Meritus Technology

← Insights

Specification misinterpretation in China sourcing: ISO 2768, AWS A2.4, and ASME B46.1 gaps that quietly cost projects time

2026-05-15 · 7 min read

Specification misinterpretation is responsible for more rework, delay, and dispute on China-sourced industrial procurement than any other single cause. It is also the most preventable. The problem is not that Chinese engineers cannot read English specifications. The problem is that technical standards are not universal, and the gaps between ISO, ASME, DIN, AS/NZS, and the Chinese GB/T family are real, undocumented in most RFQs, and therefore invisible to the factory until incoming inspection rejects the batch.

1. Dimensional tolerancing: ISO 2768-m vs GB/T 1804

A drawing issued to ISO 2768-m includes a general tolerance note. The factory applies it correctly in Chinese standard terms. The bilateral tolerance range in ISO 2768-m at a given nominal dimension differs from the equivalent GB/T 1804-m tolerance. The parts are within GB/T tolerance and out of ISO tolerance. The factory has done nothing wrong on its own standard. The buyer has parts that do not meet the drawing. Documented bridging at PO drafting prevents this entirely: either re-issue the drawing with explicit feature-by-feature tolerances, or include a bilateral standards reference table that the factory signs off on before any cutting begins.

2. Welding symbology: AWS A2.4 vs ISO 2553

A factory trained on ISO 2553 will misinterpret an AWS A2.4 drawing for groove weld geometry. The two systems use different reference-line conventions, different conventions for arrow-side versus other-side, and different default geometry assumptions for partial-penetration welds. The weld looks correct visually after fabrication. The throat thickness and penetration profile are wrong. NDT may or may not catch it depending on the technique used. The fix is a pre-fabrication WPS review: the welding engineer on the buyer side or the procurement representative confirms each weld symbol with the factory engineer before any welding starts, and the resolved geometry is documented in the WPS that travels with the order.

3. Surface roughness: ISO 4287 vs ASME B46.1

Ra values in micrometres are the same number in any language. The cut-off length used to measure them is not. ISO 4287 and ASME B46.1 specify different default cut-off lengths for Ra measurement. The same Ra number on a drawing therefore means a different thing depending on which standard the factory's surface roughness tester is calibrated to. A finish measured at the wrong cut-off length can pass on the factory floor and fail at the buyer's incoming inspection. The fix is to specify the cut-off length explicitly on the drawing (typically λc = 0.8 mm for most general engineering surfaces) and to require the inspection report to record the cut-off length used.

What this means at PO drafting

These three gaps are not fabrication errors. They are coordination errors at the boundary between two regulatory cultures (Western industrial standards and the Chinese GB/T framework). All three are visible and fixable at PO drafting, not at FAT. The cost of fixing them at PO is negligible. The cost of fixing them at incoming inspection is the cost of rework plus the schedule slip.

What we do

Meritus Technology runs a technical review step on every order that goes through our desk. The engineer assigned reads the drawing package against the factory's declared standard base and documents every discrepancy before work starts. Bridging notes are embedded in the factory drawings. The PO references the resolved standard. The factory signs off in writing.

If you are sourcing from China and want to see the checklist we use, we are happy to share it.

Contact: sales@meritustech.com · +86 134 8206 1802 · www.meritustech.com


Discuss with our team: sales@meritustech.com · +86 13482061802

Discussion on LinkedIn: View the original post and join the comments