结论:指标不是写给自己看的,是写给执行的人看的
性能指标写进技术要求之后,会被两拨人使用:检测机构按它设计试验,审评人员按它判断产品。这两拨人都不了解你的产品,他们能依据的只有字面。
所以判断一条指标写得好不好,标准很朴素:把这句话交给一个没见过你产品的工程师,他能不能据此设计出一个试验并给出明确结论。 答不上来的,就还需要改。
三条标准的具体含义
| 标准 | 要回答的问题 | 缺失时的表现 |
|---|---|---|
| 可测 | 测什么、用什么测、怎么测 | 实验室来电询问,进度停滞 |
| 可判 | 达到什么算合格 | 数据出来了但结论下不了 |
| 可复现 | 换个人、换个机构结果一致吗 | 两份报告结论不同,都没错 |
可测:测量对象要具体到位置
「外壳强度满足使用要求」这类表述无法执行。可测的写法要说清三件事:测量对象是哪个部位、用什么量、用什么方法或依据哪个标准的方法。
比较常见的疏漏是测量对象不明确。产品上有多处类似结构,指标只写了「连接处」,实验室不知道测哪一处。这时候要么逐处都测(成本上去了),要么来回确认(周期上去了)。
可判:限值要有方向和边界
可判要求给出明确的合格边界。写「不小于某值」要说明是单个样品都不小于,还是平均值不小于;写一个范围要说明包不包含端点。
还有一类容易出问题的是判定对象的数量。测三个样品,其中一个略低于限值,算不算合格?这取决于判定规则,而多数技术要求根本没写判定规则,只给了一个数字。判定规则缺失时,检测机构通常按较严的口径执行,也就是全部样品都要满足。
可复现:条件的颗粒度是关键
可复现要求把影响结果的条件都写出来。哪些条件会影响结果,取决于测什么,常见的包括:
环境的温度与湿度;样品的预处理方式与时长;加载的速率、方向与作用点;测量的时机,是加载中读数还是卸载后读数;重复次数与间隔。
条件写到什么颗粒度算够,可以用这个办法判断:假设两家机构同时测这一条,有没有可能因为理解不同而得出不同结果。如果有,就说明条件还不够细。
指标值怎么定
指标值有两个约束:下有标准的强制要求,上有量产的实际能力。
在这两者之间怎么取,建议用实测数据而不是设计目标值。做法是取一批量产状态的产品(不是精心调试的样机)实测,看数据的分布,在分布的不利一侧之外留出裕度,作为指标值。
裕度留多少,取决于工艺稳定性和批次间差异。稳定性好的,裕度可以小一些;波动大的,裕度要留够,否则出厂检验会频繁出现临界情况。
用样机数据定指标是后续麻烦的主要来源之一。 样机是工程师反复调出来的,量产件不可能每台都达到那个水平。
条件描述的常见缺口
把几类容易漏掉的条件列出来,撰写时逐项对照:
预处理。 样品在试验前需不需要在特定条件下放置、放多久。这一项对高分子材料类产品影响很大。
安装或固定方式。 样品怎么固定在试验装置上,固定方式不同受力就不同。
配置状态。 可调部件调到什么位置、可选配件装不装。
试验顺序。 多个项目在同一样品上执行时的先后顺序,前面的项目可能影响后面的结果。
判定时机。 试验结束后立即判定,还是恢复一段时间后判定。
定稿前的核对清单
逐条过一遍,问六个问题:测什么部位;用什么方法;合格边界是多少;判定规则是什么;哪些条件会影响结果、写全了没有;量产件能不能稳定满足。
六个问题都能回答,这一条就可以定稿。这个过程在送检前做,成本是几个小时;在检测阶段发现问题,成本是排期加样品。
与其他文件的一致性
技术要求里的参数会在说明书、标签、宣传材料里重复出现。这几处的数值必须一致。
实际操作中不一致的情况相当常见,原因是这几份文件由不同的人在不同时间编写。建议指定一个参数表作为统一来源,其余文件都从这里取值,修订时统一更新。
指标之间的相互影响
单条指标看着合理,放在一起可能互相冲突。常见的是性能与耐久、性能与安全之间的矛盾:提高输出能改善性能指标,但加剧发热,影响温升和寿命;提高强度要求会增加重量,而重量本身可能也是一项指标。
定稿前建议把所有指标列在一张表里通读一遍,找出互相牵制的组合,确认现有设计能同时满足。逐条定指标而不做整体检查,是后续出现「这项过了那项不过」的常见原因。
留给未来的余地
技术要求一旦定稿并随资料提交,修改就要走变更程序。所以撰写时可以适当考虑产品的迭代空间。
具体做法是:指标值按当前能力定,但表述方式上避免把自己框死。比如结构描述写到功能层面而不是具体零件型号,这样后续更换等效零件时不必修改技术要求。当然,这要在满足明确性要求的前提下做,不能为了留余地把话说含糊。
我们的做法
我们在报价阶段就会通读技术要求,把执行不了的条款标出来与委托方确认。标注的类型有三种:缺方法的、缺判定的、条件不足的。
这个动作不收费,但能省掉的返工往往不小。 见过的情况是,一份技术要求里有五六条存在执行问题,如果直接开测,每一条都要停下来确认一次,整体周期会被拉长两三周。
如果你的技术要求正在写或者已有草稿,可以发过来先看一遍可执行性,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,其他问题见常见问题。
English version
Conclusion. Performance indicators are not written for your own benefit. They are used by two groups who do not know your product: the laboratory, which designs tests from them, and the reviewer, who judges the product against them. Both have only the wording to work from. A useful test of any indicator is whether an engineer who has never seen your product could design a test from it and reach a definite conclusion.
Measurable. State what is measured, where on the product, with what quantity, and by what method. A common gap is an ambiguous measurement location: where a product has several similar features and the indicator refers only to a connection point, the laboratory cannot tell which one, so either everything is tested or the schedule stalls in clarification.
Judgeable. Give a clear acceptance boundary. Where a minimum is stated, say whether every sample must meet it or only the mean. Where a range is given, say whether the endpoints are included. Sample-level decision rules are frequently absent, leaving only a number; in that case laboratories generally apply the stricter reading, requiring every sample to comply.
Reproducible. State every condition that affects the result: temperature and humidity, sample preconditioning and its duration, loading rate, direction and point of application, whether readings are taken under load or after release, and the number and spacing of repetitions. A practical way to judge whether conditions are detailed enough is to ask whether two laboratories could reasonably interpret them differently and obtain different results.
Setting values. The floor is whatever the applicable standard mandates. The ceiling is what production can achieve consistently. Between the two, derive the value from measured data on production units rather than tuned prototypes: look at the distribution and leave margin beyond its unfavourable edge. How much margin depends on process stability. Deriving indicators from prototype data is one of the main causes of later batch-release failures.
Commonly missing conditions. Preconditioning, mounting or fixing method, configuration of adjustable parts, the order in which items are run on the same sample, and when judgement is made relative to the end of the test.
A checklist before finalising. For each indicator ask: what part is measured, by what method, what is the acceptance boundary, what is the decision rule, which conditions affect the result and are they all stated, and can production units meet it consistently. Answer all six and the line is ready.
Consistency with other documents. Parameters recur in the instructions for use, the label and marketing material, and they must agree. Designate one parameter table as the governing source and draw all other documents from it.
How we handle it. We read the technical requirements at the quotation stage and flag clauses that cannot be executed, grouped as missing method, missing acceptance criteria, or insufficient conditions. This costs nothing and commonly saves weeks: a document with five or six unexecutable clauses means stopping to clarify each one after testing has started.
Send us a draft and we will review executability. Phone or WeChat: +86 132 4819 8029.