Retail Media - Saint or Sinner?

My post yesterday, asking "anyone paying for ads through "retail media"? got so much inbound, I'm writing this short post to answer the burning question "what the heck am I getting when I buy "retail media"? Note that I am not talking about the in-store screens or the ads that run on the retailers' websites. See below for the FouAnalytics data showing ads running on the crappy sites and apps you tried to avoid in open programmatic years ago. When you buy "retail media" where do you expect your ads to run?

Context and history - "retail media networks" (RMN)

Let's start with some context, to understand why some advertisers think "retail media" makes sense. Everyone know what retargeting is -- showing ads to users that had visited your site before, because that seems to indicate interest. Everyone also knows what remarketing is -- showing ads to users that had purchased from your site before, because academic theory says "getting an incremental sale is cheaper than getting a sale from a completely new customer." Both of these are fine as concepts. But like most things in adtech and digital marketing, what starts off as a good idea devolves into something that is mostly useless and costly when fraudsters use it to commit fraud. For example, ad fraudsters simply have their bots visit your site first, before loading pages on cash-out sites; this way, the bots get higher CPM ads due to "retargeting." In case after case of retargeting, as measured by FouAnalytics, we see higher incidence of dark red (note the color coding in each row).

You may be wondering how bots do this for remarketing? Bots deliberately add items to cart and abandon the cart. Advertisers who are desperate for the sale will pay higher CPMs to show ads to users that abandoned their cart, not knowing those were bots. In the mobile app world, fraudsters harvest cookies and device IDs of real human users, send them back to their "command and control" servers, and re-distribute said cookies to their network of bots so they can impersonate that person. Some of these are persons who have purchased before, and advertisers using remarketing will end up showing higher CPM ads to these impersonator bots.

Targeting down to the individual NPI number

Over the years, many advertisers believe that in digital advertising, you can show the right ad to the right person at the right time. And yes, in theory, you can. But what actually happens in practice? Pharmaceutical advertisers upload lists of NPI numbers (national prescriber IDs) of the doctors they want to target with their ads to data brokers like LiveRamp. These data brokers find matching cookies, from various cookie pools, such that when those cookies show up in the bid request, the ad buyers can show ads to that specific NPI number -- thinking they were showing ads to a specific doctor. Again, that's nice in theory. But, the reality is that data brokers usually identify 2 - 3 dozen cookies that are probably associated with that NPI number. They don't know for sure, but their magical matching algorithms suggest those cookies are related to that single doctor. This article is not about how inaccurate cookie targeting is, but you can surmise that it is pretty bad. That cookie may be for the right iPad in the doctor's house, but the ad was still shown to the 4 yr old playing kids' coloring app on said iPad; the oncology ad was not shown to the oncologist, even though the pharma company had uploaded a list of NPI numbers they wanted to target.

Not only are the ads shown on bad apps and sites, some of these fraudulent mobile apps also click through to the landing pages of the advertisers, to trick them into thinking the ads are performing. When FouAnalytics is used to measure the landing pages, we see many examples of mobile apps clicking through to the url (landing pages of the advertiser). The advertiser thinks the hyper-targeted ads are working, whereas in fact all the clicks were faked by the same bad apps that loaded the ads fraudulently in the first place.

Retail media is the latest shiny object

This brings us to the question at hand. What exactly am I buying when I buy "retail media?" The theory here is the same as retargeting and targeting by uploaded lists. Retailers have first party (1P) audiences -- their customers and/or their website visitors. They upload these cookies to data brokers to find matching cookies. This way, whenever these audiences show up elsewhere (other sites and apps) they can show ads to them, as if they are "the retailer's audience" -- hence the name "retail media." But what about the cookie matching inaccuracy issue? What about the bot-visiting-retailer-site-first issue? Right, it's all the same. And we've seen this movie before, many times over. It's all the same crap. It's like the retailer doing audience extension. Of course, there's no malicious intent, but the outcome is the same.

Now that I have measured ads in "retail media" campaigns, I can show you it's pretty much the same crap. Advertisers using FouAnalytics in-ad tags, measured where their ads went. In the following 4 campaign examples, what problem(s) do you see in the data grids? I'll give you a minute to think about it, and will reveal the answer below.

Too much of the ads went to the top 2 mobile apps. (Apparently humans have nothing else to do during the day but play solitaire and crossword). All of the rest of the sites and apps observed are the same sites and apps you can buy on open programmatic, for far cheaper. Every vendor selling "retail media" will claim that "oh yes, these are our users, and they are playing these mobile apps and visiting these sites." This is exactly how the agencies selling "ads targeted down to the NPI number" explained why ads were showing up on sus apps and sites.

I invite you to test whether the clicks from retail media yield ANY more attentive users on your landing pages.

So is retail media a saint or sinner?

I am not saying that all retail media is bad. I am just saying that you need to know what you are buying (measure with FouAnalytics in-ad tag) and whether those users arriving on your site did anything -- i.e. were "attentive" (measure with FouAnalytics on-site tag).

I have already set up the advertisers who asked about this with in-ad tags and on-site tags so they can "see Fou themselves" -- i.e. what they are getting in their "retail media" buys (example data above). If you want to measure with FouAnalytics too, just message me privately.

Further reading: https://www.dhirubhai.net/in/augustinefou/recent-activity/newsletter/

P.S. Yes, this is just like the problems observed in (fr)audience networks. If the ads show on the retailer's own site; it's fine. If the ads are shown offline in the retailer's stores, it's fine. But when the ads show up on thousands of other bad sites and apps (that no one has ever heard of) that's a sign that bad guys have infiltrated your "retail media" buys as well. No need to panic, just need to measure it and add bad sites and apps to your block list.

https://www.dhirubhai.net/pulse/audience-networks-drive-massive-volume-eat-up-your-budgets-fou/

P.P.S. Even if you buy direct, be sure to ask if the publisher/seller is doing "audience extension." In the example below, CTV ads were not running on large screen TVs, but on crappy sites (as video ads). That is not what the advertiser paid for when buying CTV ads at CTV prices. The same kind of fraud and abuse is happening because ad buyers are chasing "scale" that real RMN don't have.