Criteo Review: What Star Ratings Cannot Tell You

What buyer reviews measure, and what they structurally cannot

Star ratings for advertising platforms are written by the people who log in. That makes them genuinely useful for a narrow set of questions: is the interface navigable, does reporting export cleanly, does support answer, how long did onboarding take. Those are real considerations and worth reading before a demo.

What no rating can carry is the question your finance director will ask: did the spend produce sales that would not have happened anyway. That is not a property of software. It is a property of an experiment, and it depends on your category, your margins, your creative and your competitors. A platform rated highly by three hundred users can still be the wrong media buy for you, and a platform with mixed reviews can be exactly right.

There is also a selection effect that applies to every vendor, including the ones we compete with and including us: the people invited to review are usually the ones the vendor is comfortable inviting. That is normal commercial behaviour rather than manipulation, but it means the distribution skews upward everywhere, and the text carries far more information than the average.

Read ratings for texture. Do not let them settle a budget decision.

And ratings are harder to read than they look

There is a second problem with review data in this category, and we know it from our own mistake rather than from theory.

We published a page on another vendor that recommended the product on the strength of an aggregate rating — a healthy score across a decent-looking number of reviews, pulled from a site that collects scores from several platforms. When we finally opened the platforms it claimed to aggregate, the numbers could not be reconciled with it in either direction. The rating was the argument of the page, so the page was deleted rather than corrected.

Three habits came out of that, and they apply to any platform review you read, including this one.

• A score without its count is not a fact. Always take both together, and treat a small count as a set of anecdotes rather than a rate.
• Open the platform that issues the number, not the site that repeats it. Aggregators and search snippets have contradicted live pages badly enough, often enough, that we now treat a repeated rating as unverified.
• Date the review, then verify support on a reference call. Many reviews of large ad-tech platforms predate the current product, the current team and occasionally the current owner.

That last one is worth taking seriously with any commerce media platform, because the product surface has moved a long way since most of the reviews sitting in the search results were written.

The primary sources that beat any rating

Criteo happens to be unusually documentable, which is an advantage worth taking.

Its own site states the scale: 200+ retailers and premium media owners, 17,000+ global clients, 740M daily active users, 5B product SKUs, access from 60+ third-party demand-side platforms and 100k algorithm tests a year. Those are company claims rather than audited figures, but they are published, dated and attributable — a different class of evidence from an anonymous testimonial.

More useful still, Criteo writes its commercial architecture into its Commerce Max terms of service. That document names a Demand Side Platform Fee and, for managed campaigns, a Managed Service Fee, each expressed as a percentage of working media spend, plus retailer audience data charges and possible third-party technology fees. No rates appear — those live in your order form — but knowing the shape of the invoice before a call is worth more than a star rating, and very few competitors in this category document it at all.

Third, it is publicly listed and reports on a schedule. Two quarters of that tells you more about the health of a network you are about to depend on than any review site can. And its own advertiser pages now describe incrementality testing, so anyone telling you the platform cannot measure lift is describing an older product.

The general principle is worth more than the specific vendor: read what a company is contractually and legally obliged to say before you read what its customers volunteered.

A worked example: reading the takeover coverage properly

In July 2026 the strongest test of that principle arrived in the form of takeover reporting, and it is a good exercise in reading a primary record.

Bloomberg reported on 6 July 2026, and Reuters reported the same day citing two people familiar with the matter, that Vista Equity Partners and Quinti Capital had approached Criteo at a premium of more than 50% to recent trading. Nasdaq-listed shares closed 21.4% higher at $23.17 that day, giving a market value of roughly $1.16 billion. None of the parties confirmed terms, Criteo's board had not said how it would respond, and any completed transaction would require shareholder approval.

Two things to notice. First, that is reported, not settled — an approach covered by two wire services is a real fact about the world and not a decision by anyone. Second, secondary coverage has circulated a much larger equity figure that does not reconcile with the share count, which is exactly the failure mode described in the section above: a confident number, repeated onward, that nobody re-derived.

None of this is a reason to avoid a vendor, and we are not going to pretend it is. Ad tech consolidates constantly, and every firm on your shortlist carries some version of the same exposure. What it is, is a reason to negotiate three clauses in any multi-year agreement: the right to terminate without penalty on a change of control, a price-protection period that survives an ownership change, and a data-portability commitment that lets you export your campaign history in a usable format. Ask all three of every vendor, including us. The speed of the answer is itself information.

What the 2026 record shows about direction

A platform review dated last year describes a product that has since moved, so read the announcements rather than the retrospectives.

On the network side, Criteo published an expanded partnership with Loblaw Advance in August 2026, an onsite retail media partnership with Lidl in May 2026, and a South African expansion with Massmart in July 2026 — three retail networks added or deepened in four months. On the product side, it published two updates on its integration with OpenAI, in May and June 2026, making it one of the first commerce media businesses to describe advertising inside an AI assistant publicly. In June 2026 it was named a leader in an analyst matrix for retail media network and monetisation platforms, and in July 2026 it completed a redomiciliation from France to Luxembourg.

That is a company investing in network breadth and in new surfaces rather than defending a single one. It also points to the piece of diligence that matters most here, and it is a neutral one: retail media networks are commercial relationships that get renewed, expanded or ended. Before you sign, confirm that the specific retailers you care about, in the countries you sell in, are live today — and ask what notice you would receive if one changed.

Five checks worth more than the ratings

Run these before you commit budget. They work for any commerce media platform, not only this one.

• Named retailer coverage. List your top five retail relationships by revenue and ask for written confirmation that each is live, in your markets, with the ad formats you want.
• Fee denominator. Ask what the percentage is charged on, and whether the budget you state is inclusive of fees. The same rate on two different denominators is two different prices, and this is where most quote comparisons go wrong.
• Prospecting against retargeting, separately. Ask what share of a comparable account's spend reached shoppers who have never seen the brand, and what those two segments returned on their own. A blended figure hides the answer, and it hides it in the direction that flatters the vendor.
• Exit terms. Read the termination clause before the pricing. Criteo's own Commerce Max terms describe termination without penalty on short notice, which is more flexible than several competitors offer and is a genuine point in its favour.
• Proof design. Ask how they would demonstrate incremental sales rather than attributed sales, who would build the holdout, and whether the design goes into the scope of work before spend starts.

A vendor who answers four of the five crisply has taught you more in an hour than a week of reading reviews would.

The question that survives every platform review

Whichever platform wins, someone still has to run it, and the result still has to be tested.

reMKTR buys Amazon DSP — the demand-side platform, not the Delivery Service Partner courier programme — as a managed service on our own seats. We are part of Full Circle, a full-service Amazon management company with $500M+ in managed spend across 100+ brands, and we hold 109 live Amazon DSP advertiser seats. Across 30 advertisers in July 2026 that book delivered 6.04x on ad spend at a $5.49 cost per acquisition across 57,137 attributed purchases, on 78.4 million impressions, with 20.1% of the purchases coming from shoppers new to the brand.

We publish the method with the numbers because the method is the argument. Last-click attribution cannot prove incrementality and never could; holdouts and matched controls can. We reconcile in Amazon Marketing Cloud, where display and sponsored ads stop double-counting each other.

Here is the sort of thing that reconciliation surfaces, and it is why a platform's own report is a weak review of itself. On one account — a publishing brand — organic order share went from 13.5% before display started to 27.9% afterwards, and was still climbing four weeks after every dollar of display spend had stopped, while sponsored-ads cost of sale on the same account compressed from 48% to 37% and held. One account, one window, told as a story rather than a rule. But it is the strongest argument we have that display should be judged on the whole account rather than on its own attributed column, and no star rating on any platform will ever contain that.

One last diligence question that has nothing to do with software. The most common reason a client loses confidence in a vendor is not bad results — it is a false impression of inactivity while the work happens out of sight. That is what we observe across our own client base. So ask any platform or agency how the invisible weeks get reported, and whether the bad ones get reported first.

Two honest signposts. Dr. DSP when you would rather run the Amazon display capability yourself as a product than hire a team for it. Full Circle when the problem is broader than media — listings, catalogue or availability — and no platform choice reaches it.

Who Criteo genuinely suits

Brands selling through many retailers, in many countries, who need shopper audiences bought across those networks under one contract. That is a real capability with real scale behind it, and an Amazon specialist is not a substitute for it. If that describes you, the rest of this page is interesting rather than decision-relevant.

Retailers and marketplaces monetising their own inventory are a different customer again, buying a different product from the same company, and we do not compete there at all.

Where it fits poorly is the narrow case we are built for: a brand whose revenue is overwhelmingly one marketplace, whose real question is whether display added incremental orders on that marketplace. That question gets answered against that marketplace's own purchase data, in a clean room, with something held out — and by someone who is not paid on the answer.