Why the Best-Looking Die Casting Numbers Are the Least Reliable
Executive Summary
Individually plausible numbers can be collectively impossible.
- Most supplier pages are audited one figure at a time, which is exactly the wrong unit of analysis. Contradictions live between figures, not inside them.
The tightest number on a page is usually the least real.
- Marketing writes capability tables; engineers write case studies. When the two disagree, the case study is the measurement and the table is the aspiration.
A percentage is not a tolerance.
- "±0.5% dimensional accuracy" means ±0.05 mm on a 10 mm feature and ±2.50 mm on a 500 mm one. It cannot be verified, so it cannot be relied on.
Every pass rate contains a failure rate that nobody prints.
- A 95% first-article pass rate is a one-in-twenty chance that your programme needs an unplanned second sampling round.
The Audit Method Everyone Uses Is the Wrong One
Buyers shortlisting Aluminum Die Casting suppliers evaluate them the way they proofread: one line at a time. Is 600 tons enough? Is A380 the right alloy? Is ISO 9001 current? Each answer is yes, and the supplier passes.
This misses almost everything, because a number that is wrong on its own is rare. A number that is wrong in combination is routine. Nobody publishes an impossible clamp force. Plenty of pages publish a clamp force that contradicts the machine list beneath it.
The shift is small and changes everything: stop asking whether each figure is plausible, and start asking whether any two figures can be true simultaneously. You need no equipment and no plant visit. Everything required is already on the page.
Below are five pairings that do most of the work.
Cross-Check 1: The Capability Table Against the Case Studies
This is the highest-yield check available, and it takes four minutes.
Read the capability table on any die casting services page. Write down the tolerance. Then read the case studies at the bottom of the same page and write down every tolerance quoted there. On detailed metal casting services pages, the two rarely match — and the direction of the mismatch is always the same.
One page reviewed for this article states ±0.01 mm in its capability table, then ±0.005 mm in its quality section, then quotes ±0.02 mm on a medical housing, ±0.03 mm on a flange flatness, and ±0.05 mm on an automotive bore. Five precision claims, spanning a factor of ten, on one page.
Which one is true? Almost certainly the loosest.
Case study numbers behave differently from table numbers. They were produced by a real part, measured by a real inspector, and often approved by the customer named in the story. A capability table has no such constraint — nobody signs off a webpage. When a supplier's own success stories quote looser tolerances than the specification table above them, the specification table is aspirational.
A second reading follows immediately. ±0.01 mm is not a casting tolerance at all; it is a machining tolerance. As-cast capability runs around ±0.10 mm per 25 mm within a single die half and roughly ±0.25 mm across a parting line, where the two halves can shift relative to each other. Anything an order of magnitude tighter than that describes a secondary CNC operation on the die cast metal — with its own fixture, cycle time and price, either quoted separately or absorbed silently into a piece price you thought was for as-cast die Casting Parts.
"Was the ±0.01 mm in your table achieved as-cast or after machining, and can you send the CMM report for the feature?"
Cross-Check 2: The Percentage Against Arithmetic
"±0.5% dimensional accuracy" appears on supplier pages because it sounds rigorous. Convert it and it dissolves.
| Nominal | ±0.5% resolves to |
| 10 mm | ±0.05 mm |
| 50 mm | ±0.25 mm |
| 100 mm | ±0.50 mm |
| 300 mm | ±1.50 mm |
| 500 mm | ±2.50 mm |
Tolerance does not scale with size, because function does not scale with size. A bearing seat needs the same few hundredths whether it sits in a 40 mm cover or a 400 mm die cast housing. Expressing precision as a percentage produces bands that are punishingly tight on small features and meaninglessly loose on large ones.
There is a harder objection. A specification that resolves differently for every dimension cannot be checked, so it can never be breached. A claim that cannot fail is not a specification — it is a decoration. When it appears next to a hard millimetre figure elsewhere on the page, the two are not describing the same reality.
Cross-Check 3: The Alloy List Against Process Physics
An alloy list is the easiest place to catch a page written by someone who has never stood next to a machine, because alloys cannot be diplomatic. Either the cast metal flows and freezes predictably, or it does not.
One aluminium list reviewed here runs: ADC10, A380, ADC12, Pure Aluminum, DM6, HA6, HA4, 6063.
Two entries do not belong.
6063 cannot be high pressure die cast. It is a wrought extrusion alloy with silicon around 0.2 to 0.6%, against the 7 to 13% that gives casting alloys their fluidity and narrow freezing range. Injected into a steel die it hot-tears. Seeing 6063 on a custom die casting alloy list means the list was assembled from an aluminium reference rather than from the melt deck — and the same applies to 6061, which appears on these lists even more often. Both belong in aluminum fabrication from extrusion, or in gravity aluminum casting using a different alloy entirely.
Pure aluminium is not a die casting material either. Without silicon it has poor fluidity, high shrinkage and an aggressive tendency to solder to the die face.
Neither error is fatal on its own. Both tell you the same thing about the page: no foundry engineer read it before publication. That is worth knowing before you rely on any other number it contains, including the ones you cannot check.
Cross-Check 4: Every Pass Rate, Read Backwards
"First-article inspection pass rate exceeding 95%." Read backwards: roughly one new tool in twenty fails its first article. That is not scandalous — it is normal, and honest for a supplier to publish. But if your programme is the twentieth, you need a T2 sampling round in the schedule that nobody quoted. The number is an argument for a contingency, presented as an argument for confidence.
"99.98% leak test pass rate." Two leakers per ten thousand parts. Entirely acceptable for a decorative cover. Not obviously acceptable for a brake or fuel circuit, where the relevant question is not the rate but the containment: does the test catch all of them, and what happens to the ones it does not?
"99.80% on-time delivery." One late shipment in five hundred. The useful follow-up is never the percentage — it is which orders were late, by how long, and whether the lateness clustered around new tooling launches, which is where it usually lives.
Key point: a pass rate without a sample size and a time window is a claim, not a measurement. Ninety-five per cent of twenty tools and ninety-five per cent of two thousand tools are different statements.
Cross-Check 5: The Headline Figure Against Its Missing Qualifier
The last check is the simplest. For each impressive figure, ask what condition it must have been measured under, then look for that condition on the page.
"Minimum wall thickness 0.8 mm." Under what flow length? A 0.8 mm wall fills readily 30 mm from the gate and cold shuts 200 mm away, where the metal front loses superheat and two fronts meet without fusing. Thin-wall capability is a ratio, not a number, and the page that omits the denominator measured it on a coupon beside the gate.
"Maximum mould life 1,000,000 cycles for zinc, 100,000+ for aluminium." These figures are broadly reasonable — zinc runs cooler and is far kinder to tool steel than aluminium, so an order-of-magnitude difference is real. What is missing is the maintenance schedule underneath them. Thermal fatigue cracking is the normal ageing mechanism of H13, managed by scheduled polishing and localised weld repair, not prevented. A shot count published without a maintenance interval and a statement of who pays for it is half a commitment.
"MOQ 500–1,000 pieces." Against what tooling amortisation? A low MOQ with tooling amortised into the piece price is a different commercial structure from a low MOQ with tooling invoiced separately, and only one of them survives a volume change. This matters most in OEM casting programmes, where the second year is where the structure gets tested.
What We Publish, and What Qualifies It
Applying our own test to our own page.
| Parameter | Value | Qualifier |
| Tolerance, as-cast, same die half | ±0.10 mm / 25 mm | NADCA linear |
| Tolerance, across parting line | ±0.25 mm | Datum must be stated |
| Tolerance, machined | ±0.02 mm | Secondary operation, quoted separately |
| Minimum wall, aluminium | 1.2 mm | Within 80 mm flow length of a gate |
| Minimum wall, zinc casting | 0.8 mm | Hot chamber, within 60 mm of a gate |
| Clamp force | 160 – 3,500 tons | Matched to projected area, not part weight |
| Alloys | A380 · A383 / ADC12 · A360 · A413 · ZAMAK 3/5 | A360/A413 where bright anodise is required |
| Tool life, aluminium | 100,000 – 150,000 shots | With polishing every 20,000–30,000 shots |
| Tooling lead time | 5 – 8 weeks | From drawing freeze, excluding DFM iteration |
| Quotation turnaround | 24 hours | Complete packages; others get a question list |
| Certification | ISO 9001 · ISO 14001 · CE · RoHS · SGS | Issued in our own legal entity name |
We publish no percentage-based accuracy figure, because one cannot be verified.
In-house scope: tooling design and manufacture, cold chamber aluminum die casting, hot chamber zinc, 3/4/5-axis CNC machining, blasting, anodising, powder coating, plating, insert installation, sub-assembly and leak testing — one aluminum alloy processing manufacturing chain with no undeclared handovers. Our metal fabrication and custom metal fabrication lines — laser cutting, press brake forming, TIG and MIG welding — often let a cast body carry fabricated brackets, removing side actions from the die; for low volumes, structural fabrication needs no tooling at all.
Typical aluminium casting products and aluminium casting parts: LED heat sinks, telecom enclosures, motor housings, gearbox covers, valve bodies and mounting brackets. We hold no stock — every custom casting and custom fabrication job is built to drawing or sample, with permanent tooling ownership and 30+ years of factory processing experience behind it.
Frequently Asked Questions
Q1: A supplier's capability table and case studies quote different tolerances. Which should I believe?
The case studies. They describe parts that were actually made, measured and shipped, often to a named customer, whereas a capability table carries no such verification. When a page claims ±0.01 mm in its table but its own automotive case study reports ±0.05 mm on a bore, the ±0.05 mm figure came from a CMM and the ±0.01 mm figure came from a marketing brief.
Q2: Is ±0.01 mm achievable in aluminium die casting?
Not as cast. Realistic as-cast capability is around ±0.10 mm per 25 mm within one die half and roughly ±0.25 mm across a parting line, because the die halves shift relative to each other. Reaching ±0.01 mm requires secondary CNC machining with its own fixture and inspection step. The number can be true of the finished part while being untrue of the casting process, which is precisely how it gets published without anyone lying.
Q3: What is wrong with a "±0.5% dimensional accuracy" claim?
It is not a tolerance, because tolerance does not scale with size. The same claim resolves to ±0.05 mm on a 10 mm feature and ±2.50 mm on a 500 mm one, yet a bearing fit needs the same absolute band regardless of overall part size. A specification that resolves differently for every dimension cannot be breached, and something that cannot fail cannot be relied on.
Q4: How do I sanity-check an alloy list?
Look for wrought alloys among the casting alloys. 6061 and 6063 are extrusion alloys with silicon around 0.2 to 0.8%, far below the 7 to 13% that makes casting alloys flow and freeze predictably, so they hot-tear in a steel die. Pure aluminium has poor fluidity and solders to the die face. Any of these on a die casting alloy list means the page was compiled from a materials reference rather than by someone who runs the melt deck.
Q5: Is a 95% first-article pass rate good or bad?
It is honest, and it is a planning input rather than a reassurance. One new tool in twenty failing its first article is normal for the industry; the risk is that the number is presented as confidence when it should prompt you to build a T2 sampling round into your schedule. Ask how many tools that percentage covers and over what period — 95% of twenty is a very different claim from 95% of two thousand.
Q6: Why does a minimum wall thickness figure need a flow length?
Because filling depends on how far the metal must travel before it freezes. A 0.8 mm wall fills reliably near the gate and cold shuts a few hundred millimetres away, where the flow front loses superheat and two fronts meet without fusing — producing a part that looks complete and leaks under test. A minimum wall published without a stated flow distance was almost certainly measured on a test coupon next to the gate.
Q7: Are the mould life figures suppliers publish trustworthy?
The shot counts are usually plausible; the omission is the maintenance schedule. Zinc genuinely reaches far higher cycle counts than aluminium because it runs cooler and is gentler on H13 tool steel. But thermal fatigue cracking is the normal ageing mechanism of any die and is managed rather than prevented, through scheduled polishing and localised weld repair. Ask for the maintenance interval and who pays for it before the tooling agreement is signed.
Q8: What if a supplier publishes very few numbers at all?
That is a different failure, not a safer one. A page consisting entirely of "depends on the application" and "confirmed at quotation" transfers all interpretive risk to you and commits the supplier to nothing. The trustworthy pattern is a small number of figures, each carrying its own qualifier — flow length with wall thickness, datum with tolerance, maintenance interval with tool life.
Q9: How long should this cross-check take?
Fifteen minutes per supplier. Compare the capability table with the case studies, convert any percentage into millimetres at your actual part size, scan the alloy list for wrought grades, invert every pass rate, and check whether each headline figure carries the condition it must have been measured under. It finds more than a plant audit does at this stage, because it tests whether the supplier's own claims survive contact with each other.