A good sample can prove that one acceptable part was made. It does not automatically prove that the supplier has established a repeatable production process.
That distinction matters when a buyer releases a batch after receiving a polished sample, a few selected dimensions or photographs. The sample may have been produced by a different machine, specialist operator, temporary tool, alternative material lot or extra manual rework. Unless those conditions are visible, approval can be based on a result that normal production will not reproduce.
The purpose of first-article inspection is therefore not simply “check one part”. It is to confirm, at an agreed point, that the product definition has been understood and that the planned process can produce verifiable output before production risk expands.
Define what kind of sample is being requested
The words sample, prototype, pre-production sample and first article are often used interchangeably, but they may support different decisions.
- A concept or appearance sample may confirm size, colour, ergonomics or visual direction while using non-production methods.
- An engineering prototype may test fit or function before the design is frozen.
- A process trial may explore tooling, parameters or special processing.
- A first article should normally be produced to an approved baseline using the intended or clearly disclosed production process and should be supported by defined verification records.
State which decision the submission must support. If non-production tooling or manual finishing is allowed, record that limitation. The buyer can still approve a development sample without treating it as evidence for batch release.
IAQG describes the 9102 first-article standard as applying when a new part is introduced or when a change invalidates the original first-article inspection result. It aims to provide consistent documentation and verify that production processes can produce parts conforming to engineering and design requirements. IAQG — 9102 First Article Inspection Requirement That standard is sector-specific and should not be imposed automatically on every order, but its underlying distinction between documented process verification and informal sample approval is broadly useful.
Freeze the baseline before inspection
The submission should identify the part number, drawing revision, model revision, specifications, approved deviations and purchase-order reference. If the sample was made before a design change, it cannot silently approve the new version.
Create a ballooned or otherwise indexed characteristic list so each reported result can be linked to the governing requirement. Include dimensions, notes, material, special processes, marking and any functional tests within the agreed scope. Open questions must be closed or explicitly excluded from release.
The supplier should also declare the manufacturing route: material source or lot, machine or process type, tooling status, external processors and inspection methods. This need not reveal proprietary parameters. It must be detailed enough to show whether the first article represents the planned batch.
Review evidence in layers
An effective first-article package commonly covers five evidence layers.
1. Product-definition evidence
Confirm that the supplier used the correct revision and has accounted for notes, general tolerances, referenced standards and approved clarifications. A dimensional report cannot compensate for a missing drawing requirement.
2. Material evidence
Check the specified grade, condition and inspection document. Link heat, batch or incoming identity to the first-article piece where traceability is required. A generic certificate with no connection to the sample is weak evidence.
3. Manufacturing and special-process evidence
Identify processes that cannot be fully verified by final dimensional inspection: heat treatment, welding, plating, passivation, coating, nondestructive testing or other outsourced operations. Review the required process certificate, parameters or qualification evidence at the level agreed for the project.
4. Dimensional and functional evidence
Report actual values for defined characteristics, not only a global “pass”. Confirm measurement method, datum alignment, part condition and units. Functional tests should include test conditions and acceptance limits.
5. Visual, packaging and identification evidence
Appearance, deburring, cleanliness, marking and protective packaging can be approval characteristics when specified. Use photographs to supplement controlled criteria, not replace them.
Confirm how the part was selected
An unannounced “golden sample” selected after extensive sorting gives a different level of confidence from a part taken from a planned production setup. Ask how many pieces were made, which piece was submitted, whether rework occurred and what happened to the others.
For multi-cavity, multi-machine or multi-tool production, one piece may not represent every source of variation. Define whether the first article covers one setup or all distinct production streams. The correct scope depends on risk and process design.
If the supplier intentionally performs extra operations on the submitted piece, those operations should either become controlled production steps or be removed before approval. Hidden craftsmanship is not a scalable control plan.
Approval is a decision with conditions
Avoid vague messages such as “sample looks good” when the intention is to release production. Use a recorded disposition:
- approved for production;
- conditionally approved subject to listed actions;
- rejected and resubmission required; or
- approved only for development, not batch release.
List every open deviation and its scope. A temporary concession should state affected quantity or lot and expiry. If a characteristic cannot yet be verified, identify who owns the missing evidence and whether production may proceed at risk.
Approval should also define what changes trigger review or a new first article. Typical triggers include drawing revision, material or source change, new tooling, process relocation, long interruption, new external processor or a corrective action that changes the process. The trigger set should be proportionate and contractually clear.
Do not confuse FAI with process capability
A conforming first article is a snapshot. It is not statistical proof that thousands of parts will remain conforming. Stable batch production may require setup approval, in-process checks, tool-life controls, capability studies, lot sampling and final release.
Likewise, a failed first article should not be solved only by reworking the sample. The supplier should identify whether the cause was unclear requirements, programming, tooling, measurement, material, external processing or operator execution, then update the process evidence before resubmission.
ASQ describes quality plans as documents that identify quality standards, practices, resources, specifications and the sequence of activities relevant to a product or project. ASQ — Quality Plans The first article should connect to that wider production control, not sit as an isolated certificate.
Maintain a first-article change log
The approved submission should become a controlled reference, not a folder that nobody revisits. Maintain a concise log of the baseline, approval date, conditions, corrective actions and later changes. When a trigger occurs, decide whether a full, partial or no new first-article inspection is justified and record the rationale.
A partial FAI can focus on affected and potentially affected characteristics, but the impact review should consider interfaces. Changing a coating may affect thickness, electrical contact and thread fit; changing material condition may affect machining and heat treatment. “Only one note changed” does not necessarily mean only one result needs review.
The log also separates the approved production route from later convenience changes, giving purchasing a clear question to ask on every repeat order.
First-article release checklist
- Is the purpose of the submission defined?
- Does it identify the approved drawing, model and specification revisions?
- Were intended production equipment, tooling and external processes used, or were deviations disclosed?
- Is each required characteristic indexed to evidence?
- Are material and special-process records connected to the sample?
- Are measurement methods and actual results clear?
- Were rework, selection and non-production operations disclosed?
- Are all cavities, tools, machines or lines represented where necessary?
- Is the disposition recorded with conditions and open actions?
- Are production quantity, effective lot and change triggers stated?
- Does the batch control plan address variation beyond the first article?
The procurement decision
The release record should be available to purchasing as well as engineering and quality. If the order is placed before conditional actions are closed, the purchase commitment should state the commercial consequence of failure and the point at which further work must stop. This prevents a technical “conditional approval” from being interpreted as unrestricted commercial release.
The buyer should release production only when the evidence answers the intended decision. A beautiful sample with an unclear process is weaker than a representative part with transparent records and controlled corrective actions. The goal is not paperwork for its own sake; it is to stop uncertainty from multiplying into a full batch.
Morning Sunlight Asia can manage first-article baselines, review supplier evidence and follow release actions before batch production in China. Discuss Your Requirement to define an approval package proportionate to the part and its risk.
如果採購方只看精修樣品、少量尺寸或照片就放行批量，樣品可能來自不同設備、資深操作員、臨時治具、替代材料批次或額外手工返修。這些條件若不透明，批准的可能是量產無法複製的結果。
首件檢驗的目的不是「檢查一件」，而是在風險擴大之前，確認產品定義已被理解，且規劃中的製程能產生可驗證的合格輸出。
先說清楚樣品類型
外觀樣品、工程原型、製程試作與正式首件支援的決策不同。外觀樣品可以用非量產方法確認顏色與尺寸；工程原型用於設計驗證；製程試作探索工具或參數；正式首件則通常應依核准基線、使用預定量產方法或清楚披露例外，並附上約定的驗證記錄。
IAQG 說明 9102 首件標準適用於新零件導入，或原首件結果因變更而失效的情況，目標包括一致記錄以及驗證生產製程能否符合工程和設計要求。IAQG — 9102 該標準具有行業適用範圍，不能自動套用到所有採購，但「正式製程驗證不同於非正式看樣」這一原則具有普遍價值。
檢驗前凍結基線
首件包應標出零件號、圖紙和模型版本、引用規格、核准偏差及採購單。可用氣泡圖或索引清單，把每項結果連到尺寸、圖紙註記、材料、特殊工序、標識與功能測試。未關閉問題要明確排除或列為放行條件。
供應商還應說明首件所採用的材料批次、設備或工藝類型、工裝狀態、外協工序和量測方法。無需披露專有參數，但資訊要足以判斷它是否代表計畫中的量產。
五層證據
- 產品定義：版本、註記、一般公差、標準和核准澄清是否完整。
- 材料：牌號、狀態、證書及爐批身份是否與首件相連。
- 製程與特殊工序：熱處理、焊接、電鍍、塗裝或無損檢測等不能只靠終檢尺寸證明的工序，是否有約定證據。
- 尺寸與功能：是否報告實測值，並說明量測方法、基準、零件狀態、單位和測試條件。
- 外觀、標識和包裝：是否按明確準則審查；照片只能補充標準，不能代替標準。
確認樣品如何被選出
從大量產品中挑出的「最好一件」，與從計畫量產設定中正常抽取的零件，證據強度不同。要詢問共製作多少件、提交哪一件、是否返修，以及其他件的結果。多模穴、多設備或多工裝生產時，一件未必代表所有變差來源。
如果首件接受了額外手工作業，該作業要麼成為受控量產步驟，要麼在批准前移除。依賴隱藏的熟練工補救，不是可擴展的製程控制。
批准必須帶有明確處置
不要以「樣品看起來不錯」代替量產放行。應記錄為：批准量產、附條件批准、拒絕並要求重提，或只批准開發而不批准量產。所有偏差要列明批次、數量和期限。
還應定義哪些變更觸發重審，例如圖紙、材料來源、工裝、製程地點、外協方或長期停產後重啟。觸發條件要與風險相稱。
首件不等於製程能力
合格首件只是某一時點的證據，不是幾千件都會穩定合格的統計證明。量產仍可能需要開機確認、製程中檢查、刀具壽命控制、能力研究、批次抽樣和最終放行。
ASQ 將品質計畫描述為識別產品或項目相關品質標準、實務、資源、規格及活動順序的文件。ASQ — Quality Plans 因此，首件應連到整體量產控制，而不是一張孤立證書。
建立首件變更履歷
批准資料應成為受控參考，而不是完成後無人再看的資料夾。建議保存一份精簡履歷，包含產品基線、批准日期、附帶條件、改善事項、關閉證據及後續變更。當觸發條件出現時，應記錄為何需要完整首件、局部首件或不需要重做。
局部首件可以聚焦直接受影響和可能連帶受影響的特性，但必須先做影響分析。例如更換塗層會影響厚度、導電面和螺紋配合；更換材料狀態可能改變切削和熱處理結果。「只改了一條註記」不代表只需重驗一個數字。
履歷還能把原核准製程與後續便利性變更分開。每次重複訂單，採購方可要求供應商確認設備、工裝、材料來源、外協方和檢驗方法是否仍與批准狀態一致。若不同，就先評估，不應等到批量完成後才補做首件文件。
首件放行清單
- 提交目的是否明確？
- 是否引用正確圖紙、模型及規格版本？
- 是否使用預定量產設備、工裝和外協流程，偏差是否披露？
- 所有要求是否有索引證據？
- 材料和特殊工序記錄能否連到首件？
- 量測方法及實測值是否清楚？
- 返修、挑選和非量產作業是否披露？
- 必要時是否涵蓋所有模穴、設備或產線？
- 放行結論、條件和待辦是否被記錄？
- 量產批次與重新首件觸發條件是否明確？
把條件批准轉化為可執行的商務狀態
首件「附條件批准」最容易在部門間失真。工程方的意思可能是完成兩項改善後才能投產，採購系統卻把它理解為可以立即交貨。放行記錄應列出每項條件、責任人、截止日期、關閉證據及未完成時必須停止的工序。若買方在條件未關閉前仍下達訂單，也要明確由誰承擔繼續生產的風險和費用。
首件完成後，核准資料應能被現場實際使用。供應商的流程卡、程式、檢具、外協訂單和檢驗表需要引用相同基線；只把首件報告存入品質資料夾，不會自動改變量產方法。量產啟動時可抽查這些控制點，確認改善事項已真正進入製程。
若首件不合格，重新提交應附上原因、影響範圍、更新後控制和新結果。只提交一件返修合格品，無法證明問題已從製程中消除。對反覆失敗的特性，採購方還應重新評估圖紙清晰度、供應商能力與量產風險，而不是無限延長樣品階段。
首件資料包不應與量產現場脫節
首件報告批准後，應確認關鍵結論已進入量產文件。例如核准的刀具補償、特殊去毛刺方式、遮蔽區、量測治具和外協注意事項，若只留在工程郵件中，換班或換人後就可能消失。可以在量產開機時抽查流程卡、程式版本與檢驗表是否引用正確狀態。
買方還應保存一份可識別的核准樣或數位記錄，但不能讓「金樣」取代圖紙。實物樣會磨損、氧化或因環境變色；數位照片也有視覺限制。當樣品與圖紙衝突時，應由工程方明確決定基線並更新文件，不能讓供應商自行選擇較容易達到的一個。核准樣若被借出、損壞或更換，也應保存身份和批准記錄，避免不同地點各自使用不同參考。
若量產分布在不同設備或外協地點，還要確認每個執行點都能取得相同核准基線，並理解哪些條件只適用於首件試作。這可以防止一個地點按最新要求工作，另一個地點仍沿用早期樣品方法。
所有抽查結果都應保存。
採購判斷
只有當證據真正支援預定決策時才應放行量產。製程不透明的漂亮樣品，弱於一件具有透明記錄和受控改善措施的代表性零件。目的不是增加文件，而是防止小範圍不確定性擴大成整批風險。
晨陽亞洲可在中國境內管理首件基線、審查供應商證據並跟進量產放行事項。歡迎 Discuss Your Requirement，共同確定與零件風險相稱的批准資料包。