When the Drawing Is Tighter Than the Machine: Setting Specs a Factory Can Hold

Every avoidable return on a listing traces back to a promise the product could not keep. Sometimes that is a design fault. Far more often it is a number on a drawing — a tolerance so tight that the factory cannot hold it on a normal day, on a normal machine. A Amazon FBA inspection China visit is frequently the first time anybody reads the drawing and the real run together, and after a few hundred visits the pattern becomes impossible to unsee.

🎯 The Rejection That Started on the Drawing

Here is the sequence. A buyer writes a tolerance that sounds careful, the factory quotes without comment, the first run lands with a scattering of units outside the window, and everyone blames the line. The line was doing what the machines could do. The drawing had asked for something else.

A tolerance is a window you allow. Process spread is the variation the line actually produces, unit after unit, hour after hour. When the window is narrower than the spread, a share of every run falls outside it — by arithmetic, not by negligence. Nobody has to be careless for a batch to miss a spec that never fitted the process in the first place.

🧮 Where the Window Usually Goes Wrong

Each of those rows is a decision, not a law of physics. The fix is to decide which dimensions carry the promise and let the others breathe.

🔧 Six Steps to Set a Window the Factory Can Hold

1. Pick the dimensions that carry the promise. Assembly fit, safety, a claim you make in the listing. Keep the controlled list to five to twelve dimensions, not fifty.
2. Measure the process on a real run. Twenty-five to fifty consecutive units off one line on one day, not the single golden unit everybody keeps in the office.
3. Set the window from that spread plus the drift you expect across the whole run and the season, not from one quiet Tuesday.
4. Write the same window in four places: the drawing, the spec sheet, the approved-sample record and the factory's own check sheet. A number that lives in only one of them is not controlled.
5. Fix the measurement. Name the gauge, the datum and the points, so two inspectors reading the same part land on the same number.
6. Re-check on the next lot. A shift means the process changed — tooling, mould, sub-supplier, material batch — not that luck ran out.

📊 Spending the Tolerance Budget Where It Pays

A drawing with three controlled dimensions that hold is worth more than one with forty that do not. Where sampling is part of the brief, the sample size still comes off the lot quantity, and the AQL calculator keeps the report and the PO quoting the same table.

💰 A Window in the Field

An importer of 2,800 height-adjustable laptop risers, FOB $12.40, had written the leg's inner tube to a window of plus or minus 0.4 mm. Measuring forty consecutive units from one line showed the line's own spread running to roughly 0.55 mm, so a measurable slice of every run was never going to fit. The assembly that actually mattered turned on one dimension, the outer rail fit; the other number had been copied from an old drawing nobody could date.

The controlled list was cut from eleven dimensions to four, the functional one kept its tight window and the decorative one was widened to what the process could hold. The following lot came in at 1.2% majors, with the drawing, the spec sheet and the factory's check sheet finally quoting one set of numbers. Two visits at $169 per man-day, reports within 24 hours of each, against a stream of returns that had looked random for months. That is the ordinary saving; across 2,000+ inspections a year we see the same story repeated under different part numbers.

🚫 What an Inspection Cannot Do About a Bad Tolerance

The honest boundary matters here. An inspection measures units and reports numbers; it cannot rewrite your drawing, and it should not pretend a single visit is a formal capability study. We can sample twenty-five to fifty units and tell you the spread we see and where the batch sits against your window. Turning that into ongoing process control is the factory's job, and a report that says otherwise is overselling.

🧾 Four Lines for the Spec Sheet

• Controlled dimensions listed. The handful that decide fit, function and safety, with their windows.
• Measurement method named. Gauge, datum and points for each controlled dimension.
• The window repeated everywhere. Drawing, spec sheet, sample record and factory check sheet all agree.
• Next-lot re-check stated. A shift in the numbers is treated as a process change, not bad luck.

A spec sheet that a factory can actually hold is one of the cheapest quality tools there is. Our published pricing keeps the visit cost visible before you commit, and a short note about the order is enough for us to say which dimensions are worth controlling.

❓ FAQs

How tight should a tolerance be?

As tight as the process can hold with margin for drift across the run and the season, and no tighter. Measure the spread on a real run first; the drawing should follow the evidence rather than lead it.

Is a tight tolerance not simply better quality?

Not when it is tighter than the process can hold. Past that point it stops describing quality and starts generating rejects, re-spray and arguments with a factory that is building to capability, not to your wish.

How do I know whether a failure is the line or the drawing?

Measure a short consecutive run and compare the spread to the window. If the units cluster inside the spread and a share sits outside the window, the drawing is the problem. If the spread itself is wild, the process is.

Can an inspection set my tolerances for me?

No, and it should not claim to. An inspection reports the measurements it sees on the sample and where those sit against your window; deciding which dimensions to control, and how wide, stays with you and your factory.

What does it cost to check dimensions before shipment?

Visits start at $169 per man-day, with the report typically issued within 24 hours. Dimensional work usually rides inside a standard pre-shipment visit once the controlled list is short enough to measure properly.