When DAU misleads product leads
Daily active users count who showed up. Engagement asks whether they came back to do the job the app exists to do. Those two stories diverged, in public, for a Bangkok grocery team last year.
The team’s Songkran packet led with a record DAU. The screenshot was real. The interpretation was not. When a Studio Seat student cut the same days by acquisition channel, the bump sat almost entirely in users who arrived that week with a free-delivery code. Returning grocery shoppers — the people who already had the app — barely moved.
DAU did its job: it counted a crowd. It could not tell a steering group whether the product had become more engaging. That is not a failure of the metric. It is a failure of using a crowd count as a quality score.
A working substitute
We do not ban DAU. We demote it. Next to it we want: share of actives who completed a core task, week-4 retention for the non-promoted cohort, and a note on whether the window included a festival or a code blast. If those three disagree with DAU, DAU loses the headline.
The grocery student wrote a one-sentence limitation: the spike measures promotion reach, not engagement quality. Her CPO kept the sentence in the board appendix. That is the win. We did not invent a new north-star religion. We stopped a promotional weather event from becoming next year’s target.
What this means for benchmarking
Mobile Engagement Benchmarking that still leads with DAU will happily compare a code-fuelled week to a quiet competitor. Peer sets only work if both sides are measuring intentional return, not inbound traffic. If your vendor dashboard cannot split campaign-driven opens from self-directed task opens, the homework is to export and split it yourself. The tool’s default view is not a defence.
If this sounds like the room you need, the flagship Engagement Benchmark Studio spends Module 01 on signal inventory for exactly this reason.