Subscribe
Share
Share
Embed
10 minutes of expert insights every weekday. Your morning ritual for staying ahead in retail media.
Billed for the Ad Before the Ad Exists: Amazon Sponsored Prompts
===
[00:00:00] Kiri Masters: Well, that's cute. That was my thought when I first saw Sponsored Prompts. That was at Amazon Unboxed last October, where Sponsored Prompts were announced as a way to nudge shoppers into longer conversations [00:00:15] with Alexa for shopping, formerly known as Rufus, and monetize the exchange along the way.
[00:00:20] They seem like a fairly simple way for Amazon to both flex the usage of Alexa for shopping and goad shoppers [00:00:30] into longer conversations by asking follow-up questions. And bonus, a seemingly low-lift way to monetize those conversations. Boy, did I underestimate the new [00:00:45] version of little blue links. Maybe not in terms of immediate ad revenue, but more in terms of how they disrupt what it means to actually be an ad.
[00:00:57] Now last week, Andrew Bell did a massive deep [00:01:00] dive into the patent behind sponsored prompts, which changed my thinking. As a disclosure, Andrew Bell is the VP of research at ReFi Buy, which is a company that I advise.
[00:01:12] Now, the big reveal in this piece is that [00:01:15] with a sponsored prompt, an advertiser is billed for an ad before the ad even exists, and that has big implications for brand side advertisers. Let's jump in
[00:01:28]
[00:01:30] Kiri Masters: So here's how a normal sponsored click works. A brand bids, They win the bid. They pay to send a shopper to a page that the brand built, a product detail page, a brand [00:01:45] store, a category page, an offer, something like that. The brand or their agency wrote the copy for that page. They got it approved, and they knew exactly what the shopper would see when they arrived there.
[00:01:59] A sponsored [00:02:00] prompt, though, inverts that order. The shopper taps a question, the tap gets billed, and then Amazon retrieves evidence and writes the answer. And that answer is [00:02:15] assembled from product reviews and Amazon's first-party signals in addition to what content the brand provides. Each answer is also generated anew, different each time, depending on the [00:02:30] shopper and the context of their question.
[00:02:33] So this raises all kinds of implications about product claims, brand positioning, and understanding which points of your USP are [00:02:45] actually resonating with customers
[00:02:47] Now another important insight here is that the advertiser is billed before any of this content is assembled, and no other types of ads have worked this way before, [00:03:00] at least not exactly. Google has responsive search ads, dynamic search ads, Performance Max. These are all different ways of assembling ad copy on the fly [00:03:15] that have, has been in place for a number of years now.
[00:03:18] Dynamic search ads in particular even generate the headline dynamically from the page content and the query. But this time there are a couple of key [00:03:30] differences. Number one is the sequence. In Google's dynamic formats, the ad, as in the headline, the description, the whole unit, it is assembled before it is shown to the user.
[00:03:43] You see a complete ad, [00:03:45] then you decide whether to click, and the click is what gets billed. ~The creative is existing~
[00:03:51] ~Get rid of that. ~What the patent describes for Amazon-sponsored prompts is the click landing on a question, the click getting billed, and then [00:04:00] Amazon going to retrieve evidence and generate a custom answer. The persuasive content for the ad is built after the money changes hands. Google assembles then charges.
[00:04:14] This is [00:04:15] charge then assemble
[00:04:17] The other key difference in this scenario is who supplies the raw material. With Google's dynamic co-copy, , it is your stuff that is being recombined, your headlines, [00:04:30] your descriptions, your landing page, assets that you wrote and you had your legal team approved. In the sponsored prompt system, the answer is built from those things, but also from [00:04:45] the review corpus, all of the product reviews customers have written, and also Amazon's first-party signals.
[00:04:55] Now, those are things that you didn't author and you [00:05:00] can't fully control. So it is not your copy that is being dynamically arranged. It's Amazon writing a fresh argument about your product from source material that you don't own. Now, so this is raising a [00:05:15] couple of obvious challenges here. The first is liability for claims that you never approved.
[00:05:22] So this is the big one. In every other format, you can control or at least see the creative before it [00:05:30] ran. So this manifests as legal risk and brand risk, and you think about categories especially like over-the-counter pharmaceutical products. This could be a huge concern. Other [00:05:45] sensitive categories
[00:05:46] Who might be concerned with allergies and safety
[00:05:51] All manner of concerns here from a legal perspective
[00:05:56] the other challenge is that the spend [00:06:00] shifts from buying attention to controlling what gets said. So you're not really buying a placement, you're paying for Amazon to make an argument, and you can't control all of the source material that Amazon is [00:06:15] pulling from.
[00:06:16] So advertising in this sense no longer looks like optimizing a bidding strategy and having really great persuasive creative and images. It becomes an operational cost of trying to [00:06:30] optimize the source material, which may or may not really be possible. But you can bet your tail that scrappy brands with nothing to lose are going to try and goose that [00:06:45] source material.
[00:06:46] there is gonna be a part two to this topic that I'll get on my high horse a little bit more about that particular thing tomorrow. [00:07:00] Are manual processes slowing down your RMN momentum? With GrowthLoop's composable commerce media solution, retail media and ad teams can build and launch precise audiences in [00:07:15] minutes, activate across DSPs and ad platforms, and deliver transparent ROI reporting that drives advertiser revenue.
[00:07:26] learn how retail media leaders at Costco, [00:07:30] Fanatics, and goPuff use GrowthLoop to deliver stronger results for their brand partners fast. Visit go.growthloop.com/breakfast. That's [00:07:45] go.growthloop.com/breakfast but wrapping up for today, big question here is, is this form factor actually going to be the [00:08:00] enduring r- one? And that is the big question that rides above all of this, which kind of dictates how seriously brands need to consider this as a, both a challenge and an opportunity, which I'll talk about more tomorrow, [00:08:15] and think about how they invest and how they organize themselves and governance and things like that.
[00:08:22] I would say we're in the very early days of ads in AI and ads [00:08:30] in retailer chat assistance. For now, most of these ad formats kind of look mostly like the ad units we've lived with for the last ten, 15 years. Sponsored product ads, banner ads, video ads. [00:08:45] Sponsored prompts is the first genuinely new shape that we've seen in a while.
[00:08:51] Whether it's the enduring shape or even a major one is still an unknown. Now, Walmart [00:09:00] launched sponsored prompts in Sparky a few months ago, suggesting that the format could become more widespread and adopted by other retailers. But I would say on the Amazon side, my reading [00:09:15] is the expansion of this format has overall been quite relaxed, which is its own kind of signal to me As I mentioned, I'm going to be doing a two-parter on this [00:09:30] tomorrow.
[00:09:30] There was just too much to try and squeeze into one episode. So in part two, we're going to go through what that billing sequence, means in terms of consequences, [00:09:45] get more into the product review ecosystem, and where the work actually moves in this
[00:09:54] new buying structure. So make sure you're subscribed to get the podcast [00:10:00] tomorrow. We'll also link up to Andrew Bell's full patent breakdown so you can read through it all yourself. Lots of really great comments from the community on LinkedIn there as well, and I'll catch you tomorrow
[00:10:13]