Subscribe
Share
Share
Embed
10 minutes of expert insights every weekday. Your morning ritual for staying ahead in retail media.
SNIPS - This Agency Abandoned MTA, Here's Why
===
[00:00:00] Kiri: There's a particular kind of vertigo that you get when you watch a podcast that you started get better without you. I launched the Ecommerce Braintrust podcast in 2017, and it is run now [00:00:15] by my former colleagues at Acadia. And I still listen to, to most of the episodes because these guys are in the trenches every day doing the thing
[00:00:26] I continue to have a lot of respect for that team. A [00:00:30] recent episode about measurement Really got me thinking because this is the question on a lot of brands' minds Which is if every platform that you're advertising on says that [00:00:45] it is winning, that it has the best audiences and the best results, why don't our business results reflect that?
[00:00:53] And if we add up all of these ad-attributed sales from the various platforms that we're advertising on, somehow it ends up [00:01:00] being much bigger than the actual sales that we brought in? Hmm.
[00:01:04]
[00:01:05] Kiri: So on this episode of the eCommerce Braintrust Podcast, the hosts, Julie Spier and Jordan Ripley, brought in two people who have to [00:01:15] answer that question for a living.
[00:01:17] one was Sally Kazen, Acadia's head of analytics, and Ross Walker, their director of retail media
[00:01:24] Now let's listen to some highlights from this episode
[00:01:27] julie: What is different today compared to three [00:01:30] years ago when brands were dealing with a similar conundrum of like, what is each channel driving? How much is my investment returning?
[00:01:38] sally: Yeah, Julie, great question. The short answer is that we've transitioned from a world of observation to a world [00:01:45] of estimation, where platform conversions are modeled, but our math, in many cases, simply hasn't caught up.
[00:01:52] Three years ago, we were living in a more trackable and more investment efficient digital landscape. Today, we're [00:02:00] trapped in the silo paradox. Brands are spending millions across Meta, Google, Amazon and others, and every one of those platforms is acting as its own judge. They're heavily incentivized to claim credit for every [00:02:15] conversion they can possibly touch.
[00:02:16] And so like you said, when you aggregate the revenue reported by those platform dashboards, it often exceeds what's actually hitting the bank account, which we call
[00:02:25] engine inflation. The gap between the dashboard numbers and the ground [00:02:30] truth has become impossible to ignore. You could end up celebrating platform success, but in the context of a lagging business, and that combined with record CPMs results in significantly wasted budget.
[00:02:43] Most brands haven't [00:02:45] updated their measurement to match this reality.
[00:02:47]
[00:02:48] Kiri: Okay, so engine inflation, I love this term from Sally. It is a better name than double counting because it really points at the cause [00:03:00] rather than the symptoms. Every platform is incentivized to claim every conversion it can plausibly touch, so of course the sum exceeds the bank account, and that is just the [00:03:15] system working as designed.
[00:03:17] And ultimately, retailers are trying to give brands what they want, which is measurability, attribution, showing results. These are the things that brands have asked for, so we are in [00:03:30] a prison of our own making at the end of the day
[00:03:33] But also interesting from this is that Sally pairs inflated credit with record CPMs. Over-attribution alone is an [00:03:45] accounting problem, but over-attribution while you're paying peak prices for the impression is a serious issue. So what replaces it?
[00:03:56] Here's a spoiler. Multi-touch [00:04:00] attribution is not coming to save us
[00:04:03] julie: But Sally, this is something that you've moved away from completely. Can you talk a little bit about that decision and what you're using now to really be the source of truth?
[00:04:13] sally: Absolutely. And you are [00:04:15] correct. The promise of MTA was beautiful. I think we all wanted to believe. But the reality of twenty twenty-six is that we aren't seeing a full picture anymore.
[00:04:24] So between privacy restrictions, for example, Jordan mentioned a couple, fragmentation of the landscape, [00:04:30] we're seeing less and less of that actual customer journey. And in most cases, the majority of touch points are invisible to us. So if you're trying to build a business strategy on that level of visibility, it's like you're navigating a [00:04:45] new city with a map that only shows every few streets and filling in the gaps with guesswork and gut feelings.
[00:04:52] So you're correct, at Acadia, we made the call to, to move away from MTA. Instead of trying to track what's [00:05:00] essentially impossible today, we triangulate the truth by using a three-pillar framework. So the first is modern or media mix modeling, which helps with the strategy. We use a, quote-unquote, "glass box [00:05:15] versus black box" Bayesian model that analyzes total spend versus total sales.
[00:05:20] That removes platform bias, and that shows what's actually moving the business needle. And then second, and Ross mentioned some of this for specific platforms, is [00:05:30] incrementality testing. So you wanna have some kind of reality check Out in the wild, if you're using synthetic controls or lift tests, you can prove causation.
[00:05:39] Did the revenue actually happen because of the spend, or would it have happened anyway? [00:05:45] And then third is platform optimization, so the engines, and this is where it gets tactical. We feed those insights back into the bidding engines, so media buyers are optimizing for actual business [00:06:00] growth. Platform metrics like ROAS and CPA are still critical levers here too.
[00:06:06] They just need to be in service of maximizing platform performance and the business. Yeah. And so overall, you're moving from a [00:06:15] dashboard that guesses at the journey to a system that quantifies and proves the impact. And here's where this gets really critical, actually. This is a, an important point. With retail media networks or RMNs growing 30-plus percent [00:06:30] year over year, maybe even more, many brands have separate teams managing Amazon, Walmart, et cetera, versus traditional paid channels, and each are often optimizing in isolation.
[00:06:43] And so drives you to look at total [00:06:45] spend across RMNs and paid channels versus your total revenue. Without that holistic view, to some extent you're flying blind.
[00:06:53] As scientific as it is, there's certainly an art to it, like I'd mentioned a few minutes back.
[00:06:58] And MMM outputs [00:07:00] should not be the only inputs into big budget decisions. There are also other considerations like the business goals or the creative that the brand has available to use. So pairing with incrementality testing helps you ensure that, yes, [00:07:15] we believe the model and it's right, or do we need to refine in some places based on real world results and feed those insights back into the model so it can be fine-tuned. [00:07:30] 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, launch campaigns [00:07:45] faster, and prove what's working across every channel. Learn how retail media leaders at Costco, Fanatics, and Gopuff use GrowthLoop to create highly segmented [00:08:00] audiences and deliver stronger results for their brand partners.
[00:08:04] Kiri Masters: Visit go.growthloop.com/breakfast. That is [00:08:15] go.growthloop.com/breakfast
[00:08:16] Kiri: And finally, in this last clip, host Jordan Ripley asks Ross how this plays out for omnichannel brands running media across Amazon, Walmart, and other retail media [00:08:30] networks. Ross talks a- about an account that he inherited at Acadia
[00:08:34] ross: This is a really good example of a brand who's-- we took over. The way they had defined the inputs in the past was by retail media network and ad type. So they were like Walmart sponsored product versus [00:08:45] Amazon sponsored product. And our philosophy as an agency is that they are not the same thing, right?
[00:08:50] How you're spending the dollars, the goals that you're spending them against, if it's defensive versus offensive, if it's branded versus non-branded, really makes a difference in the [00:09:00] increment. That was our experience before we worked with the analytics team, right? That if you want maximum incrementality, you can't just dump budget into brand defense, like into branded keyword.
[00:09:08] What we did when we worked with Sally to build the model is say, "We don't just wanna know sponsored product versus sponsored product. We wanna put [00:09:15] different goals up against each other," right? Whether it's upper funnel display versus lower funnel branded search or non-branded search, we want a model that helps show what the, what the-- and what is the output for each one of those strategies distinctly.
[00:09:29] And [00:09:30] there's a limit to how much of that you can do. But one of the things that, that we've found is that the challenges of MMMs that I've seen from other clients were like, "Put more money into sponsored products." And I was like, that's-- And as soon as we get into the details, we see that that's just because it's showing a high [00:09:45] ROAS because it's all defensive.
[00:09:46] And as soon as we dig into that, we know that's not the right approach. And so you have to understand what the inputs are to make it more effective. And that framework of measuring incrementality by retail media network, by ad type, by goal, [00:10:00] is probably one of the most effective ways to accurately measure what's going to improve my sales velocity or what's gonna have the biggest impact on my business.
[00:10:08]
[00:10:09] Kiri: So what really went wrong here? The model wasn't broken. It was fed [00:10:15] inputs that were defined as the retailer and the ad type, and the model faithfully reported that sponsored products were performing well, which they were. Because the spend was largely [00:10:30] defensive on branded terms, harvesting demand that already existed.
[00:10:33] It was doing what it had actually set out to do. But of course, that is not going to grow the brand. It's not going to grow incremental sales
[00:10:44] The point I [00:10:45] think Russ is making is that a model that hasn't been tuned to the real outcome that a brand wants is going to basically tell you to keep buying sponsored product ads no matter what your real [00:11:00] priority is. Defensive branded search always looks efficient
[00:11:05] And this is really the whole point to me is there is a
[00:11:10] Bit of a blame game that happens arou- in our industry [00:11:15] around broken attribution. We want better measurement, and this episode is a useful corrective because it shows that the failure mode isn't just having the right measurement and metrics, it's [00:11:30] actually
[00:11:30] one level up from that.
[00:11:32] I felt this was a great episode. I always love to keep on top of Ecommerce Braintrust after all of these years. I do recommend that you check it out
[00:11:41] Thanks for listening, and I'll catch you tomorrow
[00:11:44]