Eyeshadow Palette Injection Molding Defects: How Do Buyers Catch Sink Marks, Flash, and short Shots Before Shipment?

Sink marks, flash, and short shots account for the majority of rejections on injection-molded eyeshadow palettes and powder compacts, and most are catchable before shipment with a disciplined routine: 100% visual checks under 1000–1500 lux, a weight check against the approved sample (typically ±3%), flatness verification on a surface plate (usually ≤0.5 mm across the closing face), and dimensional spread measurements across at least 10 randomly drawn units. Buyers who write numeric acceptance limits into the purchase order — flash ≤0.1–0.2 mm on non-critical edges, sink depth ≤0.1 mm on visible surfaces, parting line step ≤0.1 mm — convert vague "cosmetic quality" arguments into pass/fail decisions a factory line can execute without debate.

This guide is written for beauty brands, private-label buyers, and packaging procurement teams who source injection-molded eyeshadow palettes, eyeshadow boxes, and powder compacts but do not run molding machines themselves. It explains what each defect family is, why the process produces it, which surfaces it matters on, what acceptance limits belong in your PO, how to run T1 and first-article inspections, how AQL sampling per ISO 2859-1 works, and how to document incoming defects so corrective action actually happens. Practical verification methods are woven through each section.

What Injection Molding Defects Mean on Palette and Compact Housings

Injection molding is the dominant process for eyeshadow palette bases, lids, powder compact housings, mirrors frames, and internal pans because it delivers tight tolerances at high volume at low piece cost. Molten thermoplastic — most commonly ABS, PMMA, or PP for cosmetic housings — is injected into a steel mold under high pressure, cooled, and ejected as a finished part. The speed and economics of the process are exactly why defects matter: when a molding parameter drifts, thousands of units can be produced and partially packed before anyone notices. Unlike a sewn pouch or a printed carton, a molded housing with a dimensional defect usually cannot be reworked; it can only be sorted, regraded, or scrapped.

For a beauty brand, defects in palette housings create three categories of risk. Cosmetic risk: sink marks, splay, and weld lines on the A-surface — the consumer-facing top shell under the decoration or pressed powder — make a mid-market product look discount-tier on a shelf. Functional risk: flash inside the pan recess prevents pans from seating flat, a warped base means the lid does not close with a satisfying click, and short shots leave features unfilled. Assembly and filling risk: a housing 0.3 mm out of tolerance on the pan pocket depth can jam an automated pan-insertion line, turning a packaging problem into a filler stoppage. The defect vocabulary is therefore the shared language you and your supplier need to argue less and ship more.

It also helps to distinguish a defect from a process variation. Every molded part has a gate mark where plastic entered and faint witness lines where the mold halves meet; these are inherent, placed on hidden surfaces by design. A defect is a feature that exceeds the agreed limit — flash thicker than the PO allows, a sink mark visible at arm's length under retail lighting, or a short shot that removes required material. Buyers who grasp this distinction stop demanding the impossible ("zero parting line") and start demanding the enforceable ("parting line step ≤0.1 mm, gate trimmed flush to +0.1/–0 mm").

How the Injection Molding Process Creates Each Defect Family

Every common defect on a palette housing traces back to one of four root causes: too much or too little material packed into the cavity, uneven cooling, material contamination or degradation, or a tooling problem. Knowing the family tells you what corrective action to demand and whether the fix is a parameter change (same day) or a tooling modification (one to three weeks). The subsections below cover the eight families that matter most on eyeshadow palettes, eyeshadow boxes, and powder compacts.

Sink Marks

Sink marks are shallow depressions caused by localized shrinkage: where the molding has a thick section — a rib, a screw boss, a compact hinge, or a reinforced mirror rim — the outer skin cools first while the core keeps shrinking, pulling the surface inward. On glossy ABS lids, sinks typically appear above internal ribs or aRound mirror bosses, and are most visible when light rakes across the surface at an angle. Acceptance practice on visible surfaces is sink depth ≤0.1 mm, verified with a profilometer or a boundary sample comparison. Process fixes are packing-pressure and holding-time increases, cooling-time extension, and coring out thick sections where the design allows. If sinks persist after parameter changes, the root cause is wall geometry, and the fix belongs in the DFM review, not on the molding floor.

Flash

Flash is a thin web of excess plastic that escapes the cavity at the parting line, around ejector pins, or through slider clearances, appearing as a sharp fin on the edge of the base or lid. It results from excessive injection or holding pressure, worn or mismatched parting surfaces, trapped material on the clamp face, or insufficient clamp tonnage. On cosmetic housings flash is doubly problematic: it is visible and sharp — a cut risk on a product consumers handle near the face — and it interferes with closure fit, so a palette with flash on the hinge side may not snap shut or may pop open in transit. Common limits are flash ≤0.1 mm on non-critical edges and ≤0.2 mm on functional hidden edges, trimmed flush without stress marks or knife scars. Buyers should also require controlled deflashing tooling rather than hand knives, which are where scratches and inconsistent edges enter the process.

Short Shots

A short shot is an incomplete fill: the plastic solidifies before reaching the far end of the cavity, leaving a missing corner, an unfilled embossed logo, or a partially formed rib. Short shots concentrate in thin-wall regions — embossed brand marks, decorative flanges, deep pan pocket corners, and last-filled areas of long-flow covers. Root causes include insufficient injection speed or pressure, low melt or mold temperature, inadequate venting, and walls too thin for the flow length. Acceptance is unambiguous: a short shot on any designed feature is a critical defect and the unit fails, since no cosmetic grading makes missing plastic acceptable. The buyer-side control is simple but often skipped — define which features are "designed features" in the drawing (logo emboss depth, rib heights, flange widths) so inspectors know any partial fill there is a fail, not a judgment call.

Weld Lines (Knit Lines)

Weld lines form where two or more flow fronts meet and fuse — typically around holes such as mirror windows, or where the mold uses multiple gates. Because the flow fronts have cooled slightly before meeting, the weld can be a visible hairline and a local weak point. On a lid with a mirror window, the weld line downstream of the window is a classic location: it may show as a faint line on a glossy surface and, in severe cases, become the origin for a crack in a drop or flex event. Acceptance on visible surfaces is graded — a faint line that does not catch a fingernail or appear under retail lighting may pass as a minor, while a prominent or weak weld is a major. Fixes include raising melt and mold temperature, faster injection, gate repositioning, and vents at the weld convergence point. Buyers evaluating samples should always flex-test the weld region: hold the lid on both sides of the mirror window and flex gently — a crack or white line indicates a weak knit.

Splay (Silver Streaks)

Splay appears as silver or whitish streaks radiating from the gate direction across the surface, most visible on dark and glossy colors — which is why black and deep-plum lids show it so often. The root cause is moisture or gas: wet resin releasing steam during injection, degraded material from excessive melt temperature or residence time, or trapped air at the gate. Some buyers assume splay will be hidden under printing or spraying — it will not; coating highlights splay rather than hiding it. The fix is disciplined material handling: drying resin to the moisture specification, purging old color changes completely, and keeping melt temperature and screw speed in the validated window. Buyer controls: require moisture-check records for hygroscopic resins at lot start, and make A-surface splay a defined major defect rather than an argument at final inspection.

Warpage

Warpage is dimensional distortion — the base bows, twists, or cups so that it no longer sits flat or closes flush. It arises from differential shrinkage: when one region cools faster or shrinks more than another, internal stresses pull the part out of shape after ejection, sometimes hours or days later. Flat palettes with large unsupported spans and housings with asymmetric walls (a solid hinge boss on one side, thin walls elsewhere) are the classic offenders. Warpage shows up as a visible gap between base and lid when closed, a rocking motion on a flat table, and stress cracking at the pan pocket over time. The standard verification belongs in every buyer's protocol: place the unit on a surface plate and measure the maximum gap with feeler gauges — acceptance is commonly ≤0.5 mm across the closing face for a lidded compact, tighter (≤0.3 mm) if the housing is also the sealing surface for a powder pan. Fixes include balanced cooling, longer uniform cooling, optimized ejection, and uniform wall thickness at the design stage.

Burn Marks (Gas Traps and Die Burns)

Burn marks appear as dark, charred-looking discoloration, usually where the cavity fills last or at dead-end corners where air cannot escape. When the flow front compresses trapped air at high speed, the air degrades or ignites the resin locally — this is "dieseling" or gas trapping. On dark palettes burns are easy to miss in quick checks; on light colors they show as gray or brown smudges near ribs and deep corners. Root causes are inadequate venting, excessively fast filling, or degraded resin. The fix is primarily tooling (adding or cleaning vents at the trap) combined with a multi-stage profile that slows near fill completion. Buyers should require that last-filled corners and deep rib ends be checked specifically, because a glance at the face will not catch a burn hidden inside a pan pocket corner.

Gate and Ejector Marks

Every injection-molded part has a gate (where plastic entered) and ejector pin witness marks (where the part was pushed out). These are not defects in themselves, but become defects when oversized, protruding, or badly located on a visible surface. A protruding gate vestige inside a palette base can prevent a pan or magnetic sheet from seating flat, and a tall ejector mark under a lid can scratch the mirror during assembly. Acceptance limits should be written as geometry, not opinion: "gate vestige trimmed flush, ±0.1 mm relative to local surface; ejector marks flush to +0.15 mm, not on A-surface except where shown on the approved drawing." One more control matters: mold cavity numbering. A discreet cavity identifier on every part lets defects be traced to a specific cavity, turning "the whole shipment is bad" into "cavity 3 is producing flash, fix cavity 3."

Which Defects Matter on Visible A-Surfaces Versus Functional Surfaces

Professional cosmetic packaging inspection starts with surface classification. The A-surface is anything the consumer sees or touches in normal use and on the shelf: the palette lid exterior, the decorated top shell, the front edge of the base, and the mirror frame border. The B-surface is visible only in use — the lid interior around the mirror, the pan pocket, inner lips. The C-surface is never seen: internal ribs, the bottom of the base, areas covered by a pan or label. Each class carries a different defect tolerance, and a buyer who applies one uniform standard to all surfaces will either reject good shipments (standard too strict for C-surfaces) or ship bad ones (too loose for A-surfaces).

On A-surfaces, the working rule for cosmetic packaging is "nothing visible at 45–60 cm under normal retail lighting." Concretely: no sink marks deeper than about 0.1 mm; no splay or silver streaks; no prominent weld lines; no untrimmed flash; no scratches longer than roughly 10–15 mm or wider than a hairline (calibrated with a signed boundary sample); no gate or ejector evidence. On B-surfaces — visible only in use, such as the lid interior, pan pocket, and inner lips — functional defects still fail, but shallow sinks, faint weld lines, and minor flow marks are acceptable minors. On C-surfaces, only functional problems matter: flash that blocks assembly, short shots, and dimensional out-of-tolerance conditions.

Two defect families deserve special mention because their classification is not obvious. First, weld line strength is a functional issue even on an A-surface: a weak weld across a compact hinge will crack in transit, so weld lines near hinges and snap features should trigger a flex test regardless of appearance grade. Second, warpage is always functional: a warped base is first noticed visually, but its real cost is closure failure, pan seating problems, and rocking displays. Classify it as a major defect wherever it appears, with flatness as the objective criterion.

The practical way to enforce surface classification is a defect location map: at the T1 stage, annotate a photo or line drawing of the part with the A/B/C zones and attach it to the inspection standard. Every inspector — the factory's and yours — then grades the same area the same way. This single document eliminates most final-inspection arguments about whether a mark "matters."

Defect Families, Causes, and Preventive Actions at a Glance

The table below condenses the defect families into a single reference your team can attach to a PO or inspection checklist. Note the pattern in the third column: parameter fixes are fast, tooling fixes take weeks, and design fixes belong in the DFM review before steel is cut — which is why early engagement with the molding supplier is the cheapest quality investment a buyer can make.

Specifications and Acceptance Limits Buyers Should Write Into the PO

An inspection standard is only as strong as the numbers written into the purchase order. Verbal agreements like "no visible defects" produce disputes; numeric limits with a defined measurement method produce decisions. The specifications below cover the bulk of what matters for injection-molded palette and compact housings, and should be tailored to each project with the supplier's engineering team during the DFM review.

• Wall thickness: nominal per drawing with ±0.1 mm tolerance on critical functional walls; uniform nominal (typically 2.0–3.0 mm for palette housings), with DFM confirmation that ribs and bosses do not exceed 60–70% of the adjoining wall to keep sinks within limit.
• Draft angle: minimum 0.5–1° on visible vertical walls, 1–2° on textured surfaces, confirmed at DFM; insufficient draft shows up as drag marks and ejection whitening rather than as a measured dimension.
• Parting line: step or mismatch ≤0.1 mm on visible edges; witness line permitted but flash ≤0.1 mm on non-critical edges (≤0.2 mm on hidden functional edges), trimmed flush with no knife scars.
• Sink depth: ≤0.1 mm on A-surfaces (measured or by boundary sample); sinks on B/C-surfaces acceptable if they do not affect function or assembly.
• Flatness / warp: ≤0.5 mm max gap when the unit is placed on a surface plate over the closing face; ≤0.3 mm where the housing is the sealing face for a pan or a mirror gasket.
• Closure function: lid must open and close through 50 cycles without loosening, cracking, or hinge whitening; magnetic or snap closure must pass a defined pull or drop-retention test.
• Weight check: part weight within ±3% of the approved sample; a falling average weight signals short shots or packing drift, an erratic spread signals machine instability.
• Pan pocket and assembly dimensions: pocket depth, diameter, and flatness per drawing tolerance, because these dimensions govern automated pan insertion and powder filling.
• Material and color: resin grade and color per approved master; ΔE ≤1.0–1.5 against the master under a defined illuminant; no regrind beyond the agreed percentage unless explicitly allowed.
• Traceability: discreet cavity number molded into every part; lot and machine records retained and available on request.

The table below pairs each acceptance criterion with the way it is verified on a packaging line or at a third-party inspection. Share it with the supplier before production starts: a limit the factory does not know how to measure is a dispute scheduled for a future date.

T1 Sample and First-Article Inspection Protocol

The T1 sample — the first articles molded from the newly cut or modified tool — is the single most important quality gate in the project: the first moment the physical part exists and the last moment where changes are relatively cheap. A disciplined T1 protocol protects the buyer from paying for a tool that ships defects for the life of the program. The protocol below assumes the buyer or a third-party inspection agent reviews the T1 batch in person or via a documented remote review.

• Quantity and context: request 10–20 T1 parts per cavity, molded on the production machine at intended production settings — not hand-polished show pieces, and not parts from a different press than will run production.
• Documentation first: obtain the molding parameter sheet (melt and mold temperature, injection speed profile, holding pressure and time, cooling time, cycle time) and the cavity list; these are your baseline for every future quality conversation.
• Full dimensional layout: measure every drawing dimension on at least one part per cavity; flag any dimension with less than 20% of tolerance remaining as a process-capability risk.
• Dimensional spread: for the critical dims (pan pocket, closure interface, flatness, hinge alignment), measure 10 units and record the spread; a tight mean but wide spread predicts future out-of-tolerance parts.
• Cosmetic grading: inspect A-surfaces under 1000–1500 lux using the defect map; photograph and grade every finding against the agreed limits.
• Functional tests: 50-cycle lid test, pan dummy-fit into the pocket, weight check on a calibrated scale, and a flex test across weld regions.
• Approval with conditions: the T1 result is approved, approved-with-comments, or rejected with a written defect list referencing cavity numbers. Never approve a tool verbally; the signed T1 report is the reference for the rest of the program.

First-article inspection (FAI) then repeats a condensed version of this protocol at the start of mass production, after any tool move, mold repair, resin change, or long production gap. The key habit is comparing against the signed T1/FAI records rather than memory: when a supplier says "this is normal for this tool," the correct response is "show me where this appears in the approved sample set."

AQL Sampling per ISO 2859-1 for Palette and Compact Inspections

AQL (Acceptable Quality Limit) sampling per ISO 2859-1 (with the equivalent ANSI/ASQ Z1.4 tables used the same way) is the standard statistical method for deciding whether a finished lot of palettes or compacts passes. The lot is not inspected 100%; a sample size determined by lot size and inspection level is drawn at random and graded against defect classifications. The most common configuration for cosmetic packaging is General Inspection Level II, with critical defects at AQL 0 (or 0.065 at most), major at AQL 1.0–1.5, and minor at AQL 2.5–4.0.

A worked example makes the mechanics concrete. Suppose your PO covers 8,000 eyeshadow palettes. Under Level II, the code letter for a lot of 8,000 (3,201–10,000 range) is J, calling for a sample of 200 units. With major defects at AQL 1.5, the acceptance number is 7: 8 or more sampled units with a major defect fails the lot. With minors at AQL 4.0, the acceptance number is 14. Critical defects — sharp flash, missing plastic, cracks — are set to zero acceptance: a single critical fails the lot. Translate the defect families into those classes in writing: short shots and sharp flash = critical; sink marks over limit, flash over limit, warp over flatness limit, splay on A-surface = major; faint flow marks on B-surfaces, shallow sink on hidden surfaces, slight gate witness = minor.

Three practical notes prevent the most common AQL mistakes. First, sampling must be random across the lot — different cartons, pallet layers, and ideally different shifts; pulling 200 units from the top carton of the best pallet tells you nothing. Second, AQL measures the lot, not the promise: a passing lot with 7 major defects in 200 units still implies roughly 3.5% defectives in the population, which is why demanding retail channels often tighten to AQL 1.0/2.5 or require factory 100% A-surface visual inspection. Third, record which defects were found and where (cavity numbers), because the AQL result is also your corrective-action dataset.

Incoming Inspection and Defect Documentation

Even with a strong factory and a clean T1, incoming inspection at your warehouse or 3PL remains necessary, because molding processes drift, tools wear, and shipments get mixed. The protocol need not be elaborate, only consistent: draw the ISO 2859-1 sample from across all cartons; inspect visuals under 1000–1500 lux against the boundary sample; measure weight on 10 units; flatness on the surface plate for 5–10 units; closure cycling on 5 units; pan pocket fit with an actual pan on 10 units; and record every finding with photos.

Documentation quality determines whether defects get fixed. Every record should include: a close-up photo with a scale reference (ruler or coin), the defect type using the shared vocabulary from this article, the location on the part (referenced to the defect map), the cavity number, the carton and lot, and the count of affected units in the sample. Photos matter more than prose: a supplier that receives "sink marks on lid" may argue; a supplier that receives five photos of sink marks above ribs on cavity 3 and cavity 7, with feeler gauge readings, has an engineering problem to solve. Cavity tracking converts a diffuse complaint into a targeted tooling action — flash "everywhere" is a clamp problem, flash "on cavity 3 only" is a cavity-3 parting surface problem.

Finally, feed the data back on a fixed cadence. A monthly defect Pareto — even a simple count by type — shows whether the program is improving and gives both parties an objective basis for decisions such as adding 100% visual inspection, investing in a tool refinement, or scheduling tool maintenance. Buyers who document well get better priority in a factory's quality queue than those who send unstructured photos.

MOQ, Tooling, and Sampling Expectations: Getting Quality Set Up Before Mass Production

Most molding defects buyers discover at final inspection were actually designed in months earlier. A molded palette's quality is largely fixed at three milestones: the DFM review (before steel is cut), the T1 sampling (first shots from the tool), and the pilot run (first production-scale batch). Buyers who know what to demand at each milestone spend far less time fighting defect disputes later.

DFM review before steel cut. Before tooling steel is machined, the supplier's mold designer should return a DFM report on your drawing: wall thickness uniformity, rib-to-wall ratios, draft angles, gate location and type, weld line prediction map, vents, ejector layout, and expected gate and parting line mark locations. This is your chance to move a gate off a visible surface, thicken a wall that would sink, or relocate an ejector pin off the A-surface — changes that cost days now and weeks after the tool exists. Insist that the DFM includes predicted weld line positions, because you will want to verify them at T1 with the flex test.

T1 timing and what is reasonable. After tooling is cut, a first T1 typically lands two to four weeks later depending on tool complexity and the mold maker's queue; textures, sliders for undercuts, and hot runners extend the timeline. Expect one to three T1 iterations: a first shot to see the real part, a tooling correction round (venting, gate adjustment, polish, dimension touch-up), and a second T1 to confirm. Schedule for it; a plan that assumes first-shot perfection ends in a rushed approval.

Pilot runs and MOQ. Before committing to the full MOQ, request a pilot run — one machine, one shift, a few hundred to a couple thousand units — packed the way production will be packed. The pilot exposes what a T1 cannot: process stability over time (weight and dimension drift across the shift), decoration and assembly interaction with the molded part, and packaging fit. Compare pilot output against the T1 records; if the pilot shows weight drifting downward or flash appearing after 4,000 cycles, the tool needs attention before mass production. Where MOQs feel high, remember they often exist to amortize tooling and setup; negotiating a SMALLer first run with a pilot is frequently more productive than negotiating the MOQ itself.

Two expectations round out the setup phase. First, agree in writing that the signed T1 report, the molding parameter sheet, and the defect location map constitute the quality baseline, and that any process or tooling change afterward requires a new FAI. Second, align inspection logistics early: book third-party inspection dates against the production completion date, and make sure the factory understands the AQL classes in the PO — an inspection arriving at a factory that has never seen your AQL table wastes everyone's time.

Frequently Asked Questions

What lighting and viewing conditions should be used to inspect sink marks and splay on palette housings?

Use a well-lit inspection bench at 1000–1500 lux with neutral white light (about 5000–6500 K). Hold the part at 30–45 cm and rotate it under a raking light — light directed at a low, oblique angle across the surface — because sink marks and splay are far easier to see when light grazes the surface than when it hits head-on. Verify against a signed boundary sample under the same conditions; visual judgments are only consistent when lighting, distance, and reference are all fixed. For disputed units, a profilometer or feeler gauge converts the judgment into a number.

Is flash ever acceptable on an eyeshadow palette or compact?

Flash is never acceptable on visible A-surfaces, and never acceptable if sharp enough to scratch skin — that is a critical defect with zero tolerance. On non-critical hidden edges, residual flash of up to about 0.1–0.2 mm is commonly tolerated provided it is trimmed flush, leaves no knife scars, and does not interfere with closure fit, pan seating, or assembly. Write the limits into the PO with the trimming method: controlled deflashing tooling rather than freehand knives. If flash keeps recurring, the root cause is usually pressure or a worn parting surface — an engineering signal, not a trimming workload.

How many units should I measure to judge dimensional consistency?

A minimum of 10 units drawn at random from across the lot is the practical floor, measured on the critical dimensions: pan pocket depth and diameter, closure interface, flatness, and part weight. Ten units are enough to see the spread — if the range already approaches the drawing tolerance, the process is not stable and the lot is at risk regardless of whether the sampled units pass individually. At T1 and first-article stages, add a full dimensional layout per cavity; during production, a quick 10-unit weight-and-critical-dim check at shift start and every few hours catches drift before it becomes a lot-sized problem.

What is the difference between a critical, major, and minor defect in cosmetic packaging inspection?

A critical defect could injure the user or make the product unusable — sharp flash, cracks, missing plastic from a short shot, a loose mirror, or contamination inside the housing; the acceptance number is zero. A major defect affects appearance or function enough that a consumer would notice — sink marks over the limit on a visible surface, flash over the limit, warp beyond flatness tolerance, splay on the A-surface, or color out of tolerance. A minor defect is a small departure an average consumer would not notice — faint flow marks on B-surfaces, shallow sinks on hidden surfaces, slight gate witness. The classification belongs in the PO, agreed before production, so AQL results are decided by the document rather than email.

Can defects like sink marks and weld lines be fixed after the tool is made?

Yes, but cost and time differ sharply by root cause. Sink marks caused by packing pressure, holding time, or melt/mold temperature usually respond to parameter changes within a day; sink marks caused by wall geometry (thick ribs, heavy bosses, bulky hinges) require tooling changes such as coring. Weld lines respond partially to temperature, speed, and venting, but if the weld sits in a visible or weak location, the lasting fix is relocating gates or adding overflows — one to three weeks including sampling. This is why the DFM review matters: gate location, wall uniformity, and weld line prediction are decided before steel is cut and most influence whether these defect families ever appear.

Should I rely on the factory's internal QC or commission a third-party inspection?

They serve different purposes and work best together. Factory QC runs 100% or in-process checks and is responsible for the lot being ready; it is fast, cheap, and continuous, but grades against the factory's own standard unless yours has been formally transferred to them. A third-party inspection at final random inspection provides an independent sample-based check against your PO standard — the AQL table, defect map, and boundary samples — and produces a written report supporting a ship/hold/rework decision. For molded palettes, a sensible structure is: factory 100% visual on A-surfaces during production, 10-unit weight and critical-dim checks on the pilot, and a third-party final random inspection per ISO 2859-1 for any first order or new tool.

Conclusion

Sink marks, flash, and short shots are not mysteries; they are predictable outputs of pressure, temperature, cooling, and tool condition, and each becomes manageable the moment it is defined with a number. Buyers who ship clean palette programs do three things consistently: they write numeric acceptance limits — flash ≤0.1–0.2 mm, sink ≤0.1 mm on visible surfaces, flatness ≤0.5 mm, weight ±3% — into the PO alongside AQL classes per ISO 2859-1; they gate the program at DFM, T1, and pilot run instead of discovering problems at final inspection; and they inspect under fixed conditions — 1000–1500 lux, raking light, boundary samples, surface plate checks, 10-unit spreads — so pass and fail mean the same thing everywhere. Start with the two tables in this guide.

Adapt the acceptance limits with your supplier's engineers at the DFM stage, attach the inspection-method table to your PO, and build your defect map and boundary sample set at T1. Then document everything you find — photos, cavity numbers, locations, counts — and feed it back on a fixed cadence. Quality in injection-molded cosmetic packaging is rarely won at the final inspection table; it is won in the weeks before steel is cut and confirmed in the first hundred parts off the tool.