May 29, 2026 · 41 min · 13 segments
Most marketers are still making million-dollar decisions based on dashboards they’ve been trained to trust.But what happens when the data itself is flawed?On this episode of Deconstructing Data, David…
David FinkelsteinHostJessy LezakHostUm, we also needed it explained 'cause we saw kind of a lot of anomalies in the system.
And, um, as you all know, being in ad tech, there was kind of this explosion of, uh, ROAS being used to measure success, almost myopically so.
Um, but then we noticed a couple years ago there started, or stopped rather, being a correlation between ROAS and the business outcomes.
So how that played out is, um, you know, in environments where attribution was being measured, that return on ad spend was going up and up and up, and business outcomes would remain flat, um, or even down in some cases.
Uh, we couldn't make sense of how return on ad spend could go up, um, and not have the business revenues truly go up.
We also saw that, um, data providers who were providing a lot of these attribution datas through panels, um-Their panel data sizes were getting cut in half.
Um, so they were taking really small sample sizes and getting even smaller, looking at 1% or less, um, of audiences.
For example, for mobile geolocation data, those footfall numbers were falling to 1% of actual visitation.
So it was kind of getting hard to make sound decisions off of some of that data.
Um, and then from our own research and study, [laughs] not even research, just practical, you know, um, ads in market, we were doing log file transfer matches, um, and those match rates were getting smaller and smaller and smaller.
Um, around that same time, the multiplier effect was published by WARC, and that was kind of the first time that we started to say, "Okay, maybe there is a, like, a reason behind all of these things happening at once." Um, what the multiplier effect did in publishing that data around effectiveness, taking, you know, 30-plus years of ad effectiveness studies and finding that, um, audience segmentation was one of the smallest drivers of brand growth and profitability as compared to overall channel allocation or, um, geographic allocation and creative.
Um, so we thought, "Maybe there's something going on here larger." So we actually hired a New York Times, um, investigative journalist to come in and help put all of this together.
We kind of turned over all of our first-party research, um, and started to give her some leads and directional information, um, about where to turn, and the attribution illusion was born.
I mean, we talk a lot on the show about attribution and sort of the, uh, the, the many different factors that come into play with respect to different platforms and, and sort of everybody sort of wanting to take, uh, credit for the attribution.
Um, and so how that can, you know, when you're looking at your overall attribution, it can really distort things, right? And so you have, you know, X number of people that, you know, converted and, you know, Y number of people that converted when you look at it across, uh, all the different platforms.
And so, um, we've always seen that from our perspective as being sort of the illusion, right? In the sense that, um, you never really knew where the attribution, you know, who really was responsible for that attribution, um, because there were so many different places claiming, claiming, uh, responsibility for it.
Um, and it's interesting, um, what you said earlier, uh, also about just the measurement, um, and, and the, the sort of, uh, anomalies that can occur in, in measurement as well.
And a lot of that, you know, we wrote, um, when it comes to ROAS, you know, we wrote a white paper several years ago, um, about the effects of ad fraud, um, and just removing, um, a number, you know, the bots and click farms and the IDs that are related, uh, or linked directly to this ad fraud, um, how a simple change like that, if you could do it, um, you are able to improve your return on ad spend.
Um, we also needed it explained 'cause we saw kind of a lot of anomalies in the system.
And, um, as you all know, being in ad tech, there was kind of this explosion of, uh, ROAS being used to measure success, almost myopically so.
Um, but then we noticed a couple years ago there started, or stopped rather, being a correlation between ROAS and the business outcomes.
So how that played out is, um, you know, in environments where attribution was being measured, that return on ad spend was going up and up and up, and business outcomes would remain flat, um, or even down in some cases.
Uh, we couldn't make sense of how return on ad spend could go up, um, and not have the business revenues truly go up.
We also saw that, um, data providers who were providing a lot of these attribution datas through panels, um-Their panel data sizes were getting cut in half.
Um, so they were taking really small sample sizes and getting even smaller, looking at 1% or less, um, of audiences.
For example, for mobile geolocation data, those footfall numbers were falling to 1% of actual visitation.
So it was kind of getting hard to make sound decisions off of some of that data.
Um, and then from our own research and study, [laughs] not even research, just practical, you know, um, ads in market, we were doing log file transfer matches, um, and those match rates were getting smaller and smaller and smaller.
Um, around that same time, the multiplier effect was published by WARC, and that was kind of the first time that we started to say, "Okay, maybe there is a, like, a reason behind all of these things happening at once." Um, what the multiplier effect did in publishing that data around effectiveness, taking, you know, 30-plus years of ad effectiveness studies and finding that, um, audience segmentation was one of the smallest drivers of brand growth and profitability as compared to overall channel allocation or, um, geographic allocation and creative.
Um, so we thought, "Maybe there's something going on here larger." So we actually hired a New York Times, um, investigative journalist to come in and help put all of this together.
We kind of turned over all of our first-party research, um, and started to give her some leads and directional information, um, about where to turn, and the attribution illusion was born.
I mean, we talk a lot on the show about attribution and sort of the, uh, the, the many different factors that come into play with respect to different platforms and, and sort of everybody sort of wanting to take, uh, credit for the attribution.
Um, and so how that can, you know, when you're looking at your overall attribution, it can really distort things, right? And so you have, you know, X number of people that, you know, converted and, you know, Y number of people that converted when you look at it across, uh, all the different platforms.
And so, um, we've always seen that from our perspective as being sort of the illusion, right? In the sense that, um, you never really knew where the attribution, you know, who really was responsible for that attribution, um, because there were so many different places claiming, claiming, uh, responsibility for it.
Um, and it's interesting, um, what you said earlier, uh, also about just the measurement, um, and, and the, the sort of, uh, anomalies that can occur in, in measurement as well.
And a lot of that, you know, we wrote, um, when it comes to ROAS, you know, we wrote a white paper several years ago, um, about the effects of ad fraud, um, and just removing, um, a number, you know, the bots and click farms and the IDs that are related, uh, or linked directly to this ad fraud, um, how a simple change like that, if you could do it, um, you are able to improve your return on ad spend.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
