A rebrand is a marketing event that pretends to be cosmetic. In App Analytics it is a fracture. New icon, new bundle identifier, sometimes a new company legal entity, and suddenly the people who used you last month arrive as strangers. Thirty-day retention — already a blunt instrument — becomes a costume.
The usual stitch is an identity graph: device ID to logged-in user to email hash. Graphs are only as honest as their join keys. After a rebrand we have watched teams match on advertising identifiers that iOS had already begun to withhold, then congratulate themselves on “recovering” a series that was mostly unmatched noise.
In Retention Architecture we draw the join on paper before anyone writes SQL. Who is the same person? What evidence would falsify that? If the only evidence is a probabilistic library with a threshold nobody can explain to a CFO, the series should be published as two charts with a dated seam, not one smooth line.
Thirty days is also a problem of ritual. A weekly habit app will look dead on a thirty-day first-open cohort because the product never asked for daily presence. After a rebrand, that mismatch is blamed on the new name. It was already in the old name. The rebrand merely gave everyone a meeting in which to panic.
If you must show a number in the quarter of a relaunch, show installation quality and day-two activation on the new identifier, and keep the historical retention chart in an appendix with a caption that admits the break. Captions are cheaper than a year of arguing with a ghost series.