iOS 10 Limit Ad Tracking

Apple recently released iOS 10. It's got a bigger font and you can swipe in some new places. Also something about iMessage. But most importantly for us at TripleLift, it significantly modified a feature called Limit Ad Tracking. This is buried deep in system settings, and thus will not fundamentally upend the market, but this represents a noteworthy continuation of Apple moving away from identity-based advertising. Currently, roughly 15% of users have this selected, with the number apparently declining as of late.


Previous versions of iOS included a "Limit Ad Tracking" setting that would, when enabled, send the fact that the user had selected this, along with the user's ID for Advertising (IDFA). Apple's policies required some degree of respect for this selection, but it still technically allowed for anything a bad actor wanted to do. As of iOS 10, the same setting, when enabled, will zero out the user's IDFA. This means conversion attribution, frequency capping, unique user counts, etc - all things that were previously permitted in iOS 9 under Limit Ad Tracking, are now technically impossible (-ish).

Each app has 2 user IDs. One is for advertising (IDFA), and is persistent for that user across all apps. The other is the Vendor ID, which allows apps to access an app-specific ID that they can use - even when the IDFA is zero'd out. So if an app has the user's email, and advertising company has the user's email, they could key off the email and then user the vendor ID. This would allow an app to extend its user ID matching everywhere that they have an email or similar. Ironically, this creates a worse privacy regime than before, though it is likely (or will be) against Apple's policies.

TripleLift does not leverage user data the way other vendors do, however we do benefit from those other vendors submitting higher bids for users that they've developed rich profiles for. So this move would impact us at roughly the same way it impacts mobile app monetization as a whole. Companies that are more immediately impacted are those that primarily or entirely targeting based on IDFA, which especially includes Facebook (FAN) and Google. Apple has a long-stated bias away from user targeting, and this may simply be a continuation of that policy and/or a push towards premium, subscription models from which Apple takes a direct cut.

Factual's Business Model

Factual is a mobile data provider. Specifically, they've developed an understanding of millions (?) of locations. This means, for example, that it knows where Yankee Stadium is, and that it's called Yankee stadium. It also has a little metadata built in, including that it's a sports stadium etc. Similarly for things like PetSmart, it has data regarding the GPS locations for each branch and that it's a pet store. This is useful in a couple different ways. The first is that companies like Uber license this data, so when you search for Yankee Stadium, you don't have to put in the actual address. Their primary competence is knowing where things are on the planet, which is a relatively difficult data set to pull together, and their competitors include PlaceIQ and FourSquare (through their new Pinpoint advertising program).



But they're also very much in the ad space. Being a company focused on precise lat-long data, they can only really be effective where they can get this data. This generally means mobile app, though there are a few HTML 5 sites where this is possible. So the most logical conclusion might be enabling a brand like P&G to target users that are physically in Walmarts. Indeed, this is something Factual does. But that's a relatively limited model because it requires both temporal and physical constraints (meaning they must be in the store and looking at ads at the time they're being targeted). And it also relies on the type of stores that an advertiser would want to target. So Factual expanded their model to inferred psychographic models. That's a super fancy way of saying that they could look and see if, for example, you tend to go to pet stores every couple weeks, then you're a person with a pet - and you will be a person with a pet even when you're not in the pet store. And if you're a person that speeds along a train route every day, you're a commuter. With the pet owner example, you can now target all people that exhibit pet owning habits at any time if you're Purina. Given the flexibility of this model, you can imagine their data science team putting together some gnarly models of different types of consumers to meet different customer requests.

There is, of course, more complexity. Most exchanges don't allow third parties to harvest data about their users, combine it with other data, and sell that aggregated data. So Factual actually has to create and keep wholly separate pools of data across all the different exchanges. So you have to target the MoPub pet owners on MoPub - created only from apps that use MoPub, and the Smaato pet owners on Smaato - created only from apps that use Smaato. And so forth. That limits both the quality of the data and the workflow. However, you could apply Drawbridge or Tapad data to your MoPub pet owners segment and expand that reach, where possible. That's a really expensive option, however, since you'll be paying two separate data fees.