The site knows things the file doesn't

The site knows exactly what happened today: who was on, what arrived late, what was opened up and what it looked like inside. But it knows it in phones, WhatsApp threads and heads, and none of it is reaching the file. Knowledge held that way has a half-life. People rotate to other jobs, phones get replaced, threads scroll into the dark, and in a few months today will be a set of recollections that do not quite agree.

The gap surfaces whenever money depends on proof: the valuation that needs progress evidenced, the variation that needs the instruction found, one day perhaps a claim that needs the whole story told from paper. Capture that actually happens is designed around the team that has to do it, quick enough to keep up daily, structured enough that what is captured can be found again, and checked monthly so it does not quietly stop.