供应商审核里对检测能力怎么评价

结论：看的不是有没有设备，是数据可不可信

供应商审核走到检测这一块，常见做法是看一眼设备清单、翻几份检验报告，没什么明显问题就过了。这样的评价没有实际意义。

真正要判断的是：这家供应商给你的检验数据，可不可信。 可不可信取决于六件事——设备、方法、人员、记录、判定、以及数据被质疑时能不能追溯。任何一项缺失，数据的说服力都要打折。

六个评价维度

最后一行是区分度较高的一项。 从一批产品能查到它的检验数据，多数供应商做得到；从一台设备反查它测过哪些批次，能做到的就少了。而这个能力决定了设备一旦被发现失准，影响范围能不能划清楚。

现场该看的记录

不要只看供应商准备好的样板记录，建议现场随机抽取。抽的方式有讲究：

按时间抽。 抽一份半年前的记录，看当时的记录质量与现在是否一致。只在迎审前突击规范的，时间一拉开就露出来。

按异常抽。 问最近一次检验不合格是什么时候，调出那份记录，看处置过程是否完整。正常记录容易做漂亮，异常处置才看得出真实水平。

按设备抽。 挑一台设备，看它的校准证书、期间核查记录，再看它最近测过的几份数据。

按人抽。 挑一个检验员，看他的授权范围与实际签署的记录是否匹配。

不同类型供应商的评价重点

原材料供应商。 重点在批次一致性的证据：他们怎么证明这一批与上一批相同。通常靠出厂检验数据加上过程控制记录。要关注的是检验项目是否覆盖了对你重要的特性——供应商按自己的标准检验，未必测了你关心的那一项。

零部件供应商。 重点在关键尺寸和性能的检验能力，以及不合格品的控制。还要确认他们对下游（他们的供应商）有没有相应管控。

代工厂。 重点在过程检验和成品检验的执行。因为代工厂按你的要求生产，检验清单和判定依据由你给，要确认他们是否按你给的执行，而不是按自己习惯执行。

灭菌等特殊过程服务商。 重点在过程确认与监控记录，以及每批次的放行依据。

评价结论怎么转成管控

审核发现的问题，应当转成具体的管控措施，而不只是写进报告：

检验能力不足的项目，改为你自己进货检验或委外检测；记录不规范的，要求整改并在下次审核复查；追溯能力弱的，在质量协议里补充记录保存和查阅条款；关键特性的检验，要求供应商随货提供原始数据而不只是合格证。

最后一条实际效果较明显。 要求提供实测数据而非合格结论，一方面数据本身有用，另一方面这个要求本身会促使供应商规范检验过程。

分级管理的做法

不必对所有供应商用同一强度的评价。按物料的关键程度分级：

关键物料的供应商，做完整的现场审核并定期复审；一般物料的供应商，可以用问卷加资料审查加抽样验证；低风险物料，主要靠进货检验把关。

分级的依据应当是物料失效对产品安全有效性的影响，而不是采购金额。有些便宜的小零件，失效后果却很严重。

自身检验与供应商检验的关系

供应商做过检验，不等于你可以不做进货检验。两者的定位不同：供应商检验是他的出厂控制，进货检验是你的验证。

可以根据供应商的评价结果调整进货检验的强度——评价好的可以降低抽样比例甚至改为定期验证，评价一般的保持常规强度。这种联动能让评价工作产生实际价值，否则评价做完归档，对日常工作没有影响。

远程审核的局限

近年远程审核使用增多，但检测能力的评价在远程条件下有明显局限：设备的实际状态看不到、记录只能看对方选择展示的、现场环境（温湿度控制、洁净程度）无法判断。

如果只能远程，可以用几种方式部分补偿：要求实时视频巡视设备区域并随机指定查看对象；要求当场调取指定编号的记录而不是事先准备的；安排一次小规模的比对试验，同一批样品双方各测一次，对比数据。

比对试验是远程条件下最有效的手段，因为它检验的是结果而不是表象。

评价周期与触发式复审

除了定期复审，还应当设置触发条件：供应商发生重大变更（搬迁、设备更新、人员大幅变动）、连续出现质量问题、或者其检测数据与你的验证数据多次不符，都应当触发一次额外审核。

只按固定周期审核，中间发生的变化就被漏掉了。

评价结果的记录形式

供应商评价的记录不应当只是一份打分表。建议包含：评价日期与方式（现场/远程/资料）；检查的具体内容与抽查的记录编号；发现的问题及严重程度；要求的整改与时限；结论与下次评价时间。

抽查的记录编号要写进去。 只写「抽查了检验记录，未发现问题」，无法证明抽查的深度；写明抽了哪几份，审核你的人就能判断抽样是否有代表性。

我们的做法

有委托方会请我们对某批物料做独立验证，用于供应商评价或者争议判定。这类委托要特别注意三件事：抽样方式要能代表该批次；检验方法要与供应商用的一致，否则数据不可比；报告要写清楚方法和条件，便于双方对照。

方法不一致是双方数据对不上的主要原因之一。 我们遇到过的情况里，相当一部分并不是谁测错了，而是两边执行的方法条件不同，数据本来就不该相同。

如果你在做供应商评价、需要第三方数据支撑，或者与供应商在检验数据上有分歧，可以把双方的方法和数据发过来一起看，或者直接联系：132 4819 8029。能力范围见服务介绍，送检要求见送检要求，流程见检测流程。

English version

Conclusion. When a supplier audit reaches testing, the usual approach is a glance at the equipment list and a look through a few inspection reports. That carries little meaning. The real question is whether the inspection data this supplier gives you can be trusted, and that rests on six things: equipment, method, personnel, records, acceptance basis, and whether data can be traced when challenged. A gap in any one reduces the weight of everything they report.

Six dimensions. Equipment must exist, be within calibration, and have a calibrated range covering what is actually measured; a common failure is equipment present but calibration expired, or a range that does not cover the readings taken. Methods must have a clear basis and have been validated; a common failure is being unable to say where the method came from. Personnel must be trained and authorised for what they actually sign for. Records must include complete raw data with changes properly struck through; a common failure is a tidied report with no underlying data. Acceptance basis must be stated, not merely implied by a number. Traceability must work in both directions.

That last point discriminates well between suppliers. Tracing from a batch to its inspection data is something most can do. Tracing from an instrument to the batches it measured is something far fewer can do, and that capability determines whether the impact of an out-of-tolerance instrument can be bounded.

What to examine on site. Do not rely on the sample records the supplier has prepared. Sample at random and vary how you sample. Sample by date, pulling a record from six months ago to see whether quality then matches quality now, since housekeeping done just before an audit shows up once you reach back. Sample by exception, asking when the last nonconforming result occurred and examining how it was handled, because normal records are easy to present well while exception handling reveals actual practice. Sample by equipment, following one instrument from calibration certificate to intermediate checks to recent data. Sample by person, checking one inspector's authorisation against the records they have signed.

Emphasis by supplier type. For raw material suppliers, focus on evidence of batch consistency and check whether their inspection covers the characteristics that matter to you, since they test to their own standard and may not measure your critical property. For component suppliers, focus on capability for key dimensions and performance and on nonconforming product control, and on whether they manage their own suppliers. For contract manufacturers, focus on whether in-process and final inspection follow the list and acceptance basis you provided rather than their own habits. For special process providers such as sterilisation, focus on process validation, monitoring records and the release basis for each batch.

Turning findings into controls. Where inspection capability is inadequate, move that item to your own incoming inspection or an external laboratory. Where records are irregular, require correction and re-examine at the next audit. Where traceability is weak, add record retention and access clauses to the quality agreement. And for critical characteristics, require the supplier to ship actual measured data rather than a certificate of conformity. That last measure has visible effect: the data are useful in themselves, and the requirement itself pushes the supplier towards a more disciplined process.

Tiered management. Not every supplier needs the same depth. Grade by how critical the material is: full on-site audit and periodic re-audit for critical materials; questionnaire, document review and sample verification for ordinary ones; incoming inspection as the main control for low-risk items. Grade by the consequence of failure for product safety and performance rather than by purchase value, since some inexpensive parts carry severe consequences.

Relationship to your own inspection. A supplier having inspected does not remove your need for incoming inspection; theirs is outgoing control and yours is verification. You can adjust incoming inspection intensity according to evaluation results, reducing sampling for well-rated suppliers and maintaining it for others. That linkage is what makes evaluation produce practical value rather than a report that is filed and forgotten.

How we handle it. Clients sometimes ask us to verify a batch independently, for supplier evaluation or to settle a disagreement. Three things matter: sampling must represent the batch, the method must match what the supplier used or the data are not comparable, and the report must state method and conditions so both sides can compare. Method mismatch is one of the main reasons two sets of data disagree; in a good share of the cases we see, neither party measured wrongly, they simply ran different conditions and the results were never going to match.

Send us both sides' methods and data and we will work through it. Phone or WeChat: +86 132 4819 8029.