Why the MMP and network disagree: reconcile one install dataset step by step

A network dashboard and an MMP export can both be internally consistent and still disagree. This lab treats the gap as three labelled buckets rather than a single unexplained error.

Define the calculation before using it

Enter network-claimed installs, MMP-claimed installs and the count both systems share after an authorised identifier match. The unmatched network remainder, unmatched MMP remainder and overlap must add back to each source. Do not force a match where the identifiers differ.

Work through the synthetic example

A synthetic day with 1,200 network installs, 1,050 MMP installs and 980 in both leaves 220 network-only and 70 MMP-only rows. The useful conversation is which of those buckets is click-definition, lookback, timezone or unattributed organic—not which total is ‘the real number’.

Use the artifact and preserve its assumptions

Open the editable calculator to change the inputs and inspect the sensitivity view. The CSV records synthetic inputs and expected outputs; the JSON fixture keeps the equations available for reproduction. These calculations have been checked against the stated example. No measured campaign data is included.

The sensitivity rows vary only both by 20% below and above the entered value. They are scenarios, not confidence limits or a forecast distribution. A row outside the model’s constraints is labelled rather than turned into a plausible-looking result.

Evidence and limits

AppsFlyer documents that attribution depends on matching rules and lookback windows. This worksheet uses labelled synthetic counts; it does not reproduce a live AppsFlyer export. See AppsFlyer attribution model, especially “Attribution model / matching”.

Synthetic teaching data. A real reconciliation needs the same day boundary, app, OS and conversion definition. This does not diagnose fraud.

Background: Reading an MMP dashboard without fooling yourself and UA metrics explained: CPI, ROAS, LTV and payback. These existing articles provide context; the present calculation does not verify every archived claim.

Draft prepared 19 September 2026. Synthetic teaching example; human editorial and specialist review pending. No actual campaign outcome or recommended industry default is claimed.