For years, we've largely treated IP-based geo-targeting as one of the safest assumptions in digital advertising. If I target a ZIP code, city, or DMA, there's an expectation that my ads are actually reaching that geography.
Over the past several months, the
Intermedia Advertising team worked with the team at
Adstra on research examining the stability of the identity and location signals that underpin CTV targeting. What we found challenged assumptions I think many of us have probably taken for granted.
The study found that only 23% of residential IP addresses ultimately reached the intended geographic target when downstream IP mapping approaches were used. To me, that's much bigger than a CTV statistic.
Think about how much of our industry depends on geographic targeting. Local advertising. Retail. Healthcare. Political campaigns. Attribution. Incrementality. Cross-screen measurement. If the location signal is uncertain, every decision built on top of it inherits that uncertainty.
The only companies with deterministic knowledge of where residential IP addresses are assigned are the ISPs issuing those addresses, companies like
Comcast Advertising and
Spectrum Reach. That's exactly what makes their first-party data so valuable.
Nearly every major DSP, SSP, and adtech platform ultimately relies on a relatively small number of IP intelligence providers to power geo-targeting.
MaxMind, for example, states that "it is not possible for us to guarantee 100% geolocation accuracy" and that "GeoIP geolocation data is never precise enough to identify or locate a specific household, individual, or street address."
Somewhere along the way, our industry stopped treating those signals as probabilistic and started treating them as deterministic. I think that's one of the biggest disconnects this research highlights.
This isn't happening in a vacuum.
Truthset's State of Data Accuracy report raised broader concerns around identity quality in CTV and their research with
Go Addressable &
Coalition for Innovative Media Measurement (CIMM) highlighted challenges with downstream IP targeting.
Our research adds another piece of evidence that we need to pay far more attention to the quality and confidence of the signals powering our industry.
None of this means IP intelligence isn't valuable. It absolutely is. But it does mean we should stop pretending probabilistic signals are deterministic ones.
As marketers, we don't benefit from putting our fingers in our ears when multiple studies point in the same direction. We benefit from acknowledging the limitations, measuring confidence in our data, and building better standards around it.
This conversation becomes even more important as AI takes on a larger role in planning, optimization, measurement, and attribution. AI doesn't know whether a signal is right or wrong. It simply optimizes against the data we give it.
Thank you to
Kendra Barnett &
ADWEEK for covering. Link in the comments.