Starting a Retail Media Network: Rent a Platform, Use an Ad Server or Build Your Own

A retailer or marketplace can get the technology for its own ad business in three ways. It can rent a retail media platform that runs the ads, the auction and the brands' campaign tool. It can use an ad server through its API and build the parts around it. Or it can build the whole system. Renting lets you learn quickly whether brands will buy your ad space. Building pays off in three cases: your ads must follow your own search results and stock, your shopper data must stay in your systems, or platform fees have grown larger than the cost of your own team.

What is a retail media network?

A retail media network sells advertising to the brands whose products a retailer or marketplace sells. One format is the sponsored product: a paid listing in search results or on a category page that looks like the listings around it. Other formats are banners on the site and in the app, ads shown to the retailer's shoppers on other sites, and screens in stores.

The Interactive Advertising Bureau (IAB) and the Media Rating Council (MRC), the US body that accredits advertising measurement, published Retail Media Measurement Guidelines in January 2024. They were read on 29 October 2026, like the other sources here, and they cover ads that “appear onsite, offsite, online, and in-store”.

What does the technology have to do?

• For each search or page, choose which sponsored products to show: only products in stock that match what the shopper is looking for, and fast enough not to slow the page
• Run an auction between the brands' bids and keep each campaign inside its budget
• Give brands a place to create campaigns, set bids and budgets and read results, or give your own team a tool to do it for them
• Count impressions and clicks, and connect them to sales in your own order data
• Filter out bots and scrapers before anything is reported or billed
• Invoice brands and keep the records behind each invoice
• Keep shopper data inside your privacy rules, including when ads run on other sites

What are the three ways to get it?

• Rent a platform. Criteo's retail media page tells retailers: “Earn new revenue by monetizing your audiences and digital channels.” Its product for them is Commerce Yield. Topsort describes its Core platform as “Ad serving, auctions, self-serve and measurement in one platform.” This is the fastest start, and the vendor's system decides how ads compete, within the settings it gives you.
• Use an ad server through its API. Kevel sells an API-based ad server with a Console where advertisers manage their own campaigns. Its article on this setup, updated in April 2026, describes the order: “you first ask your search tool what listings are eligible to be shown”. The ad server then runs “the auction against that set”. Your team builds whatever the ad server does not cover; ask the vendor for that list before you sign.
• Build your own. The ad decision is part of your search and ranking, the auction rules are yours, and the data stays in your systems. You also take on everything in the list above: the decision, the budgets, the brand tool, measurement, filtering and billing.

What will brands ask about your measurement?

Brands that buy from several retail media networks compare them, so the same questions come up everywhere. The IAB and MRC guidelines answer most of them:

• When does an impression count? The guidelines say “IAB/MRC Guidelines require begin-to-render measurement of digital creatives”: an ad counts once it starts to appear on the screen, not when your server chooses it.
• Was the ad seen? The MRC requires viewable impressions when sales are attributed to ads. The guidelines accept that a retailer usually cannot give every advertiser its own viewability tag on its site, and recommend that the retailer still provides viewability figures.
• How many days after an ad does a sale still count? The guidelines say typical windows “are often 3, 7, 14, 28, or 30 days”, and that windows “must be disclosed before campaign execution and measurement”.
• Which products were counted? The guidelines recommend reporting for each product (each SKU) and being open about any “halo” products counted, meaning sales of products other than the ones advertised.
• How much of the result is estimated? Some shoppers cannot be linked to an ad, for example when they pay in cash. The guidelines ask retailers to disclose what share of the attributed sales is estimated.
• Were bots filtered out? The guidelines warn that retailers may be “subject to a high degree of automated activity such as crawling and scraping for pricing and competitive information”. They say such activity “should be filtered from reported/monetized measurement”.

The guidelines also say that all digital measurement for retail media should at least meet the MRC's requirements for general invalid traffic. If sales figures are used to price your ads, they encourage “third-party audit and validation”. Plan for an auditor to read your logs.

When does building your own pay off?

Building your own system, or building around an ad server, pays off in these cases:

• Sponsored products must follow your own search ranking and live stock. An ad system that does not know your stock will show ads for products a shopper cannot buy.
• You want your own auction rules, such as how far a paid listing may rank above what the shopper searched for
• Shopper and sales data must stay in your systems, for legal reasons or because it is part of what you sell
• Your ad revenue has grown to the point where platform fees cost more than a team that runs your own system

The second route sits in between: an ad server runs the auction through its API, and your own search decides which products may be shown.

What should we ask a firm that builds or connects it?

• How does the ad decision get the products that are in stock and match the search, and how much time does it add to the page?
• How is an impression counted, and does that match the IAB and MRC definition?
• Which attribution windows will brands see, and can they get day-level data to compare windows?
• How are bots and price scrapers filtered before anything is billed?
• Where do brands manage campaigns: in our tool, the vendor's, or both?
• If we start on a rented platform and move to our own, what happens to campaigns, history and brand accounts?
• How does a rented platform charge, and can we export all our data at any time?

Where does amBrain fit?

amBrain, which publishes this blog, builds two of the systems described above. One of its lines reads: “amBrain builds ad servers: targeting, frequency capping and reporting.” Another reads: “amBrain builds event analytics pipelines for ad tech: collection, processing and reporting of impression and click events.” Its AdTech work includes RTBBidder, a demand-side platform it built for a client.

It has been building software since 2019 and works in three formats: full delivery, a dedicated team, or engineers embedded in your team. Its line on ownership is “The client keeps full ownership of the product and the code, except our reusable components.”

This article is not a case study. It does not say that amBrain has built a retail media network for any client, and it gives no prices or timelines. If amBrain is on your shortlist, give it the same questions as any other firm.