The number I didn't publish

There was a version of the Então? Comer! case study with a satisfaction score in it. I wrote it, looked at it for a day, and took it out.

The product has no analytics. Nobody has been observed using it under conditions that could produce a number. The north star I actually work to — did a person get from a vague want to one place they were willing to open on a map — is a direction, not a metric, and I knew that when I typed the number.

The temptation is obvious. Every portfolio has percentages in it. A case study that says “no usability study yet” next to one that says “SUS 78” reads as the weaker of the two, right up until someone asks how the study was run.

What made the decision easy was noticing what the invented number was competing with. On the same project I ran the quiz combinations against the database and found 44 of 288 paths returning nothing. I wrote a script that diffed 160 rewritten catalogue fields against their sources and it caught three details I had invented while making sentences sound better. Contrast is measured per pair; the worst is 4.58:1.

Those are real, checkable, and mine. Putting a soft number beside them doesn’t add weight — it invites the reader to wonder which of the hard ones I also rounded up.

So: name the method, then let the method set the size of the claim. Counted things get numbers. Uncounted things get described as uncounted. Nothing gets promoted.

The cost is a case study that looks slightly less impressive in a stack of twenty. The benefit is that every sentence in it survives being asked about, which is the only situation where any of this matters.