Retail Media Networks Decoded

Retail Media Networks - Part II

In our previous post, we discussed how Retail Media Networks (RMNs) benefit both retailers looking for additional monetization channels and advertisers aiming for improved product placement. Now, we delve into the challenges of building such networks.

Getting Started

So, you want to use a Retail Media Network to publish ads, how do I start? Where do I publish my ads? The industry has developed two main canonical channels: Display Ads and Sponsored Listings. Both are integrated within RMNs. RMNs provide platforms for advertisers to showcase products through these mediums, optimizing visibility. Let’s go over each one:

Display Ads

Display ads are fixed spaces within a site design used for advertisements, such as banners, top/sides, or interstitials. At first sight, they resemble traditional publicity signs: there is a given space that seems fit for advertising; similarly to an agency selling billboard space, the RMN reserves it and then sells the space via a specific mechanism.

The display banner has no immediate restrictions. When the advertising space is purchased, it’s up to the advertiser to choose what to publish, be it an entire brand, specific product lines, or individual items (subject to quality guidelines set by the RMN).

So, it’s basically a billboard on the internet? Not exactly.

Unlike roadside banners where your choices are limited to picking a good location (at best!), RMNs enable performance-based decision-making through an abundance of first-party data and metrics.

RMNs handle richer data. They understand individual purchasing patterns through first-party data, enabling tailored strategies and variable payment charges per impression while complying with data privacy regulations. Additionally, unlike most platforms that provide no tracking capabilities beyond displaying banners, RMNs facilitate performance tracking to determine whether a sale can be attributed to a specific impression.

Sponsored Content

Sponsored content (or sponsored listings, sponsored items) is a format used by RMNs. It’s typically shown within item listings, like search results, recommendation carousels, landing pages, or catalogs. RMNs use relevance algorithms to display the most fitting content, such as items matching a user’s query or trending content on a landing page.

Sponsoring content lets advertisers display their items in more relevant locations by paying more. Data plays a pivotal role here – it helps select the best ad available, set pricing, track performance, and more.

TLDR: Here are some differences between the two channels, throughout the rest of the post we will focus on sponsored listings when talking about sponsored content:

The Path to Auction

So there are two channels for ads within RMNs, Display Ads and Sponsored Content (we’ll focus specifically on sponsored listings), and advertisers are willing to pay more for relevant positions. But how much more? How is this determined? The answer is real-time auctions.

When there's a chance for an ad placement, like when a new query is made, the RMN initiates an auction to decide which ad could appear there (it's important to note that paying doesn't guarantee placement, we'll explain further below). Acting as the auctioneer, the platform collects bids from various advertisers, but it only considers those that meet a basic relevance filter. In other words, considering the context, such as the processed query, current trends, and recommendation category, only ads related to relevant content are allowed to participate in the bidding process.

You may be wondering, why the filter? Showing unrelated content may hurt user experience - diminishing platform GMV that may not be compensated by advertisers’ payments.

Data is crucial in determining that relevant subset. This step usually assigns an ad score, measuring relevance for each competing bidder in a given context. For example, headphones and USB keyboards might both be relevant recommendations on a notebook sale page, but if users often bundle headphones in their purchases, then they gain a higher ad score. Note that by performing this retrieval for an auction, the RMN acts both as an exchange and also as a DSP (a DSP is a demand-side platform that essentially links advertisers to ad opportunities).

Once a relevant filtered listing is ready, every item from the set competes by submitting a bid. RMNs, acting as DSPs, leverage their vast data ecosystem and offer ML-based pricing strategies. They assess each competing item based on forecasted conversion rates, client ROI expectations, budget pacing, targeting objectives, and more. A relevant note here when we say an item submits a bid it doesn’t mean the clients set a bid; they set campaign goals. The platform then calculates the optimal bid considering those goals, enhancing budget spending for publishers (bidding high only on the best opportunities) and boosting platform revenue (strategic bidding lessens the risk of high bids, thus encouraging them when fit).

With a filtered listing and submitted bids, it’s auction time! We’ve got an auction opportunity and competing publishers - each submitting a bid and ad score. So how does it work?

Auction Mechanisms

Various auction mechanisms exist, for now, let’s focus on first-price auctions (the highest bid wins and pays said bid) and second-price auctions (the highest bid wins the auction but pays the second-highest bid, plus a cent). Participants are usually ranked by combining their bids and ad scores, often calculated as bid times ad score (the higher the result, the higher they rank).

In a generalized second-price auction, the charged price is the minimum bid that would allow the winning participant to be ranked first. A noteworthy detail is that since more than one listing position may be auctioned up to N positions of the ranking are printed. So, winning may not matter; what matters is to rank high. Even though the winner takes the most relevant position, the second takes the next one, and so on. The ranking and pricing logic is readily extensible.

Selecting the Placement

After identifying an ad opportunity and a winner (or a ranked set of ads), the next step is selecting the optimal spot(s) within the listing for the ad(s) to appear. This varies with site design: static sites have dedicated ad slots, guaranteeing printing upon auction win; dynamic designs make real-time decisions, calculating whether and where to replace positions with ads when generating new lists.

Hybrid methods exist too, combining fixed ad positions with real-time display decisions. As previously mentioned, winning an auction may not guarantee printing an ad. These potential ad display spots are referred to as ”placements” offered by the site.

What Happens After The Print?

By now you’ve understood that Sponsored Content follows some steps before appearing in your retailer's search results. Printing is one more step, but what about after publishing? Tracking performance is just as crucial as the rest of the process: Was it clicked? Did it drive orders, fully or partially?

Performance tracking and attribution call for a post on their own, but it’s worth pointing out that, unlike other ad channels, sponsored content is extremely focused on delivery. Pricing, for example, may be performed on a cost-per-click or cost-per-order basis, thus aligning retailers' and publishers’ incentives to print only those ads with higher conversion chances. Of course, tracking performance also provides feedback on every other step in the process: better audiences are built based on this data, conversion rates are calculated more accurately, thus refining bids, etc.

Wrapping Up

We hope you've achieved a clearer understanding of sponsored content, its role within an RMN, and the key components any such system should implement. If you're feeling like you skipped a beat we recommend checking out our first FAQ post in the series where we cover some essential questions regarding RMNs.