Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / Why CMOs Are Abandoning Third-Party Attribution
Transcript
- Lucas: There's this assumption in marketing that if you can just trace every dollar to a conversion, you'll automatically make better decisions. And that assumption is crumbling. Luna: You're talking about third-party attribution. The thing every vendor promises is the holy grail? Lucas: Exactly. For years, CMOs have been sold on these multi-touch attribution platforms — the ones that claim they can tell you exactly which channel, which click, which ad drove the sale. But a growing number of senior marketers are quietly walking away from them. Luna: Why now? These tools have been around for a while. Lucas: A few reasons. One is privacy changes — between iOS updates, cookie deprecation, and new regulations, the data these models rely on is just less reliable. But the bigger reason is that CMOs are starting to realize that even when the data is clean, the models don't actually answer the question they care about. Luna: Which is — did the marketing cause the sale, or did it just happen to be there when the customer was already going to buy? Lucas: Bingo. That's the incrementality question. And traditional attribution models are terrible at answering it. They're correlation machines dressed up as causation engines. Luna: I've heard that term — incrementality — getting thrown around a lot more in the last year or two. But it still feels like a niche concept for most marketers. Lucas: It is niche in practice, but it's becoming central to how top CMOs are thinking. I was reading a case study recently about a large retailer — I won't name them, but think Fortune 500, omnichannel, heavy digital spend. Their CMO decided to run a controlled experiment. They basically turned off all paid search for a subset of their market for two weeks. Luna: That takes guts. What happened? Lucas: Revenue in the test group dropped by less than three percent. But their attribution model had been telling them that paid search drove about twenty percent of total sales. So the model was over-attributing by a factor of nearly seven. Luna: Seven times? That's not a rounding error — that's a fundamental misallocation of budget. Lucas: Exactly. And it's not an isolated story. More CMOs are running these kinds of incrementality tests — geo holdouts, time-based holdouts, even person-level randomized experiments. And what they're finding is that their attribution systems are consistently overstating the impact of the channels that are easiest to track. Luna: Like search and social, where you can drop a pixel. Lucas: Right. And understating the impact of things like brand advertising, outdoor, podcast sponsorships — channels where the attribution signal is weak but the actual lift might be significant. So you end up with a self-reinforcing cycle where you keep pouring money into measurable channels because they look good in the model, even if they're not actually driving incremental revenue. Luna: So what are these CMOs doing instead? Just throwing out their attribution vendors? Lucas: Some are. But the more thoughtful ones are building hybrid approaches. They're keeping the attribution data for what it's good at — directional insights, trend spotting, campaign optimization — but they're layering incrementality experiments on top for budget allocation decisions. And they're moving more measurement in-house. Luna: In-house measurement feels like a big lift for most marketing teams. You need data scientists, engineers, infrastructure. Lucas: It is a lift. But the CMO of that retailer I mentioned told me that after they built their own measurement system, they cut their external analytics spend by forty percent and got better answers. Their team is six people — two data engineers, three analysts, one data scientist. That's not an unattainable size for a company doing hundreds of millions in revenue. Luna: So the pitch is: spend less on third-party attribution tools, invest those savings in a small internal team, and run experiments to actually understand what drives growth. Lucas: That's exactly the model that's gaining traction. And it's not just about cost savings. It's about strategic clarity. When you own the measurement, you can ask questions that the vendor's black box can't answer. Like, what happens if we shift ten percent of our TV budget to streaming? Or, how does our organic social actually influence purchase intent? The vendor's model is built to sell you on their platform. Your internal team is built to give you the truth. Luna: There's a trust issue there too. If a vendor's model says your TV ads are ineffective, you might wonder — are they just trying to push you toward their own digital solutions? Lucas: Absolutely. And the market is responding. I've talked to a few agency folks who say they're seeing RFPs where the client explicitly asks for the methodology behind attribution numbers. They want to know how the model handles view-through conversions, how it deduplicates across channels, what the assumptions are. That level of scrutiny was rare even two years ago. Luna: It sounds like the pendulum is swinging from 'trust the black box' to 'show me the math'. Lucas: Exactly. And the interesting thing is, this shift is happening across industries. I saw a survey from a marketing analytics association — I think it was published in early 2026 — that said about a third of enterprise CMOs now run incrementality tests at least quarterly. That number was under ten percent in 2022. Luna: So it's still not the majority, but it's growing fast. Lucas: Right. And the ones who aren't doing it yet are starting to feel the pressure. Because when your competitor can tell you with confidence that their brand campaign actually drove a five percent lift in store visits, and you're just guessing based on last-click data, you're at a real disadvantage. Luna: If these marketing conversations have sparked something you've actually used — a framework, a budgeting idea, a way to think about a problem — a couple of dollars a month is genuinely what keeps these going. Lucas: Yeah, that's right. We keep the show ad-free by design, and listener support is what makes that possible. If you've gotten something out of it, buy me a coffee dot com slash fexingo. Luna: And it really does make a difference. Even small contributions add up and let us keep digging into these topics without worrying about sponsors. Lucas: So, back to attribution. One of the challenges with moving to an incrementality-based approach is that it requires a different mindset from the entire organization — not just marketing, but finance and the C-suite too. Luna: Because incrementality tests don't give you a neat, tidy number for every channel every month. Lucas: Exactly. They give you a periodic, slightly messy answer. You might run a geo experiment twice a year and get a range — 'our TV spend likely drove between three and six percent incremental lift.' That's not as satisfying as a dashboard showing 'TV drove 4.2 percent of conversions.' But it's more honest. Luna: And more useful for big decisions. Lucas: Right. The CMO I mentioned earlier told me that after they switched to this approach, they actually reallocated about fifteen percent of their total marketing budget. They pulled money from performance channels that looked great in the attribution model but had low incrementality, and put it into brand and content channels that were generating real lift that the old model couldn't see. Luna: That's a massive shift. And it probably made some people inside the organization uncomfortable. Lucas: It did. The performance marketing team was not thrilled. But the CFO actually loved it because the new approach tied marketing spend more directly to revenue outcomes. That's the kind of alignment that CMOs are desperate for. Luna: So what's the takeaway for a CMO listening right now? If they want to start down this path, what's the first step? Lucas: I'd say start small. Pick one channel — maybe paid search, since it's the easiest to run a holdout for — and run a simple experiment. Turn it off for a week in a test market, compare to a control market, and see what actually happens. The results will probably surprise you. And once you have that data, you can start building the case for a more systematic approach. Luna: One experiment at a time. Lucas: Exactly. The goal isn't to have perfect measurement overnight. It's to start getting better answers than the ones your attribution vendor is giving you. Because right now, those answers are costing you real money.