were compiled in 2026 from a published comparison of fragmented and consolidated sheet metal sourcing. The multi-source premium of 3 to 5 percent of procurement spend, the volume threshold of about USD 200,000 to 500,000 per category, the on-time delivery movement from 83 to 94 percent, the disruption frequency change and the 5 to 15 percent inter-factory defect variation come from a review of multi-source strategy data that cites a 2025 American Chamber of Commerce in South China survey, and from a sourcing comparison of single and multiple supplier models. These are commercial case results and planning ranges published by other parties, not audited benchmarks or guaranteed outcomes. Nothing here states or implies a certification held by any factory; quality systems are confirmed per programme and per factory in writing before production.

The short answer Both sinter nylon powder without supports. MJF fuses a whole layer with an infrared lamp and printed agents, SLS traces it with a laser. That makes MJF faster in batches, isotropic to 85 to 90 percent of XY strength, accurate to ±0.2 mm and smoother; SLS offers a wider material range and larger build envelopes. Same physics, different heat source Both processes are powder bed fusion and both are support-free: the unfused powder around the part holds it up, so undercuts, lattices and nested parts cost nothing extra. The difference is how the heat is delivered. SLS scans a laser across each layer, point by point along the part's cross-section, so the melt pool is local and cools directionally. MJF prints a fusing agent where the part should be solid and a detailing agent at the boundaries, then passes a broad infrared lamp over the whole bed, so every part of the layer is fused at once. Almost every difference in this answer follows from that one fact. Because MJF heats the whole layer together, the temperature gradient across a part is smaller, the part is less directional, and the build is faster when the bed is full of parts. Because SLS delivers energy through a focused spot, the resolution is limited by the spot size, the cooling is directional, and the same machine can be pointed at a much wider range of polymers, including grades that MJF does not offer. Mechanical properties: closer than the marketing suggests On standard PA12 the two are in the same band. Typical tensile strength is about 45 to 50 MPa for SLS and 48 to 52 MPa for MJF, so the difference is smaller than the scatter between powder lots. Elongation runs the other way: SLS is usually the more ductile of the two at roughly 15 to 25 percent against 10 to 20 percent, which matters for snap fits and parts that must absorb impact rather than resist it. The number that actually decides structural parts is anisotropy. In SLS the laser's directional cooling leaves the Z axis noticeably weaker, typically 80 to 85 percent of in-plane strength. MJF, fusing the layer evenly, reaches about 85 to 90 percent. Neither is truly isotropic, and both are far better than FDM, where the Z axis can be a fraction of XY. If the part is a bracket loaded across the build direction, that five-point gap is worth checking against the safety factor before choosing on surface finish. Neither process is isotropic. The question is whether your load path runs across the layers. Accuracy, minimum feature and minimum wall Tolerance on both is usually quoted as a combination of an absolute figure and a percentage, and the percentage takes over on large parts. SLS is commonly quoted at plus or minus 0.3 mm or 0.3 percent, whichever is greater; MJF at plus or minus 0.2 mm or 0.2 percent. On a 30 mm feature that is 0.3 against 0.2 mm, a difference you can feel in an assembly. On a 300 mm part both are dominated by the percentage term and by how the part is oriented and cooled in the cake, so the gap narrows. Feature size separates them more clearly. SLS resolves detail down to roughly 0.3 to 0.5 mm because the laser spot sets the limit, and a reliable minimum wall is around 0.8 to 1.0 mm. MJF, with the detailing agent sharpening the boundary, resolves about 0.2 mm and holds walls down to roughly 0.5 mm, which is why it can print fine embossed text, small holes and thin lattice struts that SLS rounds off. If the design depends on 0.5 mm ribs or 1 mm text, that single difference decides the process. Surface finish and the post-processing load Both come out grainy, and both are bead blasted as a standard step. The difference is how much work the blast has to do. SLS has the characteristic sandy, orange-peel surface, with roughness typically in the region of 150 to 300 microinches, and often needs more than a light blast before it can be dyed or painted evenly. MJF comes out noticeably smoother, around 100 to 250 microinches, with crisper edges, and takes dye more uniformly. That difference has a cost consequence that does not appear in the machine rate. Finishing is labour, and a part that needs less of it is cheaper to deliver even when the printing itself is priced higher. For a batch of cosmetic covers the MJF premium can disappear entirely in the finishing line; for a batch of internal brackets that will never be seen, it is pure added cost. The last two rows are the ones that rule a process out. Everything above them is a preference; those are limits. Throughput and cost per part MJF wins on batches because the lamp fuses the whole layer at once. A representative comparison for fifty small nylon brackets is roughly 8 to 12 hours on MJF against 16 to 24 hours on SLS, and the gap widens as the bed fills, because MJF's time per layer barely changes with the number of parts on it while the laser has more outline to trace. That is why MJF quotes fall faster with quantity and why the same part can be cheaper on one process at ten pieces and on the other at five hundred. Powder economics matter too. Both processes blend used powder with fresh, but the ratios differ: SLS commonly runs 30 to 50 percent fresh, while MJF machines typically need around 30 percent fresh for PA12, with better consistency across reuses because the powder sees less thermal exposure. On an expensive nylon that refresh ratio is a real cost per part, and it is one of the numbers a supplier should be able to state rather than estimate. Material range and build envelope SLS has the wider palette. Beyond PA12 and PA11 it runs glass-filled, carbon-filled, aluminium-filled, flame-retardant and flexible blends, and it is the practical choice when the part needs a property that standard nylon does not have. MJF is more constrained, centred on PA12, PA11, glass-bead PA12 and TPU, with fewer third-party materials. If the specification names a flame-retardant or carbon-filled grade, the decision is often already made. The envelope runs the other way. SLS platforms go considerably larger, with build volumes up to about 750 by 750 by 550 mm on the biggest machines, while MJF's common platform is around 380 by 284 by 380 mm and large-format MJF is rare. A part that is 500 mm long, or a batch that only makes sense if it is nested in one tall build, may be an SLS job regardless of which process is better on the numbers. Part archetypeBetter fitWhy cosmetic cover, few hundred offMJFsmoother as-built, less finishing fine text and small holesMJF0.2 mm detail against 0.3 to 0.5 snap fits and impact clipsSLS15 to 25 pct elongation flame-retardant or filled gradeSLSwider powder palette part longer than 400 mmSLSbuild envelope up to 750 mm one prototype this weekeither, ask the queueneither route needs tooling The last row is not a dodge. For a single part the process difference is usually smaller than the difference between a supplier with a build running tomorrow and one that has to wait for a full bed, so ask for the next build date before arguing about isotropy. The decision, in four questions First, how many parts and how soon: a batch that has to ship this week usually goes to MJF, a single large part usually goes to SLS. Second, does the design depend on fine detail: text under 1 mm, holes under 1 mm, walls under 1 mm point to MJF. Third, does it need a resin other than nylon: flame-retardant, carbon-filled or elastomeric blends point to SLS. Fourth, does it fit: anything beyond roughly 380 mm in the MJF envelope has to go elsewhere. If all four answers are neutral, choose on finishing cost and on isotropy, in that order, because those are the two differences that persist into the delivered part. One caution on comparisons: published property tables are usually tested on optimally oriented specimens, and neither process is isotropic. Ask for the orientation used, and if the part carries load across the build direction, derate the Z figures rather than the XY ones. The order matters. A large part fails the last question even if it wins the first three. Both processes sit inside the 3D printing capability, with powder parts listed under SLS parts; where the part needs metal rather than nylon, CNC machining is the route to compare against. Send the model, the quantity and the load direction, and we will quote SLS and MJF on the same drawing with the finishing included Scope and sources. The comparison of tensile strength at 45 to 50 against 48 to 52 MPa, elongation at 15 to 25 against 10 to 20 percent, Z strength at 80 to 85 against 85 to 90 percent of XY, tolerance at 0.3 against 0.2 mm and the corresponding percentages, minimum detail of 0.3 to 0.5 against 0.2 mm, minimum wall of 0.8 to 1.0 against 0.5 mm, build volumes, powder refresh ratios of 30 to 50 against about 30 percent, and the surface descriptions are from a comparison of SLS and MJF for nylon 12. The mechanism of each process, the near-isotropic claim for MJF and the material lists are from a multi jet fusion technical guide and a process comparison. The build time example of 8 to 12 hours on MJF against 16 to 24 hours on SLS for fifty small brackets, the roughness bands of 100 to 250 and 150 to 300 microinches and the build size of 380 by 284 by 380 mm for the common MJF platform are from an MJF service reference. Property values are typical published figures for PA12 and vary with powder lot, orientation and machine; they are planning values, not a specification.

The short answer Every process needs its own mandatory step: support removal on FDM and SLA, powder removal on SLS and MJF, solvent wash and UV cure on resin. Beyond that, bead blasting, vapour smoothing, dyeing, epoxy sealing, annealing, inserts, machining and plating are chosen to match the part's job. Two buckets: mandatory and chosen Post-processing is quoted badly because it is quoted as one line. In practice it splits into work that cannot be skipped without the part being wrong, and work that is chosen to meet a specification. The mandatory list is short and process-specific: supports and rafts have to come off an FDM or resin part and the witness marks dressed; loose powder has to come out of an SLS or MJF part, including internal channels; a resin part has to be washed of uncured resin and then cured under UV to reach its stated mechanical properties. Skip any of those and the part is not a cheaper version of the finished item, it is a different and usually unusable item. The chosen list is where the money and the arguments are: bead blasting, sanding and polishing, vapour smoothing, dyeing, painting, epoxy sealing, annealing, thread inserts, secondary machining, plating and, for metal, hot isostatic pressing. Each one buys a specific property and each one has a side effect on dimension or surface. The useful discipline is to name the property first, cosmetic, colour, strength, tolerance or conductivity, and then pick the shortest route to it rather than the longest menu. If the part mustThe sequence isWhat to watch look mouldedsmooth, prime, paint or texturebuild-up on mating faces hold a colourdye the nylon, or paintdye lot-to-lot match resist wear or fluidseal, plate or coatadded thickness, masking carry loadanneal, then machine critical facesshrinkage after annealing take a threadinstall a heat-set insertboss wall thickness hold a toleranceprint oversize, machine afterfixture and datum choice Supports, rafts and loose powder FDM and resin processes build on supports, and the marks they leave are the first thing a buyer notices. Breakaway support on FDM leaves a rough patch that has to be sanded through a grit progression, typically from 80 up to 2000 for a cosmetic surface, and soluble support removes the problem inside cavities at the cost of a wash station and time. Resin parts printed with supports on a cosmetic face will show witness pits no matter how carefully the support is placed, so the orientation decision is really a finishing decision made at the quoting stage. Powder-bed processes have the opposite problem: nothing to break off, but powder everywhere. Depowdering is blowing and blasting loose powder out of the part, and on SLS and MJF it is followed by bead blasting to even out the surface and remove the last of the cake. That blast is also what makes the part paintable without a primer, because it gives a uniform matte key. Parts with internal channels need the powder removed before any coating, or the coating seals the powder in. The middle column cannot be skipped. The right-hand column is where the specification and the cost live. Resin parts: wash, cure and the tacky failure A stereolithography or DLP part straight off the platform is not fully cured. It comes out coated in uncured resin, and its mechanical properties are a fraction of the datasheet until it has been washed in solvent, usually isopropyl alcohol, and post-cured under UV. A part that is wiped rather than washed, or cured for too short a time, stays tacky, creeps under load and can be chemically reactive on skin. That is not a cosmetic defect; it is a part that will fail a functional test and, in a medical or consumer context, fail a safety review. The practical consequence is that resin printing has a fixed labour content per batch that does not scale down with part size. Small parts in a full build are cheap per unit; one large part in a half-empty build carries the whole wash and cure cost. When comparing a resin quote with an SLS or MJF quote, ask whether wash and cure are inside the price or billed as a separate operation, because that single line often explains a gap that looks like a material or machine difference. Vapour smoothing: what it buys, what it moves Solvent vapour dissolves the outer skin of an ABS or ASA part so that the layer valleys flow together. Done properly it takes a visibly layered FDM surface to something close to a moulded finish, with roughness down to about 0.8 micrometres, in roughly twenty minutes, and it improves inter-layer bonding at the same time, which is why impact strength can rise by 10 to 25 percent. It is the single most effective cosmetic step available for a thermoplastic print. It is not dimensionally neutral. The process removes something like 20 to 50 micrometres of surface, rounds sharp edges, and moves dimensions by roughly 0.05 to 0.15 mm depending on exposure. Any feature that must hold a tolerance after smoothing has to be either masked, machined afterwards, or printed oversize. Vapour smoothing is therefore the wrong last step on a part with a press fit; it is the right step on a cover or a handle where appearance matters and fits are loose. Build-out and soak-in are different problems. A coating grows a part; a dye or a vapour changes it without adding material. Colour: dyeing against painting Nylon powder parts take dye exceptionally well because the sintered surface is porous. Dyeing in a hot bath penetrates 0.3 to 0.5 mm into the surface, so the colour cannot chip or wear off the way paint can, and colour matching to a reference is achievable to within a small colour difference. That makes dyeing the default for functional parts, jigs, housings and anything that will be handled. The limits are that it works best on light-coloured nylon, deep blacks and saturated reds are harder, and the process adds moisture that has to be dried out before the part is measured. Painting gives colour plus surface in one step, and it can hide layer lines that dye cannot. The cost is thickness: a primer and a topcoat add tens of micrometres and can bridge small holes, fill engraved text and change a press fit. It is also the right choice when the colour must match an existing product exactly, or when the part needs a gloss, metallic or soft-touch effect that a dye cannot produce. Functional finishing: annealing, inserts, machining, plating Annealing is the cheapest strength upgrade for an FDM part. Holding at 80 to 120 degrees Celsius for one to four hours relieves the internal stress locked in by extrusion; annealed PETG shows 15 to 20 percent better impact resistance and less creep under sustained load. The side effect is shrinkage, roughly 0.3 to 0.5 percent linear for nylon, which has to be allowed for before printing or removed by machining afterwards. For metal parts the equivalent step is hot isostatic pressing, which closes residual porosity and can add up to 40 percent fatigue life and 30 percent elongation at break, at the cost of several days of lead time. Threads are the other common functional finish. A printed thread in nylon is weak and wears; a heat-set brass insert gives a reusable thread that survives repeated assembly, provided the boss around it has enough wall, usually at least twice the insert diameter. Secondary machining covers everything printing cannot hold: a bearing bore, a sealing face, a flat that has to sit square. And plating, applied over an electroless copper seed and then built up electrolytically, turns a plastic part into a conductive or metallic-looking one for shielding cans, RF enclosures and display models. How to specify it in a quote Do not write "finished part". Write the property and the measurement: surface roughness or a comparison sample, colour to a reference with the acceptable difference, the dimensions that must be held after finishing, and whether the finish is inside or outside the quoted tolerance. State whether the part will be measured before or after finishing, because on a smoothed or painted part those are two different parts. And ask which operations are done in-house: the split between printing and finishing is where lead times and quality arguments usually start. Answer them in order. A part that needs colour and a tight tolerance will usually be dyed first and machined second. The processes behind these finishes are described on the 3D printing page, with resin parts under SLA printing and powder parts under SLS printing. Send the model with the finish, the colour reference and the dimensions that must survive finishing, and we will quote the printing and the finishing as separate lines Scope and sources. The split between mandatory steps and chosen operations, and the list of operations by material, are from a post-processing and finishing guide. The vapour smoothing figures of 20 to 50 micrometres removed, roughness down to 0.8 micrometres in about 20 minutes, impact strength up 10 to 25 percent and dimensional movement of 0.05 to 0.15 mm are from a post-processing techniques guide, as are the epoxy build of 0.3 to 0.5 mm per side, the annealing window of 80 to 120 degrees Celsius for one to four hours, the 15 to 20 percent impact gain on annealed PETG, the 0.3 to 0.5 percent shrinkage on annealed nylon, the dye penetration of 0.3 to 0.5 mm, the colour difference tolerance, the 80 to 2000 grit sanding progression, the 40 percent fatigue and 30 percent elongation gains from hot isostatic pressing, and its 10 to 14 day lead time. The bead blasting step for powder parts and its role as a paint key are from the same guide. Ranges are typical workshop values and will vary with geometry, exposure time and resin grade; they are planning figures, not a specification.

The short answer A two-plate mould opens at one parting line, gates at the edge and ejects the part with its runner. A three-plate mould adds a floating runner plate and a second parting line, so a pin-point gate can sit anywhere on the face and the runner drops away unaided, for 15 to 40 percent more tool cost. One plate changes everything The two-plate tool is the default because it is the simplest thing that works. A top clamp plate, a cavity plate, a core plate, a support plate and an ejector housing; the sprue comes down the centre, a runner cut into the parting plane carries the melt to each cavity, and a gate at the edge of the part lets it in. When the tool opens, the part, the runner and the sprue stay together on the moving half and are pushed out as one piece. Every consequence that follows, cheap tooling, short stroke, fast cycle, and a runner that has to be cut off by hand, comes from that single choice of where the runner lives. The three-plate tool moves the runner to its own plane. A runner stripper plate sits between the top clamp plate and the cavity plate, the runner is cut between the stripper plate and the cavity plate, and a narrow drop passes through the cavity plate to a pin-point gate on the face of the part. Now the gate is not tied to the parting line at all: it can be in the centre of a round cover, at several points across a large flat panel, or on a hidden internal face. The price is an extra plate, longer guide pins, puller bolts and latch locks to sequence the opening, and more daylight in the press to let three plates separate. How a three-plate tool opens The opening is a controlled sequence rather than a single motion. First the cavity plate separates from the runner stripper plate, which is what breaks the pin-point gates and leaves the runner behind on the stripper side. Then the stripper plate moves off the top clamp plate and the runner is pulled off its sucker pins so it falls clear, or is taken by a sprue picker. Only then does the main parting line open and the ejector system push the finished parts off the core, already degated. The ordering is enforced mechanically, so a worn latch or a stretched puller bolt does not just slow the tool down, it changes the sequence and produces short shots or runner damage. That mechanical sequence is also the source of the maintenance difference. A two-plate tool has one parting surface to keep parallel and one set of guides. A three-plate tool has two, plus latches, stop bolts and sucker pins that all have to be checked on a schedule, and a middle plate that can wear at its guides. None of that is difficult, but it is work that a two-plate tool does not need. The gate breaks on the first opening, before the part moves. That is why the part arrives degated with no trimming station. Gate position is the real decision Most of the argument about plate count is really an argument about where the plastic enters the part. A two-plate tool can use an edge gate, a fan or tab gate on a flat part, or a submarine gate that runs under the parting line and shears off as the part is ejected. That last one gives automatic degating on a two-plate tool, which is why the "three-plate means automatic" shorthand is too simple: a well-placed submarine gate solves a lot of parts at the lower tool cost. Gate typeToolWhere it entersDegatingBest for Edge or sidetwo-platepart perimetermanual trimsimple parts, tight budget Fan or tabtwo-platewide edgemanual trimflat panels, low warp Submarine or tunneltwo-plateunder the parting lineautomatic on ejectionsmall parts, hidden gate Pin-pointthree-plateanywhere on the faceautomatic on openingcentre gating, cosmetics Direct spruetwo-platetop centremanual trimsingle cavity, deep parts Hot tip or valvehot runneranywhere on the facenone neededhigh volume, engineering resin The rule that follows is not "three-plate is better". It is: if a submarine gate placed near a corner gives acceptable fill and puts the weld line where you can live with it, the two-plate tool is almost always the lower total cost, in tooling and in cycle time. Choose three-plate when the gate genuinely has to be central, when several gates are needed across a large or asymmetric face to balance flow, or when a multi-cavity runner cannot otherwise reach every cavity with similar flow length. What the extra plate costs Tool cost is the first line and it is well documented: quoted premiums over a comparable two-plate tool run from about 15 to 30 percent at the low end to 20 to 40 percent, depending on part size, cavity count and how much sequencing hardware the design needs. Cycle time moves the other way from what buyers often assume: the extra opening stroke and the thicker runner that has to cool add time, so the three-plate tool is usually slightly slower per cycle even though it saves labour downstream. The press also needs more daylight, which matters on older machines with limited opening stroke. Runner scrap is the hidden line. In a cold-runner three-plate tool each shot carries the sprue, the main runner and the drops through the cavity plate, so the shot weight is higher than in a two-plate tool of the same cavitation. That material is either reground and blended back within the limits the resin allows, or sold as scrap. Either way it is a recurring cost that scales with every shot, unlike the tool premium which is paid once. The runner and degating rows are the ones that repeat every cycle. The tool cost row is paid once. The runner is material, not just scrap A worked example makes the trade visible, and it is arithmetic rather than a quotation. Take a two-cavity tool running a 45 g part with a 55 g cold runner in the two-plate version. Move to a three-plate layout with drops and the runner grows to about 70 g per shot. At 3.00 US dollars per kilogram of resin, the extra 15 g per shot is about 0.045 dollars per shot, or roughly 9,000 dollars over 200,000 shots, before any credit for regrind. Set that against a tool premium of, say, 8,000 dollars on a 40,000 dollar tool and the three-plate route is not automatically the expensive one; it depends entirely on the shot count and on how much of the regrind the customer's specification allows back into the part. Two cautions on that arithmetic. Regrind is not free material: it carries a handling and drying cost, it degrades with each pass, and many specifications cap it or forbid it in cosmetic, food-contact or structural parts, which turns the runner into pure scrap. And the runner weight is a design variable: a well-laid-out runner in either tool can often be shortened, and a hot runner removes it altogether. When three-plate is the wrong tool Three-plate tooling is a poor fit for very large flat parts, where the middle plate can deflect under clamp force. It is a poor fit at low volume, where a 15 to 40 percent tool premium on a few thousand parts never amortises against saved trimming labour. It is a poor fit for shear-sensitive or heavily filled resins, because the pin-point gate is deliberately tiny and raises shear at the gate; the same small gate freezes early, which limits packing pressure into thick sections. And it is a poor fit where press daylight is already tight, since the third plate needs room to open. There is also a design cost that rarely appears in the comparison tables. A pin-point gate leaves a small circular vestige on the face of the part, and a drop leaves a marks where it passed through the cavity plate. On a Class A surface those marks have to be placed deliberately, or hidden behind a label recess, a rib or a mating part. A three-plate tool gives you gate freedom but not gate invisibility. Hot runner is the third structure A hot runner keeps the melt in a heated manifold and nozzles at processing temperature, so there is no runner to eject and no runner scrap at all. It gives the gating freedom of a three-plate tool with a shorter cycle, typically 10 to 30 percent faster because the thermal mass of the runner no longer has to be cooled, at a tool premium of roughly 30 to 60 percent plus a temperature controller. It earns its place at high volume, with expensive resins where regrind is unacceptable, or where gate vestige has to be minimal. Below a few hundred thousand parts a year, the arithmetic usually favours a cold runner. Three structures, one axis. The premium is paid once; the runner cost is paid on every shot. The process these tools belong to is described on the injection moulding page, the parts they produce are listed under plastic injection moulding, and where a tool is not justified at all, CNC machining covers the low-volume route. Send the part drawing, the annual volume and where the gate mark is allowed, and we will price the two-plate, three-plate and hot-runner options side by side Scope and sources. The plate stack, the three-step opening sequence, the gate types that suit each structure and the note that a hot runner offers three-plate gating freedom without runner scrap are from a two-plate and three-plate mould comparison. The tool cost premium of about 15 to 30 percent, the longer cycle from the extra stroke and the thicker runner, and the comparison of runner scrap per shot are from a mould structure reference. The premium of 20 to 40 percent, the elimination of manual degating, the cycle reduction of 15 to 30 percent for hot runners and the note that hot runners are standard above roughly 500,000 parts a year are from an injection mould guide. The gate location limits, two parting lines and the additional daylight requirement are from a two-plate and three-plate FAQ. The 9,000 dollar runner calculation is arithmetic on the stated weights and resin price, not a quotation; the runner weights are illustrative and should be taken from the actual layout.

The short answer Most moulded parts land between 1.5 and 3.0 mm: ABS 1.5 to 3.5 mm, PC 1.5 to 3.0, 30 percent glass-filled PA66 1.2 to 3.0. Keep the wall uniform, taper any change at 3:1, and hold ribs at 40 to 60 percent of the wall. Thin walls fail on flow length, thick ones on sinks. Why the wall is the first number you fix Wall thickness sets three things at once, and only one of them is strength. It sets how far the melt has to travel before the cavity is full, it sets how long the part has to stay in the tool before it is stiff enough to eject, and it sets how much material shrinks where. Because cooling time rises faster than thickness, roughly with the square of it, a wall that is twice as thick does not take twice as long to cool; it takes several times as long, and that time is paid on every shot for the life of the tool. This is why thickness is usually the first line of a DFM report and the last thing a designer wants to change. The second effect is dimensional. A thick section cools from the outside in: the skin freezes while the core is still molten and shrinking, and the difference has to go somewhere. In a soft unfilled resin it pulls the surface in and leaves a sink mark; in a stiff or glass-filled resin it leaves an internal void instead. Uneven walls produce the same argument at a larger scale, because a thin section freezes before the adjacent thick one and the two shrink by different amounts. The result is warp, and warp is a tolerance problem that no amount of inspection sorts out after the fact. A working range by resin The numbers below are starting points for a normal part with a moderate flow path, not limits. They move with glass content, flow length, gate position and the mould's cooling design, and the resin data sheet plus a flow simulation always outrank a table. ResinCommon nominal wallPractical floorWhat usually goes wrongCommercial tolerance ABS1.5 to 3.5 mm0.75 to 0.8 mmsink over bosses and ribsplus or minus 0.10 mm Polycarbonate1.5 to 3.0 mm1.0 mmhigh viscosity, stress marksplus or minus 0.12 mm PA66 with 30 pct glass1.2 to 3.0 mm0.8 mmwarp from fibre orientationplus or minus 0.15 mm PP1.0 to 2.5 mm0.6 to 0.9 mmhigh shrinkage, warpageplus or minus 0.15 mm POM1.0 to 3.0 mm0.5 to 0.8 mmshrinkage along flow directionplus or minus 0.10 mm PEEK1.5 to 3.0 mm1.0 mm and upneeds hot tool, high meltplus or minus 0.10 mm The tolerance column is there because thickness and tolerance are usually specified on the same drawing and priced together. A commercial grade of plus or minus 0.10 mm and a precision grade of plus or minus 0.025 mm are not a small step in tool cost; on a part with features on both faces they can be the difference between a two-plate tool and one with a ground, re-worked cavity. Put the tight tolerance only on the dimensions that touch another part. These are floors, not design targets. A part drawn at the floor has no margin for a longer flow path or a colder tool. Uniformity beats thinness Given the choice between a part with a 2 mm wall everywhere and a part with 1.2 mm walls and a 4 mm boss, the first is easier to mould and usually cheaper, even though it weighs more. That is counter-intuitive until you watch a tool run: the thin part fills faster but warps, the thick boss stays molten after the rest has frozen and has to be packed, and the cycle is set by the slowest section to cool, not the fastest. Material saved in the wall is often paid back in cycle time and scrap. When a thickness change is unavoidable, taper it. The standard rule is a transition length of about three times the thickness difference, so a step from 3 mm to 2 mm should spread over roughly 3 mm of length rather than happening at a shoulder. A gradual taper keeps the melt front moving as one front; a step creates a hesitation effect where the thin side freezes first and the thick side over-packs, and it concentrates stress exactly where the part will later crack. Ribs, bosses and the 40 to 60 percent rule A rib is a way to buy stiffness without buying thickness, and it only works if the rib is thin enough at its base. The usual band is 40 to 60 percent of the adjacent nominal wall: 60 percent is safe in low-shrinkage resins such as polycarbonate, while high-shrinkage grades such as PP and PE want the bottom of the band. A rib drawn at the full wall thickness is a thick section in disguise and will telegraph a sink mark onto the opposite face, which on a cosmetic part is a scrapped moulding. The rest of the rib geometry follows the same logic. Height is usually capped at about three times the wall, because a taller rib is hard to fill and hard to vent at the tip. Spacing of at least twice the wall keeps enough steel between ribs for cooling channels and prevents a weak, hot core. A base radius of 0.25 to 0.5 of the wall reduces the stress concentration at the root and improves flow into the rib. Bosses are the same problem in cylindrical form: an outer diameter of about 2.5 times the screw's major diameter, wall thickness at 60 percent of the nominal wall or less, and a connection to the side wall by ribs or gussets rather than standing as an isolated column of hot plastic. Each rule protects a different failure. Break the thickness rule and you get a sink; break the spacing rule and you get a hot, weak core. Thin walls and the flow-length limit A thin wall is not simply a scaled-down version of a thick one, because the melt has to travel further relative to the section it is travelling through. The working measure is the flow path divided by the wall thickness. Most unfilled grades fill a ratio of roughly 100 to 200 before injection pressure runs out; glass-filled grades manage less because the fibres raise viscosity, and a long, thin part may need a second gate rather than a higher pressure. When the ratio is exceeded, the failure mode is predictable. The part short-shots at the far end, or it fills but the last region to arrive is over-packed and stressed, which shows up as a weld line, a gloss difference or a crack in service. Raising melt temperature and pressure buys a little; moving the gate buys a lot. In thin-wall packaging the ratio is pushed far past these numbers on purpose, with high-speed injection and hot runners, but that is a different tool and a different machine, not a tweak to a standard mould. When a thick section is the right answer Some parts need mass, and pretending otherwise causes more problems than it solves. A section that carries a thread, a bearing, a press-fit insert or a wear surface needs enough material around it to take the load without cracking, and that is often 3 mm or more regardless of what the wall average says. The same applies to a part that must resist impact at low temperature, where a thin wall in a brittle resin will shatter rather than deform. The correct response is to localise the thickness rather than spread it. Core out the middle of a thick boss and keep the walls, add ribs instead of a solid web, and place the thick section where it can be fed and cooled rather than in the middle of a large flat face. If the thick region cannot be avoided, expect a longer cycle and plan for it in the piece price rather than discovering it at first article. What to put on the drawing State the nominal wall, the tolerance on it, and the rule for changes. A note such as "nominal wall 2.0 mm plus or minus 0.10 mm; all thickness transitions 3:1" tells the toolmaker what the design intends and gives the inspector something to measure. Dimension the wall at a defined plane, not across a drafted face, and keep the critical dimensions on a separate list rather than applying a tight block tolerance to everything. Wall first, ribs second. A rib sized against an unsettled nominal wall will be resized when the wall moves. The process context for these numbers is on the injection moulding page, the parts made with them are listed under plastic injection moulding, and where a wall cannot be moulded thin enough, CNC machining covers the machined alternative. Send the part and its annual volume, and we will return a wall and rib review with the cycle-time effect of each change Scope and sources. The common design ranges and practical floors for ABS, polycarbonate, nylon, POM, PP and PEEK, the 40 to 70 percent rib guidance and the worked example of a 2 mm wall with a 6 mm boss are from a wall thickness and rib design reference. The 3:1 transition taper, rib base radius of 0.25 to 0.5 of the wall, boss outer diameter of about 2.5 times the screw diameter and the rib height limit of three times the wall are from a moulding design guide. The commercial and precision tolerance classes of plus or minus 0.10 and plus or minus 0.025 mm, the nominal wall bands of 1.5 to 4.5 mm and the rib ratio of 40 to 60 percent are from a design for injection moulding guide. The flow-length-to-thickness ratio of roughly 100 to 200 for unfilled grades is a working planning figure rather than a specification and should be confirmed by flow simulation for the actual geometry. Thickness and tolerance figures are planning values, not a quotation.

The short answer Most moulded walls need 1 to 2 degrees of draft per side, ribs and bosses 0.5 to 1 degree. Add 1 to 1.5 degrees for every 0.025 mm of texture depth, which takes a typical 0.06 to 0.08 mm grain to 3 to 5 degrees. Core-side features need more than cavity-side ones. What draft actually does during ejection A cooling part shrinks. On the male side of the tool, the core, it shrinks onto the steel and grips it; on the female side, the cavity, it pulls away. That asymmetry is the whole reason draft is not a cosmetic detail. With a vertical wall the part stays in contact with the core for the entire ejection stroke, so the ejector pins have to overcome friction over the full depth of the part. With 1 to 2 degrees of taper the wall separates from the steel within roughly the first tenth of a millimetre of travel, and the rest of the stroke runs with clearance. The consequences of getting it wrong are visible before they are dimensional. Drag marks and scuffing appear on the wall, the surface stress-whitens where the plastic was stretched, ejector pins leave a witness ring or punch through a thin floor, and the part distorts as it leaves. On a textured face the damage is worse, because the texture acts as a field of micro-undercuts and a dragging part shears the grain off the tool. The repair is not a process tweak: it means re-machining or re-texturing steel. Draft is measured from the direction of pull, not from the parting line, and it is always quoted per side. That convention matters when a wall has draft on both faces: a 1 degree per side taper on a 4 mm wall closes the section by about 0.14 mm over 4 mm of depth, and a supplier quoting "1 degree total" is describing half of what the design needs. The baseline by surface condition Start from the finish, not from the resin. A polished cavity releases easily and 1 degree per side is a workable minimum for shallow parts in stable engineering resins, with 2 degrees the safer default for anything with depth. A fine matte or light etch needs 1.5 to 2 degrees. A medium leather or sand grain needs 2 to 3 degrees. A heavy grain or deep chemical etch needs 3 to 5 degrees, and the deepest patterns run beyond that. Ribs and bosses are the exception in the other direction: they are narrow, they grip the core hard, and 0.5 to 1 degree per side is the accepted minimum, with the taper running from the base toward the tip on core-side features. Depth of draw0.5 deg per side1 deg per side1.5 deg per side2 deg per side 10 mm0.09 mm0.17 mm0.26 mm0.35 mm 25 mm0.22 mm0.44 mm0.65 mm0.87 mm 50 mm0.44 mm0.87 mm1.31 mm1.75 mm 100 mm0.87 mm1.75 mm2.62 mm3.49 mm The table is the number most designers never calculate: the side offset the draft consumes at the far end of the feature. It is plain trigonometry, offset equals depth multiplied by the tangent of the angle, and it explains why deep ribs and tall bosses run out of material at the tip. A cosmetic wall that is 2.4 mm at the parting line with 1.5 degrees of draft over 50 mm is 0.9 mm thinner at the other end, which is no longer the wall that was analysed. If a feature has to stay thick at the tip, either reduce its depth or accept less draft and pay for it in ejection. One part carries several of these at once. The value that governs is the one on the face with the coarsest texture and the deepest draw. Texture is the largest single adder The industry rule is to add 1 to 1.5 degrees of draft for every 0.025 mm, or 0.001 inch, of texture depth. Typical moulded-in grains are 0.06 to 0.08 mm deep, so the adder works out at 2.5 to 3.75 degrees on top of whatever the smooth face needed. A face that would have run at 1 degree polished therefore lands at 3.5 to 4.75 degrees once it is grained. A deeper industrial pattern at 0.1 mm would add 4 to 6 degrees by the same arithmetic, which is why heavy textures and tight wall positions fight each other. Two practical consequences follow. First, the texture specification has to be settled before the tool is cut, because the draft that the grain requires changes wall positions, rib tips and mating faces. Adding a grain after the tool exists usually means re-cutting walls or accepting a drag mark. Second, the texture supplier and the mould maker should see the same number: the depth in the texture standard, not the pattern name. A pattern quoted as "medium leather" by one supplier is not necessarily the same depth as another's. Core side, ribs and shut-offs carry different numbers Because the part shrinks onto the core, core-side faces need more draft than cavity-side faces at the same texture. The working adjustment is 0.5 to 1 degree extra on core-side features beyond the cavity baseline, and more on deep cores where cooling shrinks the part tighter. Cored holes deserve their own rule: a perfectly cylindrical core will grip the part, so 0.5 degree per side is the practical minimum even when the surrounding wall has 2 degrees. Shut-off faces are a separate case again. Where cavity and core steel meet metal-on-metal to blank off a hole or an opening, the faces slide against each other under clamp pressure. A 1 degree angle is too shallow here and will gall, wear and eventually flash, so 3 degrees is the usual minimum. Ribs and bosses take 0.5 to 1 degree per side, with the draft measured base to tip on core-side features and tip to base on cavity-side ones; getting that direction wrong creates an undercut that a slide then has to solve. Read the band as the minimum, not the target. Where a face is both textured and deep, the two adders stack. Depth changes the answer more than material does Resin choice matters less than most design guides imply. Soft unfilled grades such as PP and PE release more easily and tolerate the low end of each band; rigid and filled grades such as PC, ABS and glass-filled nylon grip harder and need the middle or top. But a 5 mm deep feature in the easiest resin still needs less draft than a 100 mm deep one in the most difficult, because the contact area and the shrink grip both scale with depth. Depth is the variable to check first. The usual correction for a deep draw is not more draft on the whole part, which would move every wall, but a local increase: steeper taper on the deep feature only, a polish in the draw direction to cut friction, or a redesign that splits the depth. On cores forming deep pockets, additional draft plus better core cooling, which reduces shrink grip, does more than another half degree of taper alone. When a face has to stay vertical Some faces genuinely cannot be tapered: a sealing land, a mating face that must sit square against another part, an optical window, a bore that takes a bearing. Moulding them at zero draft is possible and it is not free. The tool then needs draw-direction polishing, tighter steel tolerances, often a sleeve ejector rather than pins, air poppet valves to break the vacuum, or extra wall thickness so the shrink grip is lower. Every one of those is a line in the tooling quote, and a face marked "must stay vertical" without a reason will be priced conservatively because the supplier has to assume the worst. There is one more option worth knowing: stripping. A small undercut or a zero-draft lip can sometimes be forced off the core by the flexibility of the resin itself, which avoids a slide. That works on unfilled, ductile grades for features up to roughly 2 to 3 percent of the part diameter, provided the edge is rounded. It does not work on polycarbonate or glass-filled nylon, which are too stiff and will crack or whitening instead. How to put it on the drawing Write the default, then write the exceptions. A single note such as "all vertical faces 1.5 degrees per side unless dimensioned otherwise" gives the toolmaker a baseline, and marking the faces that must stay vertical, with the functional reason beside them, prevents a late argument. Give the texture standard and its depth on the same drawing rather than in a separate cosmetic document, because the draft requirement depends on it. Two habits avoid the common failures. Do not dimension across a drafted face as though it were parallel; give the dimension at a stated plane, usually the parting line. And check both halves of the tool: a wall drafted correctly on the cavity side but left vertical on the core side will still drag, and it is the core side that the part is gripping. Texture and depth are adders, not alternatives. Apply the baseline, then add both. This answer covers the number to put on a drawing. The cost side of the same decisions, which design changes actually move the piece price, is set out in injection moulding, and the parts that come out of those tools are listed under plastic injection moulding. Where a wall or a feature cannot be moulded as drawn, CNC machining is the usual bridge for the first articles. Send the model and the texture specification, and we will return a draft check face by face before the tool is quoted Scope and sources. The baseline of 1 degree per side for standard ejection, half a degree on ribs and bosses, and the list of zero-draft mitigations including draw polishing, sleeve ejectors and air poppets are from a draft angle guide for buyers. The texture adder of 1 to 1.5 degrees per 0.001 inch, typical texture depths of 0.0025 to 0.0030 inch and the resulting 3 to 5 degrees on a textured face are from a plastic part design guide. The feature-level bands for polished, light, medium and heavy texture, the 0.5 to 1 degree rule for ribs and bosses, the 0.5 degree minimum on cored holes and the 3 degree minimum on shut-off faces are from a moulding design guide and a draft angle reference. The stripping guidance of 2 to 3 percent of part diameter on unfilled ductile resins comes from the same design guide. The side offsets in the table are calculated from the tangent of the stated angle and the stated depth; they are geometry, not a supplier claim. All angles are working minima for planning, not a substitute for the mould maker's review of the specific part and resin.

The short answer Choose die casting when the housing must carry load, spread heat, block electromagnetic interference or survive fire. Choose injection moulding when it must insulate, stay light and reach a low piece price. Tooling runs 15,000 to 60,000 dollars against 3,000 to 25,000, so volume decides the rest. Start with the material, not the process Die casting and injection moulding look similar on a factory floor: a steel tool, a clamp unit, a shot of molten material, a part ejected every cycle. The difference is what is being shot. Die casting runs aluminium, zinc or magnesium at 620 to 700 degrees Celsius; injection moulding runs thermoplastics at 180 to 350 degrees Celsius. That single fact fixes almost every downstream property. A die-cast housing is electrically conductive, thermally conductive, fire resistant, dimensionally stable at high temperature and roughly three times stiffer than the same wall in glass-filled nylon. A moulded housing is an insulator, weighs a third to a half as much, can be moulded in colour with a textured surface, and can integrate living hinges and snap fits that metal cannot produce in one shot. So the first question is not which process is cheaper; it is which set of properties the housing has to deliver. A motor housing that must sink 40 watts and survive a short circuit is a metal part, and no amount of cost engineering will make a moulded cover do that job. A consumer device cover that must be light, warm to the touch, radio-transparent and available in five colours is a moulded part, and casting it is an expensive way to make it heavier. Tooling, tool life and the volume where each pays The tooling difference is larger than the material difference. A single-cavity aluminium die casting tool typically costs 15,000 to 60,000 dollars and takes 8 to 20 weeks; a comparable single-cavity plastic mould costs 3,000 to 25,000 dollars and takes 6 to 10 weeks. Die casting tooling carries thicker plates, hot-work steel and slides that survive thermal cycling, which is where the money goes. Moulds can use pre-hardened steel and simpler cooling, because the melt is far cooler. Tool life runs the other way. An aluminium die is commonly rated 100,000 to 150,000 shots before heat-check refurbishment, with premium steel and good thermal management reaching 300,000 or more; zinc dies last considerably longer because the melt is cooler. A hardened plastic mould commonly lasts several times longer than an aluminium die, and its wear parts are cheaper to replace. The practical effect is that die casting carries a higher tooling cost over a shorter tool life, which is only recovered by the per-part economics at volume: plastic housing programmes usually justify tooling from about 1,000 to 10,000 parts, die casting from about 5,000 to 50,000, depending on size and complexity. Volume, parts per yearMoulded housingDie-cast housingWhat decides it under 1,0003D print or machinemachine or fabricateneither tool amortises 1,000 to 5,000tool pays, mould itbridge tool or machineplastic tooling is 3 to 5 times cheaper 5,000 to 20,000competitivetool pays if the part is metalfunction decides first, then price 20,000 to 100,000lowest piece pricecompetitive on functionmetal properties against unit cost above 100,000multi-cavity mouldmulti-cavity die, premium steelcavity count and tool life The tooling and tool-life rows are the ones that decide the business case; the material rows are the ones that decide whether there is a case at all. Wall thickness and the geometry each process allows Housing walls sit in a narrow band in both processes, for opposite reasons. Die cast aluminium is practical from about 1.0 mm to 3.0 mm, with 1.5 to 2.5 mm the comfortable middle for a structural housing; thinner sections lose fill and thicker ones create isolated mass that cannot be fed, producing shrinkage. Moulded ABS runs 1.1 to 3.6 mm, polycarbonate 1.0 to 3.8 mm and glass-filled nylon 0.8 to 2.9 mm, with the same rule that walls should be uniform: a thick section cools last, pulls material from the skin and leaves a sink mark on the far face. Geometry follows the material. Die casting needs draft on every vertical face, generous fillets at internal corners and ribs that are thinner than the wall they stiffen, with slides for undercuts; it can produce very fine detail, sharp logos and textured surfaces straight from the die, and it can cast in bosses and mounting pads that would otherwise be separate parts. Injection moulding allows thinner ribs, deeper draws, snap fits, living hinges, threaded inserts moulded in place and multi-shot colour, and it can hold a Class A cosmetic surface with a texture that would need secondary work on a casting. Both processes punish a thick section more than a thin one. The design rule is uniformity, not a single minimum number. Cycle time, tolerance and what happens after the shot A housing-sized aluminium part cycles in roughly 20 to 90 seconds, a plastic housing in 15 to 60 seconds, so cycle time rarely decides between them; the difference is smaller than the difference in material cost. Tolerance does differ, but less than buyers expect, and it is governed by standards rather than by slogans: castings are toleranced under the ISO 8062 system and moulded plastics under ISO 20457, and in both cases the tolerance you get depends on the feature, the tool and the material. As a working rule, a die casting holds around plus or minus 0.1 to 0.3 mm on as-cast features and a moulded part plus or minus 0.05 to 0.2 mm, with machined or reamed features on either reaching 0.05 mm or better. Post-processing is where the two diverge most in cost. A casting usually needs deburring, trimming, shot blasting, and often machining on mating faces, seal grooves and threads, plus a conversion coating or powder coat; a moulded part usually needs gate trimming and nothing else, with colour and texture built into the tool. That is why a die-cast housing is rarely the cheapest part at the same wall thickness: the secondary operations are part of the process, not an option. The hybrid housing is often the right answer Many production housings are both. A die-cast aluminium chassis carries the load, spreads heat from the power stage and provides the electromagnetic shield; a moulded cover closes it, provides the visible surface, the colour and the user interface, and keeps hands away from a hot or live surface. The joint is designed once, with a gasket groove in the casting and a matching lip in the cover, and the assembly is usually cheaper and lighter than an all-metal box and more capable than an all-plastic one. The design rules for the hybrid are worth stating. Put the tight tolerances and the sealing groove in the metal part, because it is more dimensionally stable and can be machined. Use the plastic part for features that need to flex, clip or insulate. Keep the thermal path in the metal and the user-contact surfaces in plastic. Decide the split before either tool is cut, because the parting line between the two parts is the one interface that cannot be fixed later without changing both tools. One common false economy: casting a housing that will be almost entirely covered by a plastic shell. If the metal is hidden, its cosmetic surface, its texture and much of its as-cast tolerance are being paid for and never seen. In that case either expose more metal, or move the part to moulding and use metal only where the function demands it. A decision sequence for a new housing Ask whether the part must be metal: structural load, heat spreading, electromagnetic shielding, fire rating or a grounded chassis. If yes, die casting is the default at volume, and machining at low volume. If no, ask whether the annual volume clears the plastic tooling threshold, which for most consumer housings is a few thousand parts. Then ask whether the surface must be cosmetic in colour, in which case moulding wins on both cost and finish. Finally, check the tool: draft, wall uniformity, gate position, ejector marks and, for a casting, the thermal balance of the die. Most housing failures at first article are one of those five, not a material problem. Function first, volume second, tool third. A housing that fails the first question cannot be rescued by a cheaper tool. See die casting for the metal route and its alloys, injection moulding for the plastic route and its design rules, and surface finishing for the coating and anodising steps that follow either process. Send the housing drawing and the annual volume, and we will price the die-cast and the moulded route including secondary operations Scope and sources. The tooling ranges of 10,000 to 100,000 dollars for injection moulding and 15,000 to 150,000 for die casting, cycle times of 15 to 60 and 20 to 90 seconds, wall thickness of 0.5 to 4 mm and 1 to 6 mm, the break-even bands of 1,000 to 10,000 and 5,000 to 50,000 parts, and the note that a small plastic housing may cost 0.40 dollars against 1.50 die cast are from a process comparison of injection moulding and die casting. The melt temperatures of 600 to 720 degrees Celsius for aluminium against 200 to 300 for polymers, the die face temperature of 180 to 320 degrees Celsius, the H13 die life of 100,000 to 150,000 shots for aluminium and 500,000 or more for zinc, and the failure modes of thermal fatigue against wear are from a die casting die reference. The process sequence, the housing and enclosure applications of each route and the secondary operations that follow are from a die casting and injection moulding comparison. The ISO 8062 and ISO 20457 tolerance systems are public standards for castings and moulded plastics respectively; the tolerance figures are working values and should be confirmed for the specific feature and material.

The short answer Porosity, cold shuts, flash, shrinkage, hot cracks and die soldering cause most die casting rejects. Start with the void shape: round voids are gas from venting and shot profile, jagged voids are shrinkage from feeding and wall design. Cut a section before changing any parameter. The defect map Die casting has a short list of failure modes and a long list of causes, and the useful way to organise them is by mechanism rather than by name. Filling defects happen when metal arrives cold, slow or turbulent: cold shuts, misruns and gas porosity all come from that family. Feeding defects happen after the cavity is full, when the casting contracts and there is no liquid metal or pressure to follow it: shrinkage porosity, sinks and hot cracks. Tooling defects come from the die itself: flash from a parting line that no longer closes, soldering from a die face that is too hot, heat checking from thermal cycling. Almost every defect a buyer sees on an incoming inspection traces back to one of those three families, and the fix is different for each. Read the middle column first. Two defects that look alike on the surface often have opposite causes, which is why parameter changes made without a diagnosis usually make things worse. DefectWhat it looks likeUsual causeWhat to change Gas porosityround internal voids, blisters after heat treatmentair or gas trapped during fillventing, vacuum, slower slow-shot, less lubricant Shrink porosityjagged voids, sinks on thick sectionsno feed during solidificationhigher intensification, longer hold, uniform walls Cold shutvisible seam or line where fronts metmetal too cold, fill too slow, long flow pathpour temperature, gate velocity, die preheat Misrunincomplete corner or thin ribmetal freezes before the cavity fillspour temperature, fill speed, venting Flashthin fin at the parting line or slidesclamp too low, die worn, pressure spikeclamp tonnage, die recondition, shot profile Hot crackcrack at a sharp transitionuneven contraction, poor ductilityradii, wall uniformity, alloy choice Solderingrough torn patch, die damagedie face above about 400 Ccooling, spray coverage, surface treatment Blistersurface bubble after T6 or service heatgas just below the skindegas the melt, vent, use T5 instead Porosity: gas or shrinkage, and how to tell Porosity is the defect that consumes the most meeting time, and most of that time is wasted because nobody has looked at the void. Cut the part through the suspect section and examine the cavity at magnification. A round void with smooth walls is gas: air or vapour was trapped during filling and could not escape before the metal froze. A jagged, irregular void with dendritic edges is shrinkage: the metal contracted as it solidified and there was no liquid metal or pressure available to fill the space it left. The two need opposite corrections, and a parameter change made without that one check has roughly a fifty percent chance of being wrong. For gas, the chain is venting, vacuum and shot profile. Vents and overflows belong at the last-filled areas, and the total vent area should be a meaningful fraction of the casting's projected area rather than a token slot; a common target is above 0.5 percent. Where the part is pressure-tight or the wall is thin, vacuum assistance removes the air before the metal arrives. The slow-shot stage should advance the metal without folding air into it, typically 0.1 to 0.3 metres per second, and the fast stage then fills the cavity before the front freezes, typically 35 to 60 metres per second. Lubricant is part of the same system: an over-sprayed die releases gas that has nowhere to go, so a fine atomised spray that leaves the surface slightly damp, not wet, is the target. For shrinkage, the chain is intensification, hold time and wall design. Intensification pressure of roughly 40 to 70 MPa, applied before the gate freezes, keeps feeding the casting as it contracts; hold time should cover most of the solidification time of the thickest wall, commonly 2 to 6 seconds. Neither will rescue a part designed with a 6 mm boss beside a 2 mm wall, because the thick section solidifies last and isolated. Coring the boss, adding cooling at the heavy section and keeping walls within about 2 to 1 is the design-side fix, and it is always cheaper than process compensation. Most defects start at the edge of a window rather than in the middle. A die running at 150 C has cold-shut risk; one running at 320 C has soldering and heat-check risk. Thermal balance is the root cause most people skip Die temperature decides fill, solidification and die life at the same time, which is why it is the first thing an experienced die caster checks and the last thing a buyer hears about. Aluminium dies normally run with a face temperature of 180 to 250 degrees Celsius at steady state. Below that range, metal freezes on contact, cold shuts and misruns appear, and intensification cannot reach the cavity. Above it, the release agent breaks down, aluminium starts to react with the steel, and soldering begins; the reaction becomes aggressive above roughly 400 degrees Celsius at the die surface, which is why local hot spots do the most damage. Thermal balance is also a geometric problem. A die with a thick steel mass on one side and a thin core on the other will run hot on the heavy side and cold on the thin side, no matter what the controller says, because the heat input and the cooling capacity are not matched. The fixes are cooling lines placed by thermal simulation rather than by convenience, baffles or bubblers in deep cores, and a spray pattern that actually covers the hot zones instead of the whole face evenly. Where a die is chronically out of balance, a die coat or a local insert of high-conductivity material is often more effective than another round of parameter tuning. Defects that process settings cannot fix Some defects are designed in. Sharp internal corners concentrate stress and start hot cracks; generous radii at every transition are the cheapest fix in the whole process. Abrupt wall changes create isolated heavy sections that cannot be fed, so a gradual transition over at least three times the thickness difference is better than a step. Deep, narrow pockets need long thin cores that bend and run hot, so they should be widened or split. Undercuts add slides, and every slide adds a parting line that can flash and a surface that can solder. Heat treatment belongs in the same category. A T6 treatment on a casting with uncontrolled porosity turns trapped gas into blisters, because the gas expands at solution temperature and pushes the skin out. Where porosity cannot be eliminated, T5 ageing gives most of the dimensional stability with much less blister risk, and vacuum impregnation can seal micro-porosity in non-structural, pressure-tight parts such as housings and valve bodies. Neither is a substitute for a sound casting, but both are legitimate routes for a part that is otherwise correct. What to specify and what to inspect A die casting purchase order that says "no porosity" is unenforceable, because no process delivers it. What works is a specification by zone and function: structural sections and pressure boundaries get a defined porosity level verified by X-ray or sectioning, cosmetic faces get a surface acceptance standard, and non-critical sections get the general level. State the inspection method and the sampling frequency, because a visual check and a radiograph answer different questions. For pressure-tight parts, state the leak test pressure and the acceptance criterion, and decide in advance whether impregnation is permitted. On the process side, ask for the shot profile, the die temperature record and the die maintenance interval, and ask how the supplier verifies that intensification is reaching the cavity. A supplier who can show those three things is controlling the causes; one who can only show final inspection is controlling the symptoms, and the symptoms will move to the next order. The loop is deliberately short. One sectioned part and thirty verified shots resolve most porosity arguments faster than a week of parameter changes. See die casting for the process and the alloys behind these defects, aluminium die casting for housing and bracket work, and surface finishing for the coating steps that follow casting and can hide or reveal a surface defect. Send the drawing with the critical zones marked and we will quote the casting with a porosity and inspection specification Scope and sources. The distinction between round gas voids and jagged shrinkage voids, the troubleshooting order that starts with a sectioned part, the effect of venting and overflow design, the 400 degree Celsius onset of aluminium soldering on steel and the use of vacuum impregnation are set out in a die casting defect and prevention guide. The slow-shot range of 0.1 to 0.3 metres per second, fast shot of 35 to 60 metres per second, pour temperature of 670 to 700 degrees Celsius, die temperature of 180 to 250 degrees Celsius, intensification of 40 to 70 MPa, hold time of 2 to 6 seconds, clamping of 40 to 70 kilonewtons per square centimetre of projected area and vent area above 0.5 percent of projected area are from a review of aluminium die casting parameters. The H13 die life of 100,000 to 300,000 shots for aluminium, the T5 against T6 blister risk, the 0.5 to 1.0 mm machining allowance and the cavity-count guidance are from a die casting design and defect prevention guide. These are typical windows; the correct settings for a specific die are established on its own trial shots.

The short answer Standard EN 755-9 holds plus or minus 0.15 mm on a 3 mm dimension and 0.25 mm across a hollow, opening to 0.5 mm at 90 to 120 mm. Minimum wall is about 0.8 mm in 6063 and roughly 25 percent thicker in 6061; anything tighter than the standard needs secondary machining. Two standards, two price levels Aluminium extrusion is bought against one of two European tolerance systems, and the choice changes the price more than most buyers expect. EN 755-9 is the general standard for extruded profiles and is what a mill quotes by default. EN 12020-2 is the precision system for close-tolerance profiles, and it roughly halves the permitted deviation on wall thickness and opening dimensions. A wall of 2 mm may deviate 0.25 mm under EN 755-9 and 0.15 mm under EN 12020-2; an opening below 60 mm moves from 0.40 to 0.15 mm. Straightness, twist and flatness tighten in the same proportion. The cost of that tightening is not a line on a price list; it is paid through slower press speed, more frequent die correction, more frequent measurement and a higher rejection rate. Suppliers commonly quote 20 to 30 percent more for precision class, and that is before yield loss. The rule that follows is to buy the standard you need and no tighter: specifiers who write EN 12020-2 "to be safe" pay for a capability the part does not use, while specifiers who leave EN 755-9 on an assembly that needs precision-class openings discover the mismatch at first build, when a 0.25 mm oversize means every part needs hand fitting. Precision class is not a better version of the same product. It is a different press setting, a different die correction schedule and a different price. Dimension A and dimension B: why hollows are looser EN 755-9 does not apply one tolerance to the whole section; it separates dimensions by what they span. Dimension A covers a solid dimension not affected by wall thickness, for example the outside width of a profile or the thickness of a solid leg. Dimension B covers a dimension across a metal-enclosed hollow space, and it carries a wider tolerance because both the die and the mandrel can move under pressure while the metal flows around them. At a nominal 20 mm, A is plus or minus 0.25 mm and B is plus or minus 0.45 mm; at 60 mm the figures are 0.35 and 0.70. Wall thickness is handled separately again, as class A, B or C depending on whether the wall is a free wall, an outer wall of a hollow or a wall between two cavities. Nominal dimensionDimension A, solidDimension B, across a hollow up to 3 mmplus or minus 0.15plus or minus 0.25 over 3 to 6 mm0.150.30 over 6 to 10 mm0.200.35 over 10 to 15 mm0.200.40 over 15 to 20 mm0.250.45 over 20 to 30 mm0.250.50 over 30 to 45 mm0.300.60 over 45 to 60 mm0.350.70 over 60 to 90 mm0.400.85 over 90 to 120 mm0.501.00 This split explains a common argument at first article. A buyer measures the inside width of a hollow and finds it 0.4 mm over nominal on a 20 mm section, and the supplier is entirely within the standard, because that dimension is a B dimension. The same 0.4 mm on a solid outside face would be a reject. Knowing which column a dimension falls in, before the drawing is released, prevents that conversation. Wall thickness: the floor is set by the die, not by the drawing There is no universal minimum wall in the standards, because the achievable value depends on alloy, section shape, circumscribing circle and press capacity. In practice, a modern press with a well-designed die will run 6063 down to about 1.0 mm on a small hollow with several cavities, and some shops quote 0.8 mm on simple sections. Large hollows, wide flat faces and asymmetric shapes need more: 1.5 mm and up is normal above about 100 mm across. 6061 needs roughly 25 percent more wall than 6063 for the same geometry, and 6082 more again, with simpler sections. Two mechanisms set the floor. The first is metal flow: thin walls cool faster than thick ones, so a section with a 1 mm wall beside a 4 mm boss will fill unevenly, and the thin side may tear or fail to fill. The second is die strength: the tongue of steel that forms a narrow gap is itself thin, and it deflects or breaks under press load. That is why the useful design rule is not "minimum wall" but "wall uniformity": keep adjacent walls within about 2 to 1, and accept a thicker wall everywhere rather than a thin wall in one place that forces a slower, hotter, less stable press pass. The last row is the escape hatch: where a wall must be thinner than the press will run, extrude thick and machine the pocket afterwards. Form tolerances: straightness, twist and flatness Cross-section tolerances get all the attention, and form tolerances cause most of the assembly problems. EN 755-9 allows roughly 1.0 mm of straightness deviation per metre on lengths up to a metre, rising to about 3.0 mm over 3 to 4 metres, with local deviation within any 300 mm segment limited to about 0.5 mm. Twist is limited to about 1.5 degrees per metre on a medium-width profile. That does not sound like much until it is multiplied: on a 6 metre length, 1.5 degrees per metre accumulates to a visible spiral, enough to misalign screw holes at the far end of a frame. Stretch straightening after extrusion corrects most bow and twist, but it also leaves residual stress, and that stress relaxes when the part later passes through an anodising or powder coating line at elevated temperature. Parts that were straight at the factory can arrive slightly bowed. The practical consequences are to specify straightness and twist explicitly on long parts, to allow for relaxation in the fixture design, and to check form after finishing rather than only before it. What the die and the press contribute Achievable tolerance is a function of tool condition as much as of the standard. A die wears, and dimensions drift as it does; shops commonly plan refurbishment on a tonnage interval rather than a calendar one, because wear tracks metal pushed through the die. Billet temperature is the second lever: holding it within a few degrees keeps flow stable, while a wide swing produces dimensional variation that no inspection plan can fix afterwards. Press speed is the third: running slower improves dimensional stability and costs money per kilogram, which is exactly what a precision class is buying. For the buyer, this means the tolerance on the drawing is a claim about the supplier's process, not just about the part. A shop that cannot show its die maintenance interval, its billet temperature control and its first-article measurement method is unlikely to hold a precision class consistently, whatever the documentation pack says. Ask for the measurement method and the sampling frequency, and ask which dimensions are measured on the CMM rather than with a caliper. How to write it on a drawing Put the standard in the title block, not in a note buried in the revisions. State the alloy and temper beside it, because tolerance and alloy are interdependent: a 6061 section at 1.2 mm wall is a different capability question from the same drawing in 6063. Then mark, dimension by dimension, which values are as-extruded and which are machined, since the two carry tolerances an order of magnitude apart. A machined face at 0.05 mm and an as-pressed face at 0.3 mm can sit on the same part without contradiction, as long as the drawing says which is which. Finally, resist the blanket tolerance. Applying plus or minus 0.1 mm to every dimension forces the whole profile into precision class and can add 20 to 30 percent to the price, when two or three functional dimensions were the only ones that needed it. The efficient drawing has a general tolerance note, plus explicit callouts on the mating faces, seal grooves and fastener positions, and nothing else. The second step is the one that prevents the first-article argument: a hollow dimension is not judged against the solid-dimension column. See CNC machining for the secondary operations that hold the tight features, surface finishing for anodising and its effect on form, and sheet metal fabrication where a folded part can avoid the tolerance question entirely. Send the profile drawing and we will mark which dimensions need precision class and which do not Scope and sources. The EN 755-9 dimension A and B table from 0.15 and 0.25 mm at 3 mm up to 0.50 and 1.00 mm at 90 to 120 mm, the 0.80 mm minimum wall on a 1,100 tonne press and the note that dimension B spans a metal-enclosed space and carries a wider tolerance are from a published extrusion tolerance and capability table. The EN 12020-2 and EN 755-9 comparison for wall thickness, opening and enclosing dimensions, together with the straightness, twist and flatness figures and the 20 to 30 percent cost effect of a blanket tight tolerance, is from a guide to extrusion tolerances. The straightness of 1.0 to 3.0 mm per metre, the twist of about 1.5 degrees per metre, the effect of die wear and billet temperature and the advice to specify tolerance by standard in the title block are from a tolerance standards reference. These are standard-class values; confirm the class and the measuring method with the extruder before the drawing is released.

The short answer Extrusion wins above roughly 500 to 1,500 parts when the cross-section is constant: a die costs 1,500 to 10,000 dollars, 95 percent of the bought metal ends up in the part and as-pressed tolerance is around 0.2 mm. Below that volume, or where the geometry changes along the length, machine it from stock. The two processes answer different questions Extrusion is a shaping process: heated billet is pushed through a die and the result is a continuous profile of one cross-section, cut to length afterwards. CNC machining is a subtracting process: a block or plate is cut away until the part appears. That single difference decides most projects before cost is discussed. A frame rail, a heatsink body, a linear enclosure, a mounting track or a window section has the same section for its whole length, and extrusion produces it at a fraction of the metal and machine time. A housing with a sloped face, a bearing bore at one end, mounting pads on two sides and a cable entry on the third is not a constant section, and no die will make it one. The mistake in both directions is common. Buyers extrude a profile and then machine away 60 percent of it to reach the final shape, which pays for the die and the machining; or they machine a long constant part from plate, which turns most of an expensive billet into swarf. The useful question is not which process is better but which features belong to which process. The arithmetic: die, metal and cycle time The cost of an extruded part has three lines: the die, amortised over the order; the metal, bought by weight; and the press and cut time. The cost of a machined part has two: stock and machine time, with almost no tooling. That is why the curves cross, and why the crossing point moves with part size. On a small simple section the die is 1,500 to 3,000 dollars and 100 to 200 parts can absorb it. On a large hollow profile with internal webs, the die can exceed 10,000 dollars and the break-even moves to 500 to 1,500 parts. Where the section is long and the order is large, the die cost per part becomes negligible and the material yield advantage dominates: extrusion uses 90 to 96 percent of the billet, while machining from plate typically converts 20 to 40 percent of the stock into chips. Order sizeExtrusion routeCNC routeWhat decides it 1 to 50 partsstandard profile, cut and machinedclear winnerno die to amortise, drawing can still change 50 to 500 partscustom die if the section is simpleusually still cheaperdie cost per part is 3 to 60 dollars here 500 to 5,000 partscustom die, machined endscompetitive only for small partsmetal yield and cycle time start to dominate above 5,000 partsclearly cheaper per partexpensive unless the part is tinydie is fully amortised, press output is high Yield is the quiet cost driver on long parts. On a 3 kg frame rail, the difference between 25 and 96 percent yield is worth more than the machining rate. Tolerance is where the cheap route stops An as-pressed profile is governed by EN 755-9 and holds roughly plus or minus 0.15 to 0.20 mm on small dimensions, opening to 0.5 mm or more as the section grows and to around 1 mm across a large enclosed hollow. That is adequate for frame geometry, cover fit and most mounting faces, and it is not adequate for a bearing seat, a seal groove, a dowel hole or any surface that must mate within 0.05 mm. Those features are added after extrusion by CNC, and the economic answer is almost always to keep them few, local and on the ends of the profile rather than along its length. Machining from stock starts at plus or minus 0.05 mm and can hold 0.02 mm on a critical feature, which is why it remains the default for prototypes, low volume and any part where the geometry changes in three dimensions. The comparison that matters is not "which is more accurate" but "how many features actually need the accuracy". A frame with two machined end faces and a 0.2 mm body tolerance is both cheaper and faster than the same frame machined all over. The gap between the second and fourth row is the reason the hybrid route exists: extrude the body, machine only the features that must be tight. Lead time and the cost of being early A custom extrusion die takes two to four weeks, plus a first-article sample for approval before the production run; standard profiles ship in days. CNC parts typically quote in three to ten working days depending on material and finishing. That difference matters most at the start of a programme, when the design is still moving. Cutting a die for a section that changes after the first prototype is the single most expensive way to use extrusion, and it is avoidable: prototype from a standard profile or from machined stock, freeze the section, then cut the die. There is a second, quieter lead-time cost. Once the die exists, the section is fixed. Any later change to wall thickness, a boss position or an internal web means a new die, not a new program. Machining keeps that flexibility at the price of a higher piece cost; extrusion trades flexibility for a lower piece cost. Which side of that trade you want depends on how confident the design is, not on the volume alone. The hybrid route is usually the real answer Most production frames and enclosures are neither pure extrusion nor pure machining. The body is extruded, cut to length, then the ends are machined for mating faces, the holes and threads are added, and the part is anodised or powder coated. Because the machine only removes a few millimetres instead of a whole block, cycle time falls sharply: cutting an extruded blank instead of a solid bar can remove 60 to 70 percent of the machining time on a housing, and tool wear falls with it. The design rules are simple. Leave 0.5 to 1 mm of stock on any face that will be machined, keep machined features off the thin walls, and mark on the drawing which dimensions are as-extruded and which are machined, because they carry different tolerances and different costs. A specification trap worth naming: a drawing that applies plus or minus 0.1 mm to every dimension forces the supplier to treat the whole profile as precision grade, which can add 20 to 30 percent to the price. Put the tight tolerance only on the features that function, and leave the rest at the standard. When each route is the wrong one Extrusion is wrong when the section changes along the length, when the order is too small to amortise a die, when the part needs fewer than about two walls of constant thickness, or when the design is not yet frozen. It is also wrong when the profile would need so much machining afterwards that the die saves nothing. Machining is wrong when a long constant part is cut from plate, when the same section repeats across thousands of units, when material cost dominates the part, or when the geometry is simple enough that a die would pay for itself within one order. Sheet metal is the third option that often wins for enclosures below about 2 mm wall: it needs no die, folds in days and tolerates design change, at the cost of stiffness and joint count. Run the checks in order. A single no usually sends the part to machining or to sheet metal, and finding that out before the die is cut is the whole point. See CNC machining for the secondary operations and the low-volume route, sheet metal fabrication for the folded alternative on enclosures, and surface finishing for anodising and coating after either process. Send the section and the annual volume, and we will price the die route and the machined route against each other Scope and sources. The die cost range of 1,500 to over 10,000 dollars, the break-even band of 100 to 200 parts for simple sections, the material waste of 5 to 10 percent for extrusion against 30 to 70 percent for machining, the tolerance comparison and the 70 percent cycle-time reduction from machining an extruded blank are set out in a comparison of CNC machining and aluminium extrusion. The break-even range of 500 to 1,500 units, the cost trend by order size and the recommendation of 500 to 1,000 pieces to justify a die are from a bulk-order cost comparison. The as-pressed tolerance of about 0.2 mm, the machined tolerance of 0.05 mm, the two to four week die lead time, the die cost range of 1,500 to 10,000 dollars and the roughly 1,000 piece threshold for an enclosure are from an extrusion enclosure design guide. These are planning figures for standard sections; the die price and the crossing point for a specific part should be quoted against its actual section and annual volume.

The short answer Choose 6063 when the profile is visible, thin-walled or complex, because it extrudes at 50 to 60 feet per minute and anodises evenly. Choose 6061 when the section must carry load: at T6 it reaches 290 to 310 MPa against 186 to 240 MPa for 6063, for about 10 to 15 percent more on the metal. One alloying element explains most of the difference Both alloys are Al-Mg-Si from the 6000 series, and on a datasheet they look almost identical. The difference that matters is copper. 6061 carries 0.15 to 0.4 percent copper, magnesium at 0.8 to 1.2 percent and silicon at 0.4 to 0.8 percent. 6063 keeps copper under 0.1 percent, with magnesium at 0.45 to 0.9 and silicon at 0.2 to 0.6. That small addition of copper raises the strength reached after artificial ageing, and it also changes how the metal flows through a die and how evenly it takes an anodic film. Almost every trade-off on this page, from press speed to anodising streaks, follows from those numbers rather than from anything the sales sheet says about "premium alloy". The practical consequence is that the alloy decision should be made before the cross-section is drawn, not after. A die is cut for one alloy family and one flow behaviour; if the profile is designed around 6063 and the structural calculation later demands 6061, the section usually has to be redrawn thicker and simpler rather than simply re-quoted. Buyers who reverse that order end up paying for a die twice. Strength: what the numbers do and do not buy you In the T6 temper 6061 reaches a typical 290 to 310 MPa tensile and 240 to 276 MPa yield, with elongation around 8 to 10 percent. 6063 in T6 sits at roughly 186 to 240 MPa tensile and 145 to 214 MPa yield; in T5, the temper most architectural profile is actually bought in, the figures are lower again, near 205 MPa tensile and 180 MPa yield. The exact values depend on section thickness and on which edition of the product standard the mill works to, so treat any single number as indicative and check the mill test record for the heat you receive. Two things follow. First, the strength gap is real but narrower than the headline 40 percent suggests once you compare like tempers and like sections. Second, and more important for design, both alloys have essentially the same elastic modulus, around 68 to 70 GPa. Stiffness therefore comes from geometry, not from alloy: a deeper section in 6063 will usually deflect less than a shallow one in 6061, at lower cost. Alloy choice buys yield strength, fatigue performance and the ability to survive a proof load. It does not buy rigidity, and specifying 6061 to fix a deflection problem is one of the most common and most expensive mistakes in extrusion design. Read the speed and wall rows together with the strength rows: the same alloy content that adds strength makes the profile slower to press and harder to draw thin. Alloy and temperTensile, MPaYield, MPaWhere it is the right answer 6063-T5about 205about 180window and door sections, trim, heat sinks, enclosures 6063-T6186 to 240145 to 214drawn tube, light frames, decorative profiles needing more stiffness 6061-T6290 to 310240 to 276machine frames, brackets, truck and trailer parts, marine structure 6082-T6about 310about 260European load-bearing sections where 6061 margin is short 6005A-T6about 260about 225rail and ladder profiles between 6063 and 6082 Extrudability decides the shape you can afford Extrudability is not a vague virtue; it sets press speed, scrap rate and the thinnest wall the die can hold. 6063 is the reference alloy of the industry because it flows at low pressure: a profile that runs at 50 to 60 feet per minute in 6063 typically runs at 30 to 40 feet per minute in 6061, and the slower pass means more furnace time per kilogram, more die wear and a higher scrap rate on start-up. Where the geometry is demanding, the gap widens further. Complex hollow sections with internal webs, multi-cavity heatsink profiles and walls down to 1.0 mm are routine in 6063 and marginal or impossible in 6061. The design rule that follows is blunt: if the profile needs walls below about 1.25 mm, deep narrow slots, or several internal cavities, 6063 is not a preference but a requirement. 6061 generally needs roughly 25 percent more wall thickness for the same geometry, which erodes much of the weight advantage a buyer hoped to gain by choosing the stronger alloy. Die cost also moves: a harder alloy needs more bearing length and more robust tooling, and 6082 tooling typically costs 10 to 20 percent more than the same section in 6063. Strength is the reason to move up the list; press speed, minimum wall and anodising quality are the reasons to stay low. Surface, anodising and heat For anything visible, 6063 is the safer specification. Its lower impurity content and finer grain produce a smoother as-extruded surface, and the anodic film grows more uniformly, so clear and coloured anodising comes out even across a batch. 6061 anodises perfectly well for protective purposes but the copper content tends to produce a slightly darker, less uniform appearance and occasional streaking, which is why decorative architectural work almost never specifies it. If the part is to be powder coated rather than anodised, this advantage largely disappears, because the coating hides the substrate. Thermal behaviour runs the other way from strength: 6063 conducts about 200 watts per metre-kelvin against roughly 167 for 6061, which is why LED housings and heatsink profiles are almost always 6063. Where the extrusion is both a heatsink and a structural member, the usual answer is a 6063 body with 6061 or steel inserts at the load path, rather than compromising both functions in one alloy. Cost is per metre, not per kilogram 6063 is normally quoted 10 to 15 percent below 6061 on metal, and the gap widens once press speed and scrap are included: a faster pass uses the press less, generates less butt scrap and wears the die more slowly. But the comparison that matters is cost per finished metre or per finished part, not per kilogram. A 6061 profile that must be 25 percent thicker to extrude cleanly can easily cost more per metre than the 6063 section it replaced, while weighing more, and it may then need a machining operation to hold the tolerance the thicker wall was supposed to protect. The reverse is also true: on a heavily machined bracket, 6061 machines faster and holds a better finish, so the higher metal price is recovered at the spindle. The cheapest mistake to avoid is specifying alloy by habit. "We always use 6061" is common in machine-building and "we always use 6063" is common in architecture, and both habits overpay on half their parts. State the load, the finish and the wall, then let those three pick the alloy. A decision rule that holds up in a quote review Ask four questions in order. Does the section carry a calculated load, or survive a proof or fatigue case? If yes, start at 6061-T6 and check whether 6082 is needed. If no, start at 6063. Is the surface visible and anodised? If yes, stay at 6063 unless strength forces you up, and expect a cosmetic penalty if it does. Are the walls below 1.25 mm or the section complex? If yes, 6063 only. Finally, set the temper honestly: T5 is adequate and cheaper for most architectural and enclosure work, while T6 is for strength and for parts that will be machined after extrusion. Where the answer is genuinely mixed, split the assembly. Visible covers and cosmetic profiles in 6063, hidden structural members in 6061, fasteners and inserts in steel or stainless: that combination is common in machine guards, display fixtures and modular frames, and it usually beats forcing one alloy to do both jobs. The order matters. Alloy is fixed before the cross-section is finalised, and the cross-section is what the die is cut for. See CNC machining for the secondary operations that turn an extruded profile into a finished part, surface finishing for anodising and powder coating on extrusion, and sheet metal fabrication where a folded part competes with an extruded one. Send the section, the load case and the finish, and we will quote 6063 and 6061 side by side Scope and sources. The alloy compositions of 0.15 to 0.4 percent copper in 6061 and under 0.1 percent in 6063, the strength ordering of 6063, 6061 and 6082, the press-speed effect on die design and the 10 to 20 percent tooling premium for harder alloys are set out in an extrusion alloy selection guide. The typical tensile figures of 290 to 310 MPa and 240 to 276 MPa yield for 6061-T6, 145 to 186 MPa for 6063-T6 and the note that 6061 machines more easily are from a 6061 and 6063 comparison. The press speeds of 50 to 60 feet per minute for 6063 and 30 to 40 for 6061, the roughly 25 percent thicker wall requirement for 6061 and the 10 to 20 percent die cost increase for 6082 come from a guide to alloy selection and die design. The thermal conductivities of about 200 and 167 watts per metre-kelvin and the 10 to 15 percent cost difference are from a 6063 and 6061 selection note. These are typical values for standard tempers; confirm the properties against the product standard and the mill test record for the actual heat.