Criteo advertising is a catalog-based dynamic retargeting network that shows visitors who left without buying the exact product they viewed, on publisher sites across the open internet. Unlike Meta and Google, which serve inside their own ecosystems, Criteo positions itself as a commerce media platform that stitches publisher inventory together with a retailer network. In practice that makes Criteo an extra lower-funnel reach layer rather than a replacement. Below we cover the three pillars of setup, the real differences from Meta and Google retargeting, sensible budget bands, and the biggest measurement trap.
What is Criteo and what does commerce media mean?
Criteo is a performance advertising platform that matches an online store's product data with publisher inventory. Commerce media is the category name for advertising driven by first-party data tied directly to purchase intent. Criteo collects that data from two sources: behavioural events on the advertiser's own site, and shopping signals from partner retailers. The company reports working with 200+ retailers and tens of thousands of publishers, reaching billions of shoppers.
- Commerce Growth: self-serve dynamic retargeting for small and mid-size advertisers, priced on cost per click, with no dedicated account manager required.
- Commerce Max: onsite and offsite retail media buying across the retailer network for brands and agencies, with SKU-level reporting.
- Product catalog: the feed carrying price, stock, image and ID fields. It is the single source that decides which product appears inside the ad.
- OneTag: the JavaScript tag that carries site events to Criteo, covering product views, listings, baskets and purchases.
- Identity layer: in a cookieless context, matching relies on a shopper graph built on hashed emails and alternative IDs.
How does Criteo dynamic retargeting actually work?
Dynamic retargeting pulls the product a user viewed straight from the catalog and assembles the banner at serve time. The chain runs in four steps: site event, catalog match, auction bid, and conversion measurement. The critical point is that ad quality depends less on creative than on feed accuracy. A product that is out of stock or carries a stale price turns into wasted clicks even under the best bidding algorithm.
- Load OneTag on every page and fire viewItem on product detail, viewList on listing and category pages, viewBasket in cart, and trackTransaction on the confirmation page.
- Connect the product catalog. The ID field in the catalog must match the product ID you send in site events exactly, otherwise match rate drops and the dynamic ad falls back to a generic banner.
- Add order ID and item price to basket and purchase events. Revenue reporting and bid optimisation depend on these fields; if they are missing, ROAS is computed on bad inputs.
- Split campaigns by recency: the last 3 days, days 4-14 and days 15-30 deserve different bids and different messaging.
- Exclude existing customers and anyone who purchased in the last 7 days, otherwise part of the budget goes to users who already converted.
How is Criteo advertising different from Meta and Google?
All three serve dynamic product ads, but their inventory and data sources differ. Meta serves inside its own app family and is strong in social context. Google dynamic remarketing covers the Display network plus YouTube and Gmail surfaces. Criteo shows up mainly on publisher sites across the open internet and inside partner retailers. In other words Criteo usually opens an impression pool the other two do not reach, which makes the relationship complementary rather than competitive.
See Criteo, Meta and Google in one table
Ads Sensor reads spend and revenue from your connected accounts side by side and makes double counting visible.
How much budget should Criteo get, and which metrics matter?
Industry benchmarks show most advertisers allocating 10-20% of total paid media budget to retargeting, rising to 25-30% for high-traffic ecommerce sites. On the Criteo side four metrics deserve attention: catalog match rate, CPC, cost per conversion and incremental ROAS. The first three are visible in the dashboard. The fourth only appears through a controlled holdout test, and it is the one that should drive decisions.
Why does Criteo ROAS look higher than it really is?
Retargeting shows ads to people already close to buying, so reported ROAS is always flattering. What it does not tell you is how much of that revenue the ads actually caused. Studies suggest up to 75% of retargeting conversions would have happened anyway, with true incremental lift typically settling in the 25-30% range. That is not a reason to switch the channel off. It is a reason to set an honest budget ceiling.
- Holdout test: keep 10% of the audience out of the ads for 3-4 weeks and measure the conversion gap. This is the only honest measure of incrementality.
- Frequency cap: beyond roughly 3-5 impressions per user per day, you typically buy irritation and waste rather than extra conversions.
- Recency window: conversion probability drops sharply past 30 days, so concentrate spend in the first 14.
- Double counting: Meta, Google and Criteo each report the same conversion. For decisions use blended MER, total revenue divided by total ad spend.
- Margin check: look past ROAS to the profit left after returns and shipping. Scaling on revenue alone grows turnover and shrinks profit.
In a retargeting channel the real question is not 'what is our ROAS?' but 'would this revenue have arrived without the ad?'Industry benchmarks and incrementality studies, 2026
What are the most common Criteo setup mistakes?
In practice most problems sit in the data layer, not in campaign settings. When the catalog product ID and the site event ID disagree, Criteo cannot resolve the right item. The second frequent failure is consent ordering: if the tag fires before consent is handled correctly, data never arrives and the account reads as 'traffic yes, matches no'. The third is a feed that is not refreshed during the day, so out-of-stock items keep serving.
- Compare catalog IDs with site event IDs. If match rate is low, fix the feed first and leave bids alone.
- Refresh the feed at least daily, and more often during promotions; serving out-of-stock products burns clicks with zero upside.
- Audit consent handling so the tag fires after permission is granted. Otherwise data loss compounds quickly in European markets.
- Do not skip viewList on internal search and category pages; a meaningful share of intent signal originates there.
- Avoid aggressive scaling in the first 30 days. Large, frequent daily budget swings distort the signal while the algorithm is still collecting data.
How do you read Criteo alongside Meta and Google in one dashboard?
Judging Criteo in isolation is misleading, because at the bottom of the funnel it shares the same users with your other channels. The correct reading places every channel's spend and revenue in the same date range and compares them against total site revenue. Ads Sensor builds that combined view from Meta Ads, Google Ads, TikTok Ads, Criteo and GA4 connections, and its AI analysis reads the whole account rather than one channel at a time. Connecting accounts starts at the pre-beta signup.
- Combined revenue view: compare Criteo, Meta and Google spend against GA4 revenue in one date range. If channel reports sum to more than booked revenue, you have double counting.
- Reasoned actions: Ads Sensor prioritises risks and opportunities and states why for each one, then applies approved changes through the platform APIs.
- Before/after tracking: applied recommendations are measured automatically, so the effect of a change is not left to guesswork.
- Anomaly monitoring: spend spikes and conversion drops are watched around the clock, which catches feed-driven breakage early.
- Feed discipline: catalog quality drives Shopping performance as much as Criteo. For the detail, see the product feed optimization guide.
In short: with a clean product catalog, Criteo advertising opens lower-funnel inventory that Meta and Google do not reach. But treating its reported ROAS as real contribution makes the channel look bigger than it is. Turn it on, keep the feed tight, measure incrementality with a holdout, and decide budget from the combined table. For a complementary read on the feed side, see the Google Shopping ads guide.