The Samples Fit Perfectly, but the Production Lids Wouldn't Close
A Guide to Dimensional Tolerance Inspection - Why Dimensions Drift in Mass Production, General Tolerances (ISO 2768, GB/T 1804) and Molded-Part Tolerances, How to Mark Critical Dimensions, Calipers, Micrometers, Go/No-Go Gauges and CMMs Compared, Drawing and Spec Items, Per-Cavity Full Dimensional Reports, Checks at Sample, Production and Shipment, Request Wording, Real Cases and a Checklist

Hello, this is StartMade.

An online kitchen storage brand had 8,000 sets of airtight food containers made at an injection molding factory in Taizhou, Zhejiang. The three sample sets snapped shut with a clean click and didn't leak even when filled with water and turned upside down. But when the production goods arrived at the warehouse in Korea, some lids spun loosely on the base and others were so tight they couldn't be pressed all the way down. A full sort turned up about 960 bad sets, 12% of the order.

Tracing the cause, the lid mold had four cavities (it molded four lids per shot), and only the lids from one cavity had an inner rim diameter about 0.4mm too large. The three samples just happened to come from good cavities. The drawing had a single line, "未注公差 ±0.5", meaning any dimension without its own tolerance could vary by ±0.5mm, and the dimensions where the lid meets the base had no tolerance of their own. As far as the factory was concerned, a 0.4mm difference was within spec. The pre-shipment inspection had checked appearance, quantity and packaging, and nobody measured anything.

Dimensional problems stay invisible until someone assembles the product or a customer uses it. This article covers why dimensions drift in mass production, how to write tolerances on a drawing, which tools measure what, and what to check at each stage from sample to shipment.

Factories hold only the tolerances written on the drawing. Mark mating dimensions as critical dimensions and get a report with actual measured values from every cavity.

1. Why Dimensions Drift in Mass Production

Samples are usually the handful of parts the factory made with the most care. In mass production, molds, materials, operators and machine settings all shift a little, and those shifts show up as dimensional variation. These are the causes we see most often.

• Cavity-to-cavity variation: In a multi-cavity mold, each cavity has slightly different cooling and machining accuracy, so parts from the same mold can measure differently.
• Shrinkage changes: Plastic shrinks as it cools, and the amount changes with the resin lot, mold temperature and injection pressure. Verifying resin grade is covered in Episode 125.
• Mold and tool wear: After tens of thousands of shots the parting surfaces wear, and in machining, tool wear slowly pushes dimensions in one direction.
• Surface treatment thickness: Plating, anodizing and painting add a few to several tens of microns. Parts that fit right after machining sometimes stop fitting once they are finished.
• Different measuring methods: If the datum, the tool or the timing differs, the same part gives different numbers. This is the first thing to suspect when the factory's values don't match yours.

Plenty of orders go out with no drawing at all, just sample photos and the phrase "same as sample (按样品)". The factory then decides the tolerances on its own, and when something goes wrong there is no agreed standard to point to.

2. Tolerance Basics and Common Standards

A tolerance is the allowed range for a dimension. If the drawing says "50 ±0.2", anything from 49.8 to 50.2mm passes. Writing a tolerance on every single dimension is impractical, so drawings usually state a general tolerance class in one corner and add individual tolerances only to the dimensions that matter.

Mark Critical Dimensions Separately

Not every dimension matters equally. Dimensions where a small miss breaks the function, such as the diameter where a lid meets its base, the distance between screw holes or the space a battery sits in, should be designated as critical dimensions (关键尺寸, CTQ) and numbered on the drawing with balloon callouts. Those numbers also tell the factory which dimensions to measure more often. Molds and the injection process in general are covered in Episode 33.

3. Measuring Tools and How to Measure

Each tool has its own resolution. If a dimension's tolerance is ±0.05mm and you measure it with a ruler marked in 0.1mm, you can't tell whether it passes. A common rule is to use a tool with resolution about one tenth of the tolerance range.

Agree on measuring conditions too. Molded parts keep shrinking for hours after molding, so they are usually left at room temperature for about 24 hours before measurement. Writing the measuring temperature (typically 23±2°C), datum surfaces, measuring locations and tools into the drawing or inspection standard sharply cuts down on number disputes with the factory.

4. Items to Put in the Drawing and Spec

"按样品" or a single "±0.5" line doesn't tell the factory what to hold and how tightly. Split the items below between the drawing and the spec sheet.

For writing spec sheets in general, see Episode 59, and for mold modifications and design change control, see Episode 119. Cavity numbers cost almost nothing when they go into a new mold, but adding them later means reworking the mold, so ask for them when you order the tooling.

5. How to Check at Each Stage

Sample Stage

When the first mold samples (T1) arrive, ask for a full dimensional report on 3-5 pieces from every cavity. Measure the critical dimensions on a few of them yourself with a caliper, compare against the factory's report, and actually assemble them with their mating parts. Approve the mold only when both the numbers and the fit are right. Approving off samples from a single cavity is exactly how the case at the top of this article happened.

First Article and Production Stage

On the first production lot, a first article inspection (FAI) rechecks critical dimensions cavity by cavity. Pre-production meetings and first article checks are covered in Episode 123. For large volumes, buyers sometimes require a process capability index (Cpk) of 1.33 or higher on critical dimensions. For small orders, even a simple patrol rule such as "measure critical dimensions on one part per cavity every 2 hours and keep the records" goes a long way.

Pre-Shipment Inspection Stage

A standard AQL inspection focuses on appearance, so you have to ask for dimensional checks separately. Give the inspector the list of critical dimensions and the tools to use in advance, and have samples drawn so cavity numbers are evenly mixed. For mating parts, you can have go/no-go gauges made and give one set each to the factory and the inspector. Nobody has to read numbers; the part either goes in or it doesn't, which speeds up decisions on site. Pre-shipment and third-party inspection are covered in Episode 100 and Episode 47.

6. Sample Request Wording for the Factory

Putting the wording below in your purchase order or spec makes the factory manage critical dimensions separately and leaves less room to argue over the standard when a dimensional problem comes up.

"本产品须按我方确认的图纸（版本号V2，日期2026-10-10）生产，未注公差按ISO 2768-m执行。图纸中标注气泡编号1至8的尺寸为关键尺寸，须按图纸标注的公差管控。T1样品须按每个模穴各提供5件的全尺寸报告，并附三坐标或影像测量仪测量数据。测量须在注塑成型后常温放置24小时、环境温度23±2℃条件下进行。量产期间须每2小时按每个模穴抽取1件测量关键尺寸并保存记录，出货时一并提供。产品须刻有模穴编号。如需修改模具、更换原材料或变更加工工艺，须事先取得我方书面确认。"
(This product must be made to the drawing we approved (version V2, dated 2026-10-10), with ISO 2768-m applied to dimensions without a stated tolerance. Dimensions marked with balloon numbers 1 to 8 are critical dimensions and must be held to the tolerances shown on the drawing. T1 samples must come with a full dimensional report on 5 pieces from each cavity, with CMM or vision measuring system data attached. Measurements must be taken after the molded parts have rested at room temperature for 24 hours, in a 23±2°C environment. During production, one part per cavity must be measured for critical dimensions every 2 hours, with records kept and provided at shipment. Parts must be engraved with the cavity number. Any mold modification, raw material change or process change requires our written approval in advance.)

7. Three Real Cases

These are anonymized versions of situations we see often in consultations. Amounts and figures are examples.

Case 1 — Airtight Container Lids

This is the case from the top of the article. The factory agreed to modify the problem cavity, and from the next run the lid's inner diameter and the base's outer diameter were designated critical dimensions with their own tolerances. Cavity numbers were also engraved into the lid mold. The mold modification cost about 4,000 yuan, but the labor to sort and repack the 960 sets already delivered, plus the lost sales from the delay, cost several times that. The same defect did not appear in later runs.

Case 2 — Sheet Metal Bracket Screw Holes

A buyer sent the factory a complaint that the screw hole spacing on steel shelf brackets was "1mm shorter than the drawing". The factory replied with measurement photos insisting its values were correct. It turned out the factory was measuring between the outer edges of the two holes with a caliper, while the drawing dimension was center-to-center. They were measuring different things. Once the drawing said "hole center-to-center distance, measured with a vision measuring system" and included a sketch of where to measure, the dispute never came back.

Case 3 — Aluminum Phone Stand

Right after CNC machining, the sliding parts fit together smoothly, but the finished, anodized units were too tight to move easily. The anodic coating had built up on both mating surfaces and closed the gap. Changing the drawing to "dimensions after anodizing" and having the machining step leave room for the coating thickness fixed it. For any part that gets a surface treatment, state which point in the process the dimensions apply to.

8. Common Mistakes

• Ordering with only sample photos and "按样品", with no drawing
• Writing a single general tolerance line and never assigning tolerances to mating dimensions
• Approving a mold based on samples from one cavity only
• Not defining measuring locations, tools and conditions, so your numbers and the factory's don't match
• Not stating whether dimensions apply before or after surface treatment
• Keeping pre-shipment inspection to appearance and never requesting dimensional checks

9. Dimensional Tolerance Inspection Checklist

Drawing and Contract Stage

• Sent the factory a 2D drawing and 3D file with version and date
• Stated the general tolerance class on the drawing
• Designated mating dimensions as critical and gave them individual tolerances
• Specified measuring locations, tools, conditions and before/after surface treatment
• Included advance approval for mold modifications and material changes

Sample and First Article Stage

• Received a full dimensional report for each cavity
• Measured critical dimensions myself and compared them with the factory report
• Assembled the parts with their mating parts
• Cavity numbers are engraved on the parts

Production and Shipment Stage

• Received patrol measurement records for critical dimensions during production
• Included critical dimension measurement in the pre-shipment inspection
• Drew samples so cavity numbers were evenly mixed

Wrapping Up - Start by Writing Down One Mating Dimension

Dimensional defects get caught later than cosmetic ones, often after thousands of parts have already been made. Compared with reworking a mold and sorting inventory, writing a few more numbers on the drawing and getting samples from every cavity is a small effort.

There's something you can do right now. Open the drawing for a product you have in progress and check whether the dimensions that mate with other parts have their own tolerances. If they are covered only by a general tolerance line, build a list of critical dimensions before your next order.

StartMade supports you on the ground in China with drawings and specs that mark critical dimensions, review of per-cavity full dimensional reports, sample measurement and fit checks, and on-site dimensional inspection before shipment. If you already have assembly problems, we start with root cause analysis; if you are preparing a new mold, we start with the tolerance standards.