Latest / Retail Media Breakfast Club / Is MTA Past Its Expiration Date? (Part 3)
Transcript
- Kiri Masters's audio recordingKiri Masters === [00:00:00] Kiri Masters: Good morning. Welcome to Retail Media Breakfast Club. This is part three of my series on multi-touch attribution, which I was surprised about this, maybe I shouldn't have been, at the number of impassioned [00:00:15] comments that I've received so far on this series on the LinkedIn versions of this post. It's been a really interesting journey to explore this particular niche [00:00:30] of media measurement and learn how different retailers are approaching it, how different solution providers are approaching it and helping those retailers build those systems. [00:00:41] So in part one, I asked the question, do brands [00:00:45] really want MTA? In the second part of the series, I asked why retailers are reviving MTA. And today I'm going to talk about, is MTA past its [00:01:00] expiration date? And share a little bit more on the case for MTA, especially in its newest iterations, and whether it's really realistic to expect media buyers to act [00:01:15] even if they're able to access perfect information. [00:01:19] So a lot to get into today. Let's jump right in [00:01:22] [00:01:24] Kiri Masters: So first of all, how MTA got its bad name In case you missed it, Albertsons Media Collective [00:01:30] and Best Buy ads both launched multi-touch attribution capabilities in September. [00:01:36] Best Buy built its own, and Albertsons built its version with LiveRamp, who is really trying to avoid actually calling it [00:01:45] MTA. Jack Foster, who is the head of measurement products at LiveRamp, says the company avoids calling it MTA specifically, and that this version that it has launched with Albertsons is really kind of its own solution.[00:02:00] [00:02:00] Now, LiveRamp, the way that they sell it, they call it cross-media intelligence. And Albertsons, with their version of this product, they named it Retail Media Incrementality Based Multi-Touch [00:02:15] Attribution. And my view is when the vendor won't use the name and the retailer feels like it needs a prefix to the name, it suggests that the word MTA itself is giving the [00:02:30] ick. [00:02:30] Now, Jack Foster at LiveRamp blames what traditional MTA included as its data input. In his telling, older MTA systems had data issues that compounded. [00:02:45] So first of all, there's a chain of events here. First is that they combined two kinds of exposure data that don't really fit together. There is the row-level data, such as logs from a DSP [00:03:00] that ties each impression to a specific device through an IP address, mobile ID, or cookie. [00:03:07] And then the other type of exposure data is cohort and panel data that only says that some of a few [00:03:15] thousand people in a zip code saw an ad without saying which ones. And then the second part of this compounding problem is that to make those two kinds of exposure data line up, they randomly mark [00:03:30] people in the cohort as exposed. [00:03:33] And by the end of that chain of events, Jack says that the group that you're learning from might only be in the tens or hundreds of people, and you're about to go and make a [00:03:45] multimillion-dollar decision off of what they did or what really the model is guessing that they may or may not have done. [00:03:54] And so clean rooms then let LiveRamp match [00:04:00] publishers' own logs person by person, and, uh, they don't use these cohort feeds. So he, uh, at LiveRamp, they're su- presenting a, you know, historical example of what [00:04:15] MTA used to look like and that this time around we're doing things differently [00:04:20] Now coming into some of the, the pushback from the earlier two parts of this series. One of them was, uh, one interesting comment [00:04:30] came through from Evan Walsh, who is a senior director at the ad measurement company Incremental, and he said, Eight banners with size four font loaded around the page and can't not get credit with view-based attribution." [00:04:43] What he's calling out [00:04:45] is that better data doesn't fix things like viewability and, you know, what counts as a view. And counting those impressions accurately using the methodology that Jack [00:05:00] described still gives those ads credit for simply appearing on a page, whether or not they actually had any impact on a customer's behavior at the end of the day. [00:05:10] The second, another camp of pushback came from [00:05:15] a consultant who works with retail media networks, and they argued that MTA isn't worth building if a brand can't actually act on that output. And since that article came out, they pushed a little bit further on that view saying, "If a brand sees [00:05:30] 15% attributed to social, 25% to offsite display, 10% to CTV, 20% to onsite display, and 30% to sponsored products, what exactly should they be doing differently?" [00:05:43] To decide where they're gonna [00:05:45] spend the next dollar, this consultant argued a brand would need to know these five things. One, how much it spent on each tactic relative to the credit it received. Number two, which audiences were reached, where and [00:06:00] how often. Number three, how much the tactics overlapped and whether the order of exposures mattered. [00:06:08] Four, which tactics are saturated and which have room to grow. Five, what evidence do we have that shifting [00:06:15] the next dollar from one tactic to another would actually generate more incremental sales? Well, here is the issue. I haven't met a media buyer that has access to perfect information, plenty of [00:06:30] time, and the freedom to move money as soon as results come in. [00:06:35] Most media buyers are satisficing. They are allocating as well as they can within their constraints. [00:06:45] CPGs are often committed to spending a certain amount with a retailer under a JBP. They're capped on the total spend that they can allocate, or they're bound by other minimums in their media plan that [00:07:00] pull them away from spending optimally. [00:07:02] Even if they had this perfect amount of information, they can't always act on it, and they're short on time and people. That buyer needs to split their committed dollars [00:07:15] across one retailer's tactics without time to research overlap and saturation first. A MTA approach that exists solely within one retailer still gives them a better [00:07:30] basis than a last-click attribution model, as imperfect as that is. [00:07:36] So it is... I hate to put it this bluntly, it's, it's better than nothing but Even with perfect information, [00:07:45] I believe most media buyers wouldn't even be able to act on that if they had the world's most perfect attribution model and understood everything about it. [00:08:00] your advertisers can't wait weeks for audiences. GrowthLoop's Composable Commerce Media solution helps media teams turn first-party data into high-value audiences, [00:08:15] launch campaigns faster, and prove what's working across every channel. Learn how retail media leaders at Costco, Fanatics, and Gopuff use GrowthLoop to create highly [00:08:30] segmented audiences and deliver stronger results for their brand partners. [00:08:35] Visit go.growthloop.com/breakfast. That is [00:08:45] go.growthloop.com/breakfast Now, absent from the discussion so far, and this one is completely on me, is actually Amazon. [00:08:55] Amazon announced its MTA in October twenty twenty-four, [00:09:00] and, um, there's actually a paper in August twenty twenty-five by Amazon researchers which explained in detail how this model works. [00:09:15] And so it addresses a couple of those issues that I, I received, from people who are not, uh, fully supportive of, of MTA. [00:09:24] Number one is on those freeloading banner ads. Amazon checks its credit [00:09:30] against holdout experiments where some shoppers are randomly kept from seeing a campaign, and that's designed to stop an ad earning credit just by loading on a page, whether anyone saw it or not, or whether they changed their behavior or [00:09:45] not. [00:09:45] , And then the second pushback on how to practically integrate the insights. [00:09:50] Amazon shows its multi-touch figures next to the usual last touch ones, and then the brands still decide what to do with them. And [00:10:00] keep in mind, this is Amazon describing its own system. The paper doesn't say how accurate the model has turned out to be, and Amazon is grading its own ads, which is, you know, has that marking your own homework, pushback [00:10:14] And [00:10:15] like any retailer's MTA, it only counts the ads that were bought through that retailer and purchases that were made there. And that's the problem that marketers say that they have the most. Siloed or incomplete data was the top reason [00:10:30] that marketers gave for doubting their measurement in a July 2025 survey of 196 US marketers by eMarketer and TransUnion. [00:10:40] And just to note that TransUnion also sells its own MTA product. [00:10:45] Now, Albertsons and Best Buy have that same limit. They only report back sales and exposures made through their systems. And every one of these, Amazon's included, hands the brand a set of percentages [00:11:00] and leaves the decision with the brand [00:11:03] And Jack Foster from LiveRamp thinks that that's kind of the point. "I think that we need to trust brands," he says, "and the retailer partners to be able to make decisions and not spoon-feed [00:11:15] them an answer and have us all pretend that it's 100% accurate," he said to me. [00:11:20] And just wrapping up here, the marketers in this survey from TransUnion and Emarketer aren't finished with MTA. A third of [00:11:30] them plan to invest more in MTA specifically over the next 12 months, and that is close to what they also report as planning to invest in incrementality testing and media mix modeling Colin [00:11:45] Colburn, who is the VP of Commerce and Retail Media at the IAB USA, offered a less flattering reason in the comments on part one. [00:11:53] He said about MTA, " part of why it still has a seat at the table is because so much of the ecosystem relies on [00:12:00] it for compensation, such as performance and affiliate channels." [00:12:03] the checklist for what a media buyer can do with this information describes this idea of an ideal buyer. ~Amazon's approach is to-~ [00:12:12] ~Most buyers, most media buyers are already splitting money they've already committed, and for them, a retailer's MTA only has to beat the last click~ [00:12:12] Most media buyers are going to settle [00:12:15] for a better split of money that they've already committed. That's it for today. Thanks for listening, and I'll catch you tomorrow. [00:12:22]