“Just focusing on the install isn’t productive,” Oiknine said. “If someone drops an app, why did they drop it? Did they uninstall after time had passed and they’d generated some revenue or did they immediately uninstall after downloading? These are very different types of behaviors and it’s important to know the reason.”
Say the uninstall rate is high in a country where the average uninstall rate is fairly low, like in the United States or Europe. That’s a clear sign that there’s either market penetration for that app or that category of app or that there’s an issue with how the product performs. If the uninstall rate is high across all campaign sources, that would indicate a problem with how the app functions. But if only a handful of ad networks are producing high uninstall rates, that could mean those parties are doing something shady.
Some networks will do whatever it takes to drive downloads – because that’s the metric they’re generally paid on, regardless of quality.
Ola Cabs for one, whose acquisition strategy includes always-on paid media and referral campaigns, along with the periodic big branding push, is using uninstall attribution to figure out which affiliate networks use incentivized campaigns to bring in users, the majority of which only download an app to redeem some sort of offer or move to the next level in a game. If Ola suspects that incentivized traffic is being used, it hits pause on that partner relationship.
Fraud is also an issue that Ola is working to combat with a combination of attribution analytics and vigilance.
“Uninstalls are sometimes connected to fraudulent installs, so we’re able to use the attribution tool for fraud detection,” Gangrade said. “And we have a fraud analytics team that’s continually tracking fraud at the user level, the device level and the network level.”
But even if the reason behind an uninstall isn’t criminal, that doesn’t mean it’s not a headache for the publisher, as is the case when a user deletes an app after only using it once or removes it because there’s something confusing about the UX.
“We look to see exactly what consumers who uninstall tend to do right before they uninstalled and if we see it has to do with the app experience, we tend to rectify that very quickly,” said Gangrade, who noted that Ola rejiggered its onboarding flow after observing that users were dropping off before fully completing the sign-up process.
Oiknine also sees the uninstall attribution tool as a way to create – or suppress – targeted audiences. If certain users are likely to uninstall, it might be best to proactively exclude them or their lookalikes from targeting segments, for example. Developers are then able to append that type of information to a user’s profile in Apsalar’s DMP.
“The signal of an uninstall can be just as interesting, if not more so, than the signal of an install,” Oiknine said. “But data is only really interesting if it helps you take action that makes sense.”