Geo as an identity alternative

Geo as an identity alternative

This week’s vendor interview is with Blis, the DSP recently acquired by T-Mobile. Last week’s podcast guest was Olivia Kory of Haus Analytics, a cross-media measurement company. What do these two companies have in common? They’ve both more or less abandoned user-centric identity and instead rely on geography as the scaffolding for all their decisioning. Let’s dive in.

Geography is destiny

President Obama had a point, but unfortunately is not backed up by the evidence. You can learn a lot about a person based on their zip code. Study after study shows the income, political affiliation, health and a lot else are correlated by home zip code. The popular PRIZM system uses zip code to characterize consumers into groups with fanciful names like "Young Digerati," or "Money & Brains." This has been known for a long time.

Making zip codes sexy again

Geo targeting has historically been seen as crude cudgel in digital advertising due to concerns about accuracy, and the dominant narrative about the value of one-to-one, cookie-based targeting. However, the decline of the cookie, and the persistent inability to reach or measure users on Apple devices have laid bare for many the drawbacks of the cookie-centric approach. In the land of the unidentified, the zip code is king.

Why the return to zip codes?

• Relatively unbiased across devices and platforms

• Broad applicability

• Works across media, including (in theory) TV, social, and YouTube

• Can be linked to demographic and psychographic profiles, arguably more accurately than cookies!

• Cheaper than third-party cookies

• Same “currency” can be used for targeting and attribution.

Attribution?

While using zips for targeting has always been part of the toolbox, attribution, like that done by Haus and others, is a new wrinkle. In the social and search arenas, there’s been an important movement away from pixels (which can be blocked, and are largely an artifact of cookie tracking) to conversion APIs, or CAPIs. CAPIs let the advertiser stream all their anonymized conversion data to advertising platforms like Google and Meta for attribution.

While certainly more correlation that causation, the CAPI data can easily and efficiently be attributed to the exposed media in the same zips, and patterns between media spend and real-world cash registers can be understood. This is a trade off vs cookies and their false certainty.

But with zip code dominant targeting and a sufficiently large budget, you can actually create experiments and test and control groups to move beyond correlation. A couple years back there was a big ruckus in the measurement community around “ghost bidding” wherein some impressions would be held back as a control group in real time. The problem with ghost bidding, like much of the rest of ad tech, was that it was still dependent on identity! With zip codes you can identify statistically similar geos then hold back as needed. You can even extend this methodology to other media.

Problems with zips

There are clear problems with using zip codes as the linga franca of your digital advertising, and probably a bunch have already come to mind.

Accuracy: When zip codes are inferred from IP addresses, the accuracy is not always that great. MaxMind, a leading geo lookup vendor, includes a tool on their website for estimating IP-to-geo lookup. Here are the results I got when I pulled them for US-based postal codes:

In practice the accuracy numbers can be bolstered by looking at other signals like precise geo signals, user-input locations (like on weather apps), and really whatever you can get your hands on to calibrate the data and get better than the out-of-the-box solutions available. Still, if you think zip if 90%+ in accuracy, you are lost.

Privacy/Browsers: Apple introduces their iCloud Private Relay a couple of years ago in an effort to reduce the ability of third parties to use IP address as a tracking vehicle. It also reduces the geographic lookup granularity from zip-level to more like city-level I haven’t seen any stats on how widespread the usage of this tech is, but it remains only available to paying iCloud+ users and must be manually turned on.

Google also has its IP Protection project as part of the Privacy Sandbox (yes, it still exists). This feature is rolling out into incognito mode later this year, but there is no timeline as yet for widespread deployment. This project, if implemented broadly, would also reduce geo accuracy from zip to city.

Consumer Behavior: Consumers don’t sit at home ordering stuff online all day. OK, maybe you do. But most consumers move around, go to work, do stuff, etc. Their associated geographies also change, and so the ad exposures and outcomes may not match. On the conversion side, as well, some sales can be associated with the purchaser’s zip, while others can only be sent with the zip of the store itself.

All of these problems with zip codes are real. For the vendors that have taken on this methodology, they have to build solutions that work through the challenges, and assure customers this old, but newly popular way of doing business has staying power.

Reading list

• DoubleVerify suing Adalytics for defamation! (link)

• Walmart earnings - ads up 50% post-Vizio integration (link)

• Ads are coming to Airbnb, sometime (link)

• Google IO (link)

• IO (OpenAI’s acquisition of Jony Ive’s startup) (link)

• Publicis acquires Captiv8 (link)

• AWS putting TAM into Prebid (announced, no link)

• Magnite in TAM (link)

• A rare win for Google in court, judge tosses Rumble video antitrust case (link)