Verification is emerging as one of the biggest themes of 2017 in adtech. Brands like P&G are placing new demands on agencies and media providers for increased transparency and accountability, which in turn has led to significant advances in areas like viewability. The next battleground is clearly verifying the data accuracy of the underlying ingredients — namely the quality of data used for targeting and measurement.
In mobile, carriers are emerging with an unrivaled truth set for verifying the accuracy of data and ad offerings distributed across their networks. The same carrier networks that we rely on to ensure that our always-on devices work all the time provide a system of record that can be used to help advertisers understand what data is real and working for their campaigns. And as truly agnostic players, carriers can use this data to simply verify other data sets without ever exposing individual consumer information.
Outside of Facebook and Google, fragmentation plagues the mobile data landscape. Aggregation has become the name of the game for achieving scale — meaning assembling disparate data sets from a multitude of sources (such as SDKs from different apps) with varying degrees of accuracy, in a quest to stitch together a behavior graph of sufficient scale and reliability.
While this approach may work for building a segment or campaign, it isn’t suited for data verification. Differences in how data is collected, the frequency of observations, and even the collection methods used by the app or SDK tapping the OS (not to mention the differences between how the OSs work), all create anomalies in the data signals. For example, in the context of location data: a dating app and a weather app may access different locations (a bar vs. the airport) using different methods (wifi vs. GPS or beacons), thus yielding different levels of accuracy. Add the curbs that both Google and Apple are putting on the collection of user data like location, not to mention the impending challenges of GDPR in Europe, and accuracy at scale will only be all the more challenging.
By contrast, carriers provide unparalleled scale, data completeness, and the added benefit of mutual exclusivity, as most consumers have only one carrier. Carriers also have access to different technologies for collecting location data, from cell towers to tri/multilateration to GPS. This allows to store and use verified accurate location from cell towers and also have some records coming from multilateration with high precision up to 10 meters. As a truth set for mobile media consumption and movement, this data is the most pristine and uncorrupted mobile record that any data scientist could hope for.
Precision vs. Accuracy
Mobile data purveyors have woven a narrative about the precision of their data in order to command a premium. This narrative seems logical – after all, mobile devices are with consumers all the time – so shouldn’t the data exhaust be perfect and persistent? But while this is true in some cases, it is not universally the case. You could say that location data consistency is the hobgoblin of the media buyer. Unfortunately, it just isn’t that perfect or persistent all the time.
As an example, location vendors often use the word “precise” referring to the number of decimal places of a lat/long location fix. Buyers are often sold by claims of hyper-precision and perfectly persistent signals. But the reality is, while a small amount of mobile data delivers on this promise, a much larger portion is intermittent at best, with varying degrees of precision.
(For the uninitiated, each decimal place on a lat/long increases resolution by 10x. One decimal place is a square roughly 11KM on a side, the third decimal digit is worth up to 110 meters and can identify a large agricultural field, campus, or city block – usually considered privacy safe.)
No doubt some of the data from SDKs and apps is really precise – but precision has no impact on accuracy. How? A user’s location may be reported by an ecosystem player as coming directly from the center of Times Square (40.759000° N, 73.984500° W) when in reality that user may be anything from 2 to 2000 miles away from that. All the predictable reasons are at play here – from fraud to technology configurations and anomalies. What’s important to note is that without a truth set, advertisers have no way of knowing what’s real.
Contrast this with data from a consistently correct signal, as measured by an extensive system of record only available via carrier data; You can turn the dial in for large-scale accuracy or smaller-scale precision.
SDKs and panels measuring other SDKs and panels are a house of cards
Emerging verification services in mobile come in a number of different flavors, from apps and SDKs collecting user data, to implicit and explicitly recruited panels. Again, SDK data can be really useful for targeting, but using one SDK or app to measure another stretches methodological credulity for all the reasons outlined above. The inherent errors created by the technology, the sometimes bad behavior of the ecosystem, and the shortcomings in scale all make this data wrong for verification. Add in the data privacy issues and the very real threat of Google and Apple cracking down on background location tracking, and the fact that “panels” are recruited to represent niche population groups, and you begin to see why these approaches are not accurate for verification.
Carriers, by contrast, offer census-level data – representing every demographic and lifestyle group at large scale. Furthermore, the underlying data and technology was designed for high reliability, large scale and low latency use (ensuring that your calls aren’t dropped when you drive down the road). This offers a simple, elegant yardstick for verifying the accuracy of other mobile data sets that are best suited for targeting. And given their role in the mobile ecosystem more broadly, they provide an excellent, agnostic truth set.
Alistair Goodman brings over 20 years of experience working in marketing and product development efforts for media and data technology companies. Currently, as the CEO at Placecast, Alistair leads a team of location and mobile data management experts. Prior to Placecast, Alistair was vice-president of strategic marketing at Exponential Interactive.