Readings, not impressions

A reading is a claim you are willing to compare. Next week, next year, another person. “Looks fine” does not compare. It compares only to the speaker’s mood, and the speaker may be on a different shift when the question comes back.

I want numbers where the world gives you numbers. Level, pressure, temperature, a count of drips if you are going to stand there and count, a measurement a hood instrument actually showed. Type the number the instrument showed. Do not “correct” it because you know the instrument runs high. If it runs high, the limit or the instrument is wrong, and fixing the limit is an administrator’s job. Fixing the number is how the history becomes fan fiction.

I also want phrases, used narrowly, where a number would be a costume. Seepage at a toe drain. A smell at a wet well. The position of a valve that is either seated or not, described as seated or not, rather than “about closed.” The phrase has to be one you could hand to a person who was not there and have them look for the same thing. “Damp patch, about the size of a hand, no flow” is a phrase with a boundary. “A bit damp” is a shrug.

The product will let you type a shrug. I am not interested in a form that refuses prose. I am interested in a crew that notices, on the asset, that last week’s phrase and this week’s phrase are not the same kind of sentence. The last three readings sit on the asset for that reason. If one of them says “hand-sized damp patch” and today’s says “flowing,” you do not need a chart. You need to read three lines. Charts are where shrugs go to look quantitative.

Limits are how a phrase becomes a fail without a debate at the truck. “No visible drip” fails a visible drip. You do not need a committee. Where you cannot write the limit, say so, and make the check a note instead of a pretend measurement. Pretend measurements are worse than notes because they sort into pass and fail as if the universe agreed.

There is a social problem hiding here. Experienced people speak in impressions because impressions are fast and, in their mouth, accurate. The record is not their mouth. The record is the person who will read it after a crew change, which Ruth has written about, and the person in an office who cannot smell the wet well. Forcing a reading feels like distrust. It is a kind of distrust: distrust that a memory will be present later. That distrust is the correct engineering posture. It is not an insult.

At Riverward, the gland on RW-VLV-14 is “drip, about one per second.” They could have bought a fancy description. They did not need one. The phrase is stable enough that a change to a stream would be obvious in the three-line history. If it stays a drip for a year, the open item is a decision about packing, not a mystery about what anyone saw. Mysteries are expensive. Stable phrases are cheap.

Please do not add a field called “severity” beside the reading. Severity is an impression with a dropdown. It will drift. Two inspectors will not mean the same thing by “major,” and the pack will sort on it as if they did. If something blocks startup, mark the check as a block. If it can wait, hold it with a date. Those are decisions. Severity is a mood with a legend.

I will accept one impression field, and it is the note, clearly not a result. “Customer on site, do not prop the gate.” That is not a reading of the gate. It is a fact about the hour. Keep it off the result column so nobody graphs it.

If you are editing checks this week, read ten historical “ok” lines out loud. If you cannot tell what would have made them not ok, the check is not a check yet. Rewrite the passing sentence until a new person could fail it. That rewrite is the work. The screen will not do it for you, and I do not want it to try.