Criteo is an independent retargeting platform that brings back visitors who browsed products but left without buying, showing them dynamic ads built from the exact products they viewed. Unlike Google Ads’ and Meta’s own remarketing tools, Criteo isn’t tied to a single publisher — it can buy ad space across thousands of sites at once, and it does this through a fully automated, product-based ad engine.
How Does Criteo Work?
Three pieces work together at the core of Criteo: a tracking tag running on the site, a data feed that supplies the advertiser’s product catalog, and a bidding engine that matches the two and assembles the ad in real time.
The tag records which products a user viewed, added to cart, or bought. This data alone doesn’t produce an ad; combined with the price, stock status, and image data in the product feed, the ad is assembled automatically. If a product goes out of stock or its price changes, the ad shown to a user who previously saw that product changes the moment the feed updates — no manual intervention from the advertiser is needed.
After the user leaves the site, Criteo catches the moment that person visits a page somewhere in its ad network (the thousands of publisher sites, apps, and social platforms Criteo has deals with) and bids for that ad slot in real time (RTB). A machine-learning model that estimates the user’s likelihood of converting, based on past behavior, sets the bid amount — so the same user doesn’t get the same bid on every impression.
This mechanic resembles Google Ads’ dynamic remarketing feature; the difference is that Criteo’s inventory isn’t limited to Google’s own network, and its bid optimization runs purely on Criteo’s own data pool. That broad data pool can make the model learn faster even on lower-traffic sites — but it also means an advertiser’s data is processed in the same model alongside data from other advertisers on Criteo’s network.
Criteo’s Pros and Cons
| Pro | Con |
|---|---|
| A broad publisher network adds visibility beyond Google/Meta | Campaign control is more limited than in Google Ads or Meta Ads |
| Ads update automatically when the product feed updates, no manual work | Less suited to service businesses without a product catalog |
| The machine-learning model takes bid optimization off the advertiser’s plate | The bidding logic is a black box — why a given user got a given bid isn’t visible |
| Setup is relatively fast for e-commerce (feed + tag) | The expected minimum spend can run high for small businesses |
How Does the Pricing Work?
Criteo doesn’t publish a public price list; billing is typically cost-per-click (CPC) and runs on a budget commitment negotiated with the advertiser. The three variables that determine final cost are the traffic volume the campaign targets, the size of the product catalog, and the level of managed service requested. This structure places Criteo closer to a tool that mid-to-large e-commerce businesses set up through a sales conversation than a self-service option for small budgets.
Who Does Criteo Fit?
Criteo produces meaningful results for an e-commerce site with a product catalog and consistently high-volume cart-abandonment traffic — the larger the catalog and the higher the traffic, the faster the bidding model learns. For service businesses without a product catalog, low-traffic new sites, or advertisers who want to keep the campaign mechanics entirely under their own control, Google Ads’ own dynamic remarketing feature is usually a better starting point.
The other tools in this series solve a similar problem with different mechanics: RTB House uses a deep-learning bidding model, AdRoll takes a multi-channel approach that includes email, and Taboola and Outbrain run native content discovery instead of retargeting.