In the first hour of The Packet Ledger we ask a rude question: when your chart says “user,” what packet actually arrived? Often the honest answer is a reset advertising identifier, a vendor IDFA stand-in, or a cookie that died when someone borrowed a tablet.
That is not a philosophical problem. It is why finance directors stop signing papers. A household that shares one login looks like one loyal person. A person with three devices looks like a crowd. A privacy prompt that resets the identifier looks like churn.
We are not arguing for a perfect identity graph. Perfect graphs are a sales category. We are arguing for a written subject: device, logged-in account, household, or subscription. Pick one for the board paper. Document the others in an appendix if you must. Do not average them into a single gold line and call it product-market fit.
United Kingdom teams have a second reason to be precise. If the packet carries enough to recognise a person, you have a data-protection problem as well as a measurement one. The Signal Packets questions exist partly so counsel can sit in the same room as the analyst without a translator.
If this note sounds like an advertisement for a nine-week sitting, it is only because we have watched too many “quick identity workshops” produce a slide and no key. The work is slow because the product is allowed to be complicated. The packet is not allowed to pretend otherwise.