When a network reports more installs than your mobile measurement partner, first align the comparison. Use the same app, dates, time zone, event definition and reporting cutoff, then inspect attribution windows and models. A difference is a question to reconcile, not proof that either report is dishonest. This April 2023 article explains several common reasons the totals can diverge.
Table of Contents
Cost per install and cost per in-app action can differ when their denominators or credited conversions differ. Start with the definitions below, then compare the settings behind each report. The mobile user acquisition guide provides the wider planning context.
TLDR: Attribution? Pardon?
To understand what attribution is, we can make a simple comparison to a soccer team (or basketball, volleyball, rugby…). Imagine that the ball goes into the goal and two players had touched it before that happened. Which of the two scored the goal? We would have to watch the replay to find out, right? 
Well, this process of knowing who scored the goal could be the very definition of attribution. Now replace the goal with a conversion in an app (install, registration, subscription) and a player with an Ad Network (Meta, TikTok, Google, or even organic) and you will have the result of what Attribution is in for every app. Nothing more and nothing less than the process of recognizing an action that occurred in/around an app and attributing it to a specific source. Every time a person clicks or views your ad, the attribution process begins. With mobile attribution, you will be able to measure the effectiveness of your campaigns beyond clicks and installs and understand in which ad networks to put your money. We won’t bore you with more theory, as if you want to learn more about mobile attribution, we have an entire article explaining how it works. But with many examples and in a simple way, we promise.
Five reasons why your MMP might not be telling you “the whole truth”
Different attribution windows
An attribution window is the time allowed between an eligible interaction and the conversion being credited. Different configured windows can produce different totals. For illustration, a network with a 14-day click window could credit a first launch on day eight while an MMP with a seven-day window would not credit that click. These are hypothetical settings, not current Twitter defaults. First launch and download are also different events. 
Source: AppsFlyer Align the windows that each system supports, then check view-through credit, event definitions, time zones and reporting delay. Matching a window alone does not guarantee identical totals.
The weight you assign to each interaction with your ads
Continuing with the previous soccer/sports comparison – who is more important when counting a team’s victory, the one who scored the goals or the goalkeeper who saved two penalties against? It’s difficult to say, right? Well, here we have to play with some kind of rule that helps us decide how to distribute those credits. Without going into the technical part of attribution models, we can say that the most widespread model is the last-touch model. In this case, if a person sees 3 different ads from let’s say Google Ads then TikTok then Meta Ads, and clicks on Meta’s ad the install will be attributed to the last one, which is Meta. And what about Google and TikTok? They do count, but in a different way. But since we don’t want to make it too complex, if you’re interested in the topic, you can read more in this article about Assisted Installs. And where do discrepancies arise here? Basically, they occur if your traffic source and your MMP have a different attribution model (in addition to adding the attribution windows we’ve already seen). Because if the coach gave more importance to the goalkeeper for the victory and the club president to the striker who scored the goals, we would have a problem, wouldn’t we? Is this fixable? Well, to some extent, yes.
The concept of “install” doesn’t always mean the same thing
This is one of the less technical topics but often overlooked. The fact is, simply put, your MMP and Ad Networks (oh, and the Stores themselves) consider Installs to be different things. Like the US and UK pronunciation of tomato (/təˈmɑː.təʊ/ versus /təˈmeɪ.t̬oʊ/), but in reverse. When you understand that your MMP measures downloads through an SDK that you/your developers add to your app’s code, you might ask the question: “How do they know if a user installs the app and doesn’t just open it?” Well, there’s your answer:
- Store reports can count a download or installation before the app is opened. Check the specific store metric and the network’s own conversion definition.
- AppsFlyer records a new install after first launch. Confirm the integration and the rules for reinstalls, attribution and reporting dates in the MMP being used.
Booooom!. Did your mind just explode? Now you understand why MMPs always talk about “First Launch,” among other things, right?
.
