Digestmeasurement
Reading an MMP dashboard without fooling yourself
By UA Ledger staff — Archive date: 3 min read
Self-attributing totals, reattribution, modelled data, timezones, currency and cost import gaps: six ways an MMP dashboard quietly misleads.
An MMP dashboard looks like a single source of truth because it aggregates everything into one number per channel per day. It's closer to six separate accounting conventions stitched together, and reading it as one clean truth is the fastest way to make a bad decision from good-looking data.
Six ways a dashboard misleads
Self-attributing network totals are the most familiar trap. Meta and Google and TikTok each report their own claimed installs inside their own interface, and those numbers don't deduplicate against each other or against the MMP's own total, so summing channel-reported installs will always overstate total installs, sometimes by a wide margin.
Reattribution quietly inflates a channel's apparent contribution when a lapsed user reinstalls and gets credited to whichever channel last touched them, even if that touch was a low-intent view rather than the reason the user came back. Installs spike; spend doesn't. Check that pattern for reattribution before reading it as an organic efficiency gain.
Modelled and deterministic data share the same column on most dashboards despite representing very different confidence levels. SKAdNetwork and AdAttributionKit postbacks on iOS arrive aggregated and delayed by design, and an MMP's modelling of the gap between those postbacks and a full picture is an estimate, not a count, however precisely the dashboard displays it.
Timezone mismatches between a platform reporting in Pacific time and an MMP reporting in a studio's local time or UTC shift installs across a midnight boundary. What that looks like on the chart is an unexplained day-to-day swing that has nothing to do with performance.
Currency conversion, for any studio buying in multiple markets, applies whatever exchange rate the MMP or platform used at ingestion time, which may not match the rate finance uses for reporting, so the same cost per install looks different in two systems that are both technically correct.
Cost import gaps happen when a smaller network's spend data fails to sync for a day or two. Installs land with no matching cost, and that channel's blended CPI looks artificially strong until someone notices the gap and backfills it.
None of these six is a bug in the sense of something an MMP ought to fix outright. They're the standard behaviour of an industry built on multiple platforms reporting through different conventions into one interface, a point this site's earlier piece, Reading your MMP dashboard after AppsFlyer's round, covered from a different angle. The discipline that stops them causing bad decisions is simple to state and tedious to keep: check a suspicious number against at least one of these six before treating it as a real signal.
Related archive reading
These articles provide related context and remain subject to their stated review status.