A first-off is valuable only when it represents a defined repeat population and is inspected while the decisions that matter—including concealed work—are still observable.
The factory presents one neat casement as the ‘first article’ for a project containing fixed lights, tilt-and-turn windows, sliding doors, different glass build-ups and two finish routes. The unit passes a visual check and production accelerates. Days later, the shared handle preparation is wrong on one hardware family and reinforcement evidence is missing from already closed frames. The problem was not that the first unit looked poor. It was that nobody defined which risks it represented, which features required upstream evidence, or what release the inspection was allowed to grant.
1. Form first-off groups from shared failure routes—not catalogue families
Start with the released opening-mark register and configuration matrix. Group units by characteristics that can change fabrication or inspection: system and profile family, fixed or operable type, handing, size/weight risk, reinforcement, corner/joint route, drainage machining, glass build-up, finish, seals, hardware family, screens/accessories and any destination-specific evidence route. Record which characteristic creates each group; ‘windows’ and ‘doors’ are too broad.
Then map shared production inputs: extrusion or finish batch, machining program and fixture, hardware preparation, assembly cell or shift, glazing route, subcontracted process and inspection method. Two visually different units may share one high-risk machining program; two identical elevations may require separate first-offs after a material, machine, operator, revision or process change.
Define requalification triggers before production. A revised drawing or BOM, substitute component, new material batch where relevant, program/tool/fixture change, process move, prolonged interruption, repaired first-off or recurring NCR may require a repeat or partial first-off under the approved quality plan. The trigger should identify the affected characteristic and population rather than automatically restarting everything.
| Group driver | Released source | Shared process/input | Representative mark | Population represented | Repeat trigger |
|---|---|---|---|---|---|
| Operation/handing | Opening schedule and drawing | Machining and hardware route | [mark] | [marks/quantity] | Revision/program/hardware change |
| Size/mass boundary | Approved configuration evidence | Cutting, assembly and hardware loading route | [mark] | [risk group] | Configuration change |
| Glass/build-up | Glass schedule | Glazing and bead/gasket route | [mark] | [marks] | Glass or glazing-route change |
| Finish/material batch | Finish/sample/BOM | Supplier/batch/process | [mark] | [marks] | Approved trigger |
| Special interface/accessory | Detail and schedule | Preparation/assembly step | [mark] | [marks] | Detail/component change |
2. Build the inspection around last-observable evidence
Create a requirement-to-evidence plan before the unit is built. For every characteristic, name the released criterion and revision, inspection stage, method, equipment or record, result format and authority. Separate direct measurement, visual observation, identity reconciliation, witnessed operation and upstream retained evidence. A generic checklist should not invent requirements that the project sources do not contain.
Place hold or witness points before concealment: reinforcement and internal connectors before closure; machining and drainage preparation before assembly hides access; corner preparation before seals or finishes obscure it; glass, blocks, gaskets and sealant during glazing; hardware identity and fastening before covers; labels and accessories before packing. Photographs need unit/part identity, orientation, scale or measurement context and source linkage—not merely a tidy close-up.
Control measuring resources for the intended use. ISO 10012:2003 addresses measurement-management systems and metrological confirmation; ISO 14253-1 addresses decision rules when verifying conformance with specified limits. Neither supplies a window tolerance or lets an inspector ignore measurement uncertainty. The approved plan must define equipment suitability, checks/calibration or verification, environment, datum, method and decision rule.
| Characteristic | Released criterion/revision | Last observable stage | Method/equipment or record | Result and limitation |
|---|---|---|---|---|
| Identity/configuration | Drawing, schedule and BOM | Material issue through final assembly | Bidirectional reconciliation | Match / mismatch / unknown |
| Concealed work | Section/detail/process source | Before closure | Identified in-process record | Observed scope only |
| Machining/preparation | Controlled machining drawing/program source | Before hardware/assembly hides it | Approved gauges/measurement | Actual result and criterion |
| Geometry/joints | Approved drawing/tolerance/process source | Assembly before downstream masking | Controlled datum/method | Actual result and uncertainty rule |
| Operation/hardware | Hardware schedule/instruction | Safe completed operating state | Approved functional sequence | Observed result/limitations |
| Glazing/finish/accessories | Schedules and approved samples | Relevant in-process/final stage | Identity and defined observation | Result/exception |
3. Inspect one identity through the full production chain
Freeze the first-off identity and starting sources. Verify that materials carry accepted status and link to the correct BOM. Follow cut parts, machining, subassembly, frame/sash assembly, glazing, hardware, operation and labelling without swapping the representative unit. If a part is replaced, record the old/new identities and reason rather than presenting only the successful replacement.
Capture actual values and conditions before writing pass or fail. For a dimension, record feature, datum, point, instrument ID, reading, unit, requirement source and applicable decision rule. For an attribute, record the physical feature, observation, photo/evidence reference and criterion. For operation, identify sash/panel, installed hardware configuration, support state and exact approved check. Do not adjust repeatedly until the symptom disappears without preserving the original state and authorised work.
Challenge adjacent effects. Correct profile dimensions do not prove joint quality; correct hardware code does not prove hole location or engagement; smooth operation can hide forced adjustment or out-of-square geometry; a clean final photograph cannot recreate reinforcement or drainage preparation. Link the evidence set across stages and mark anything not observed as unverified, not passed.
| Stage | Input/status gate | Observed item and result | Evidence ID | Exception/next hold |
|---|---|---|---|---|
| Material issue | Accepted BOM-linked lots | Actual codes/batches/quantities | [record] | Hold mismatch |
| Cut/machine | Current program/drawing/tool status | Actual characteristics | [record] | Contain shared process |
| Subassembly | Concealed-work sources | Connectors/reinforcement/preparation | [record] | Do not close until disposition |
| Frame/sash assembly | Approved process and geometry criteria | Joints/geometry actual results | [record] | NCR/reinspect |
| Glazing/hardware | Released schedules/instructions | Identity/workmanship/operation | [record] | NCR/reinspect |
| Final identity | All upstream evidence reconciled | Mark/label/completeness/status | [record] | Release or hold |
4. Disposition the first-off and its represented population separately
Use a signed disposition with one of the project's defined states: released for a named population and revision; released with explicitly closed corrections and repeat evidence; held pending named actions; rejected/non-conforming; or another authorised state. ‘Sample approved’ is too vague. Name the permitted production operations and quantity/population, active holds and any characteristic not represented.
If the first-off fails, preserve the observation, stop or contain the shared process, identify all material and WIP since the last verified state, and issue an NCR. The cause may be local, program-wide, batch-related or unresolved. Do not assume that repairing the first unit clears other parts; inspect the potentially affected population and repeat the first-off under authorised scope after the correction.
Before repetition, verify production-floor sources, program/tool/fixture identity, material status, operator/work-centre instruction and inspection points against the released state. Then monitor early repeats at the frequency and scope defined by the approved plan. First-off approval confirms bounded evidence at a moment; it does not guarantee the rest of the batch.
| Decision field | Required entry | Unsafe shortcut |
|---|---|---|
| Representative identity | Exact unit/parts and evidence package | Best-looking unit |
| Population | Marks, quantity, configuration and revision represented | All similar windows |
| Disposition | Released / held / NCR under project states | Verbal go-ahead |
| Permitted work | Explicit operations and conditions | Production can proceed |
| Exceptions | Unrepresented features and open hold IDs | Blank means accepted |
| Repeat trigger | Defined change/event and authority | Only repeat after customer complaint |
FAQ
Is one first-off enough for an entire window order?
Only if the authorised risk grouping shows it represents every relevant configuration and shared process. Mixed systems, hardware, glass, finish or process routes usually require separate coverage.
Should the largest window always be the first-off?
Not automatically. Select boundary and failure-relevant configurations from approved risk analysis; size is one driver among operation, hardware, glass, joints, finish and process route.
Can photographs replace first-off measurement records?
No. Photos can support identity and observed condition, but dimensions need controlled datums, methods, equipment, actual readings, criteria and decision rules.
What happens after a first-off is repaired?
Preserve the original failure, authorised disposition and repair evidence; contain the connected population; reinspect affected characteristics and repeat the defined first-off gate before release.
Does first-off approval release shipment?
No. It is a production-stage disposition for a defined population. Final product, packing, loading, documents and shipment have separate gates.
Does ISO 2859-1 define first-off sample size?
No. ISO 2859-1 concerns AQL-indexed lot-by-lot attribute sampling. First-off grouping and coverage require a project-specific risk and process plan.
Continue the project review
Connect this guide with the proposed window and door systems, then review how Cavorlen approaches manufacturing and quality control. When your schedule is ready, send it through the project review form.
Official sources used
ISO — ISO 9001 quality management systems overview ↗ISO — ISO 10012:2003 measurement management systems ↗ISO — ISO 14253-1:2017 decision rules for verifying conformity ↗ISO — ISO 2859-1:2026 sampling procedures for inspection by attributes ↗ISO — ISO 10007:2017 quality management and configuration management ↗Related buyer guides
This article is general first-off inspection and production-control education. It is not a product specification, inspection and test plan, sampling plan, engineering review, conformity assessment, certification, commissioning or shipment release. It does not select a window system, specimen, AQL, sample size, defect class, tolerance, measurement value, test pressure, load, tool, adjustment, destructive method or acceptance criterion, and it does not state any Cavorlen design, factory process, capacity, quality result or release practice. The approved drawings and BOM, governing specifications and standards, exact system/component instructions, contract quality plan and named authorised design, technical, quality and production parties must define and disposition the actual first-off scope.