Blog

Why MMP and Ad Platform Numbers Don't Match

Lakshith Dinesh

Lakshith Dinesh

Head of Growth, Linkrunner

Last updated on 7 min read

Why MMP and Ad Platform Numbers Don't Match

Short answer: Your MMP and your ad platforms count different things. Meta and Google each claim every install their ads touched, inside their own windows, timezones and reporting dates, while your MMP gives each install to one source. A steady gap is normal. A gap that suddenly widens, or hits revenue but not installs, usually means something broke: a window, an event map, a postback or a timezone.

Your MMP says about 2,800 installs from yesterday's Meta campaign. Meta Ads Manager says about 3,100. Your analytics tool shows about 2,500 first opens. Your CFO wants one number for the board deck.

They will never all match. The real question is whether the gap is ordinary measurement difference or a tracking fault sending budget to the wrong campaigns. This post lists the 12 causes we see most, groups them into five diagnostic categories, and ends with a 20-minute checklist and a decision matrix.

Why do MMP and ad platform numbers never match exactly?

Because each system answers a different question. Meta and Google are self-attributing networks: they report every install that followed any of their own ads. Your MMP referees all networks and credits each install to one source.

Google's own guide to app conversion discrepancies lists the structural differences:

  • Third-party tools de-duplicate conversions across networks; Google Ads does not.
  • Google Ads dates conversions to the day of the ad interaction, not the day of the conversion. In February 2026 Google told advertisers that App campaign conversions move to the install date, so check which basis your account shows.
  • Google Play counts installs; third parties count the first app open.
  • Some iOS search conversions are modelled by Google Ads and never reach an MMP.
  • Google Ads uses your ad account's time zone, and conversions can take up to 24 hours to appear.

None of these is a bug. For background, see why self-attributing networks report different numbers.

What size of difference is normal?

There is no universal "normal" percentage. Normal is your own gap, measured on a fixed basis, holding steady.

  • Measure the gap per network for four weeks on a shared window and timezone. That is your baseline.
  • A gap near its baseline is measurement difference: track it, do not chase it.
  • A gap that jumps, or moves on one campaign or one platform only, is a signal.
  • Definitions explain most gaps. In Linkrunner support conversations, ROAS mismatches with Meta or Google usually trace to a definition gap first (7-day versus 30-day window, cohort versus in-period counting, view-through), and only occasionally to a real setup fault. In one Google comparison, moving both sides to the same 30-day window shrank the gap to a few per cent.

Category 1: Is it an attribution window or counting difference?

Window and counting rules cause the largest structural gaps. An attribution window sets how long after a click or view an install can still be credited.

Cause 1: click windows differ.

  • Meta's default attribution setting is 7-day click and 1-day view.
  • Google Ads tracks installs within 30 days of the ad interaction and in-app conversions within 90 days by default (Google).
  • In Linkrunner the default click window is 7 days for most networks and 30 days for Meta and Google, and each channel can be set from 1 to 30 days (docs).

Cause 2: view-through counting. Networks count installs from people who saw but did not click an ad. If your MMP reports view-through separately, or not at all, the network will show more.

Cause 3: self-attribution and double counting. If a user clicked a Meta ad and then a Google ad, both networks claim the install. Your MMP gives it to one. Add the network totals together and they will always exceed the MMP.

Cause 4: "every" versus "one" counting. Google Ads can count every conversion per user; most MMPs count one install per user. Google also counts Android re-installs as installs, which not every MMP does.

Test: compare by click date, on matching windows. Google suggests taking a week of clicks and then 30 days of MMP conversions for that cohort. For more on choosing lengths, read our attribution windows guide.

Category 2: Are events mapped and sent correctly?

Event problems hit revenue and ROAS harder than installs, so a clean install count with a broken ROAS points here.

Cause 5: event name mismatches. The postback sends an event name the network does not expect, so the network shows zero revenue while your MMP shows plenty.

Cause 6: missing or wrong parameters. Revenue events need a value and an ISO currency code. Sending amount instead of value, or a currency label instead of a code, means the network drops or misreads the event.

Cause 7: event timing. A purchase fired when the user taps "buy" counts payments that later fail. Events fired before the SDK initialises are lost. Fire purchases only after payment confirmation.

Test: send a test purchase with known values from a production build. Check it lands in your MMP with the right name and parameters, then in the network's events manager. If it is in the MMP but not the network, the postback is the problem. Webhooks, postbacks and server-to-server feeds fail in different ways, so know which one carries each event. For Meta specifically, use our Meta SDK event troubleshooting guide.

Fix: keep one event spec (names, parameters, types) shared by engineering and marketing, and test every new event before release. Linkrunner's anomaly alerts flag sudden install spikes, ROAS drops and SDK issues with a suspected cause, and affiliate partners can see each postback's delivery status (delivered, retrying, failed) in Postback Logs.

Category 3: Are timezones and currencies aligned?

Cause 8: timezones. Google Ads reports in your ad account's time zone; your MMP reports in its project time zone. An install at 2:30 am IST on 16 January is still 15 January in a US Pacific account, so daily totals shift between days even when weekly totals agree.

Cause 9: currency. If your app reports ₹ and the ad account shows USD, comparing raw values makes ROAS meaningless. Even with conversion on, tools use different exchange rates, adding a small, steady gap.

Test: compare weekly rather than daily totals. If the weekly gap disappears, it is timezone. For currency, send a known purchase (say ₹999) and check how each tool shows it.

Fix: set every tool to one timezone (IST or UTC for most Indian teams) and send revenue to networks in the currency each one expects.

Category 4: Are paid installs leaking into organic, or the other way round?

Cause 10: fraud filtering and click spam. Click spam fires fake clicks just before organic installs to steal credit. Your MMP flags these, so the network counts installs your MMP does not credit to it. Watch for very short click-to-install times, install spikes that do not follow spend, and installs with almost no later activity. Linkrunner's Fraud Control flags these in real time and can withhold postbacks for flagged installs.

Cause 11: broken deep links and web-to-app gaps. If WhatsApp, SMS, email or website links lose their parameters, those installs show as organic and paid channels look worse than they are.

Test: if organic installs spike on days you raised paid spend, attribution is leaking. Click your own links from every channel and check each lands under the right source.

Category 5: Is it just a reporting delay?

Cause 12: delays, especially on iOS. Without App Tracking Transparency consent, iOS installs are measured through Apple's SKAN postbacks, which arrive late on purpose. Under SKAN 4 the first postback covers days 0 to 2 and arrives after a random 24 to 48 hour delay; the second and third cover days 3 to 7 and 8 to 35, each with a 24 to 144 hour delay (Adjust help centre). Apple's newer AdAttributionKit works alongside SKAN. Linkrunner reports SKAN postbacks today and is ready for AdAttributionKit.

Networks also model iOS conversions for users who decline tracking.

Test: compare a cohort from a week ago, not yesterday. If the gap shrinks sharply, it was delay.

Fix: do not move budget on same-day comparisons. Wait at least 3 days for Android and other deterministic data, and longer for SKAN.

The 20-minute diagnostic checklist

  1. Define the gap (2 min). Write it down: "MMP about 2,800 Meta installs on 14 Jan; Meta about 3,100; gap about 10%." Note the metric and dates.
  2. Check windows (3 min). Compare click and view windows in your MMP and the network. Re-pull on a matched window and by click date.
  3. Check timezones (2 min). Compare the MMP project timezone with the ad account timezone. Re-check on weekly totals.
  4. Check for delay (3 min). Move the date range back a week. For iOS, make sure you are not comparing SKAN data with modelled same-day numbers.
  5. Check event mapping (4 min). If revenue or post-install events are off, confirm event names and the value and currency parameters in the postback setup.
  6. Check organic leakage and fraud (3 min). Compare organic installs with new users in analytics, and look at installs with very short click-to-install times.
  7. Log and fix (3 min). Record the date, gap, cause and fix. Expect corrected deterministic data in 1 to 2 days and SKAN data later.

Still stuck? Send your MMP's support team what you have ruled out. For a symptom-by-symptom troubleshooter, see Attribution Debugger's guide to mismatched counts.

When should you investigate and when should you accept it?

Set your own thresholds from your baseline. A sensible starting point is to investigate when the gap moves more than about 5 points from its four-week norm.

SignalPriorityLikely causes
Gap jumps overnightInvestigate nowBroken postback, SDK update, changed window or campaign setup
Revenue gap, installs fineInvestigate nowEvent names, parameters, currency
One campaign out of lineWithin a weekThat campaign's links, windows or tracking parameters
iOS gap, Android fineWithin a weekSKAN setup, modelled conversions, delay
Daily gap, weekly totals matchAcceptTimezones, reporting date
Steady gap at baselineAccept and monitorSelf-attribution, windows, counting rules

If fixing a gap would not change which campaigns you scale or pause, accept it.

FAQ

Why don't my Meta installs match my MMP?

Meta counts every install that followed a Meta click within 7 days or a view within 1 day under its default setting, including installs another network also claims. Your MMP credits each install to one source on its own window, and may flag fraudulent installs. Compare on matched windows and click dates before assuming a fault.

Why does Google Ads show more installs than my MMP?

Google Ads counts within a 30-day window, can count re-installs and every conversion, and models some iOS conversions. Your MMP de-duplicates across networks.

Which number should I report to leadership?

Your MMP's, because it de-duplicates across networks. Use each network's own numbers for in-platform optimisation.

How long should I wait before comparing iOS numbers?

At least several days. SKAN 4 postbacks arrive after random delays of up to 48 hours for the first window and up to 144 hours for later ones.

Takeaway

  • A steady gap between your MMP and ad networks is normal; a moving gap is a signal.
  • Check windows, timezones and delay first; they explain most gaps.
  • Revenue gaps with healthy installs point to event mapping.

Linkrunner is a mobile and web attribution platform that credits each install to one source across every network and flags anomalies with a suspected cause. Request a demo from Linkrunner to see how it fits your stack.

Start measuring the installs your team cares about

Bring attribution, deep links, SKAN, cohorts, and campaign intelligence into one workflow your growth team can trust.