Social Computing · Product
Tremer
Six redesigns, logs from 2,013 devices, and the day I retracted my own findings
- Role
- Founder · product design · log analysis
- Period
- Nov 2025 – Present
Tremer was redesigned six times. What moved it was not my intuition but merchant interviews — the same gap kept surfacing between what users enjoyed and what businesses would pay for, and it eventually pushed the product to B2B.
The retraction
My early analysis reported that people lingered on the screen: a median of 38 seconds. It was a flattering number, which is exactly why it deserved suspicion. Digging back into the logs surfaced the cause — background timeouts. Time spent with the app backgrounded was being counted as dwell time.
Time spent with the app in the background was being counted as dwell.
Reported 38.0s → re-analysed 1.5s. This single fault contaminated six reported metrics. All of them were retracted and the analysis was re-run.
Cross the threshold yourself
The ten seconds came out of the logs, but to a reader it is just a number. So it runs here instead — this is how long you have stayed with this block.
This is counted in your browser only — nothing is sent or stored. This project retracted six metrics over a measurement artifact, so the measuring done here holds to the same standard.
What survived re-analysis
A 10-second engagement threshold
Across logs from 2,013 devices, sessions that crossed ten seconds behaved distinctly differently from those that dropped before it.
Browsing and sign-up ran in reverse
It was not that heavy browsing led to sign-up; signing up led to heavier browsing. I had been reading the funnel backwards.
Shipped screens
The fun of a game collects the data; the authority of a finance terminal hands it back.
The better a number looks, the earlier it deserves suspicion. Killing your own finding is part of the work.