How to brief an OEM factory so the first sample is right
Why the first sample usually fails
The first sample is rarely wrong because the factory ignored the drawing. It fails because the brief was incomplete in at least one of four areas: geometry, tolerance, material or finish. The factory quotes and builds to what the drawing and the purchase order actually say, and a dimension that is not stated is left to the factory default. The sample then reveals the gap, and the review becomes a negotiation about what was intended rather than an inspection of what was built.
The second common cause is a mismatch between the drawing and the process. A feature that cannot be formed, welded or coated as drawn will be interpreted by the factory in the most practical way, which may not be the way the designer imagined. A bend radius that is too tight for the alloy, a hole that overlaps a weld, or a finish that cannot cover a sharp internal corner will all be resolved on the shop floor unless the brief addresses the process constraint up front.
The third cause is the absence of a decision maker. When the brief is assembled by several people and no one owns the final answer, the factory receives conflicting instructions and resolves them in the order it receives them. A single named technical contact on the buyer side, with authority to approve a deviation, removes most of the ambiguity that produces a wrong first sample.
Drawing standards: formats, dimensioning and CTQ flags
The drawing package should include a 3D model in a neutral format and a 2D drawing that carries the dimensions and tolerances. STEP AP214 or AP242 is the most widely readable 3D format for furniture hardware and metal parts, and it should be exported from the native CAD file rather than converted by a third party. DXF or DWG is acceptable for flat parts, but a 2D-only package leaves the factory to infer the third dimension, which is a common source of fit errors.
The 2D drawing should use a consistent dimensioning scheme. Baseline or ordinate dimensioning is easier to inspect than chain dimensioning because every critical feature is referenced to a single datum, and the datum should be identified on the drawing and matched in the inspection report. Where a feature is symmetrical, the symmetry should be stated rather than assumed, and where a dimension is not critical, it should be marked as reference so the factory does not spend capability on a non-functional requirement.
Critical-to-quality features should be flagged explicitly, with the tolerance and the inspection method. A hole pattern that must align with a mating part, a weld that carries load, a surface that must be flat, and a finish that will be visible in the assembled product are the typical CTQ items. Marking them on the drawing, listing them in a table and repeating them in the purchase order ensures the factory cannot treat them as optional.
Tolerance stack-up: specify what matters instead of defaulting everywhere
A general tolerance block is a starting point, not a substitute for analysis. If every dimension carries the same general tolerance, the factory will produce within it, and the assembly may still fail because the stack-up of several dimensions exceeds the clearance between mating parts. The brief should identify the functional interfaces, calculate the worst-case stack for each, and tighten only the dimensions that contribute to the failure mode.
Statistical stack-up is often a better tool than worst-case when the process is capable and the distributions are known. A worst-case calculation assumes every dimension is at its limit in the unfavorable direction, which is conservative but can force unnecessarily tight tolerances and increase cost. A statistical calculation using process capability allows a realistic assessment of the assembly outcome, provided the factory can demonstrate the capability that the calculation assumes.
For sheet-metal and welded components, a general tolerance of ± 0.25 mm on hole-to-hole and mating-face dimensions is a workable baseline, with tighter callouts reserved for the interfaces where the fit genuinely requires them. Applying a tight tolerance everywhere raises cost and slows inspection without improving the assembly. The brief should also state the measurement method for each critical dimension. A dimension measured with calipers has a different uncertainty than the same dimension measured on a coordinate measuring machine, and the acceptance decision depends on the method. Where a tolerance is tight enough that the measurement method matters, the drawing should specify the instrument and the number of readings, and the inspection report should record the actual values rather than a pass or fail statement.
The BOM as a contract: material, finish, hardware and alternates
The bill of materials is the document that turns the drawing into a procurement instruction. It should name the material grade and temper, the finish specification and its standard, the hardware source and part number where the buyer has a preference, and the approved alternates where the buyer is willing to accept a substitute. A BOM that says stainless steel without naming the grade leaves the factory to choose, and the choice may be 201 rather than 304.
Finish specifications should reference a published standard and include the measurable requirements. Anodize should specify the type, class, color and minimum thickness, and powder coat should specify the chemistry, the film thickness range and the gloss target. Where the finish is visible, the BOM should also state the color standard and the acceptable difference, because a color that is described in words cannot be inspected objectively.
Hardware deserves the same treatment as the main structure. A screw that is specified by size but not by material, drive type or coating can be substituted with a lower-grade part that fails in service. The brief should state the required corrosion resistance, the torque requirement and the approved brands or part numbers, and it should require the factory to obtain written approval before substituting any hardware item.
The sample inspection report: CMM data, finish verification and photo evidence
The sample inspection report should be a dimensional record, not a certificate of approval. It should list every CTQ dimension, the nominal value, the tolerance, the actual measured value and the instrument used. A coordinate measuring machine report is the strongest evidence for complex geometry, but it is only useful if the datum scheme matches the drawing and if the report identifies the program and the part revision that was measured.
Finish verification should include the measured film thickness or anodize thickness, the color reading against the standard, and a gloss measurement where applicable. A visual statement that the finish matches is not enough, because the sample is usually inspected under different lighting than the production parts will be. Photographs should be taken with a scale and a color reference in the frame, and they should cover the assembled product as well as the individual features that were measured.
The report should also record any deviation from the drawing, with the reason and the approval. A deviation that is recorded and approved is a controlled change; the same deviation discovered later in production is a defect. The buyer should review the deviation list before approving the sample, because approving the sample approves the deviations that were documented with it.
A reusable brief template and the review loop
A reusable brief follows a fixed structure: project information, product description and use case, 3D model and drawing package, CTQ table, tolerance stack-up summary, BOM, packaging and labeling requirements, inspection and test requirements, and the approval and change-control process.
The review loop should be defined before the sample is built: a design review of the brief by the factory, a first-article build, a joint inspection against the CTQ table, a documented deviation list, and either approval or one round of corrections. The target is two rounds to approval, and a process that routinely needs four or five rounds usually has an incomplete brief rather than a difficult factory.
The loop closes with a signed approval that freezes the drawing revision, the BOM revision and the approved deviations. Production should not start until the frozen revision is issued and acknowledged, and any change after that point follows the same change-control process.