Criteo alternative — you are shopping for one of three different things

Name the job in one sentence before you name a vendor

A clarification worth making up front: every DSP mentioned below is a demand-side platform, meaning ad-buying technology. Amazon separately runs a Delivery Service Partner programme that franchises parcel delivery, and it has nothing to do with any of this.

Criteo's own platform page lists five distinct products aimed at five distinct buyers — Commerce Growth, Commerce Max, Criteo GO, Commerce Yield and Commerce Grid, read on 20 August 2026. So when someone says they want an alternative to Criteo, the first useful question is which of those they ran, and the second is what the campaign was for.

In practice that collapses to three jobs:

• Job one — multi-retailer reach. You need to be present onsite and offsite across many retail banners, in one contract and one creative pipeline.
• Job two — open-web display. You want programmatic reach across the internet at large, with control over supply paths and access to log-level data.
• Job three — depth where the revenue is. One channel produces most of your sales and you want display bought against that channel's own shopper signal, and proven there.

Most shortlists go wrong by mixing candidates from all three. This page is from Dr. DSP, part of Full Circle, a full-service Amazon management company with more than $500M in managed revenue across 100+ brands. We compete only for job three.

Job one — reach across many retailers, onsite and offsite

This is the job Criteo was built for and the hardest to replace, because the alternative to an aggregator is several relationships instead of one.

Buy the retail media networks directly. Most large retailers now run their own self-service ad platform. Direct access generally means better data granularity and no intermediary margin, and it costs you an integration, a login and a reporting reconciliation per network. The rough arithmetic: below about four networks, direct is usually cheaper and clearer. Above four, the operational load starts to exceed whatever an aggregator's margin was.

Buy an aggregating platform. Skai and Pacvue both sit here, managing many retail media networks from one interface. Skai is unusual in this category for publishing its rate card: on skai.io on 20 August 2026 the tiers ran from Standard at $114,000 per year on annual billing, covering up to $4M of annual ad spend, through Advanced at $276,000 up to $10M, Enterprise at $504,000 up to $20M, and Enterprise Premier at $756,000 up to $35M, with incrementality testing listed only at that Enterprise Premier tier. Pacvue quotes on a call. Quartile also belongs here and prices in tiers keyed to spend rather than as a percentage rate — worth stating plainly because the two get conflated constantly.

Buy an agency. A retail media agency will hold the network relationships for you and charge for the team. Genuinely the right answer when you have budget but no headcount.

Job two — open-web display without the commerce network

If what you actually valued in Criteo was programmatic reach rather than the retailer footprint, you are shopping in a different aisle and the commerce-media specialists are the wrong list.

The Trade Desk is the reference point here, and it is genuinely excellent at what it does: connected TV, open-internet scale, supply-path transparency and log-level data access. It suits brands and agencies with a trading desk, or an agency partner running one. It publishes no rate card, and independent write-ups circulate spend minimums that the company itself does not state — treat any figure you read as unverified until it arrives in a proposal.

Other general DSPs deserve a look for specific strengths rather than as like-for-like swaps — audience modelling, creative tooling, a particular regional supply footprint. Amazon DSP also buys well beyond Amazon's own properties, which surprises people who assume it is confined to the retail site.

The honest warning about this aisle is retraining. A dynamic product campaign that has been optimising for a year carries learned state you cannot export, and several of the better guides in this results page put the reset period at six to eight weeks of degraded performance. Budget for that as a cost of switching, not as evidence about the new platform.

Job three — depth where most of the revenue actually lands

If Amazon produces the majority of your sales, the deciding factor stops being reach and becomes signal: what a shopper searched, viewed, bought and abandoned on Amazon lives inside Amazon, and so does the only place display and sponsored search can be reconciled against each other.

Your honest options here, and two of them are not us:

• Amazon DSP, self-service. Amazon's own product page, read on 20 August 2026, states no minimum for the self-service route. You need a trader and a weekly cadence. Cheapest in fees, most expensive in attention.
• Amazon DSP, Amazon-managed. The same page states that the managed-service option typically requires a minimum investment of $50,000, and that the minimum can vary by country. You get the platform owner in the seat.
• A specialist Amazon DSP agency. A named team, a scope of work, usually a percentage of media or a retainer. reMKTR is our own answer to this shape, and it buys across more than Amazon.
• A commerce suite. Pacvue, Quartile and Intentwise all touch this. Rithum — the company formerly known as ChannelAdvisor, taken private by CommerceHub in 2022 and rebranded in 2023 — is the enterprise end of the same idea, and its historic filings describe a fee architecture of a minimum commitment plus a percentage above it. That is dated architecture: get the current structure in writing.
• DSP as a managed product. Dr. DSP. No published price — a demo, the first 30 days free, then priced on the call against real media budget and scope, month to month.

As a benchmark rather than a promise: across 30 advertisers in July 2026, an API pull showed 6.04x return on ad spend, 78.4 million impressions at a $4.00 CPM, $1.42 blended cost per click and 20.1% new-to-brand.

Export this before you give anyone notice

Portability is decided by what you take with you, and almost none of it is retrievable after the account is closed. Do this while the contract is live and you are still a customer in good standing.

• Raw delivery data at the finest grain available — impressions, clicks, spend by placement, by retailer, by creative, by day, for the full history. Not the dashboard summary. The export.
• Audience definitions in words, not audience IDs. The IDs will not travel; the logic behind them will, and rebuilding logic from memory is where switches quietly lose their targeting.
• Creative source files, including feed templates and dynamic rules, at full resolution with fonts.
• The product feed specification your current platform consumes, and the mapping between your catalogue fields and theirs.
• Every measurement study ever run — brand lift, geo tests, holdouts — with methodology attached, not just the headline result.
• Invoices for the last twelve months — the only document that reliably shows what you paid versus what was quoted.
• The contract itself, and specifically the notice period, the change-of-control clause, any auto-renewal date and any data-deletion trigger.
• Pixel and tag inventory — every tag deployed on your site by the outgoing vendor, so removal is clean and nothing silently keeps firing.

Two asks worth making of every vendor on the shortlist, including us: written notice of any fee change, and a data-portability commitment stating what you can export and in what format. Vendors who agree easily are telling you something useful.

Keeping a clean read across the switchover

The most expensive mistake in a platform migration is not choosing wrong. It is being unable to tell afterwards whether the number moved because the platform changed or because something else did — a price rise, a stockout, a seasonal peak, a competitor's launch.

Five rules make the read survivable:

• Freeze everything else you control for the switchover window. No price changes, no new ASINs, no promotional calendar shifts, no creative refresh. If you cannot freeze them, log them with dates so they can be excluded later.
• Write down the baseline before you switch — the specific metric, the specific date range, the specific denominator. Baselines chosen after the fact are always flattering to somebody.
• Do not overlap unless you can separate the audiences. Running both platforms at once feels prudent and usually produces two reports each claiming the same conversions.
• Discount the learning period explicitly. Agree in advance how many weeks do not count, and hold that agreement when the first fortnight looks bad.
• Hold out something. Withhold display from a matched set of ASINs or regions for a defined window, then reconcile in Amazon Marketing Cloud, where DSP and sponsored ads stop double-counting each other's orders. Amazon makes it available free to eligible advertisers. This is the only method here that answers whether display was worth buying from anyone.

That last point is why we exist rather than a feature list. Last-click reporting cannot separate an ad that caused a sale from one that merely appeared near it, and display is where that gap is widest. If the problem turns out to be broader than media — content, catalogue, fees, inventory — that is a Full Circle conversation rather than a DSP purchase.