Comparing retail media business models on one retailer-advertiser exchange

Compare retail media business models on one job: an England retailer makes a defined placement available to one advertiser, approves the communication, records delivery and supplies an agreed report. This comparison judges named models, among them Tesco Media and Insight Platform, Sainsbury's Nectar360, Boots Media Group, Amazon Ads and Criteo, on the same units, evidence duties and exit conditions.

What to take away

Fields every model must fill

For every model, identify the legal contracting parties, the inventory owner, the advertiser and the approving editor. Then record the data roles, billable event, currency and VAT basis. Fix the reporting population, corrections process and exit route.

Test the commercial unit against the contract. What determines the price is the contracting model: a CPC auction clears at market, a booked placement is negotiated, and a managed network bundles media, labour and technology. Ask for a dated rate card and separate units.

Prices and thresholds remain UNKNOWN until a dated contract supplies them.

Amazon Ads sells Sponsored Products through a self-serve console on a cost-per-click basis, with no insertion order and no negotiated floor. Tesco Media and Insight Platform, run with dunnhumby, sells booked placements against Clubcard, which Tesco says covers more than 20 million members.

A Criteo managed network bundles media, labour and technology. These are different commercial units, so no single rate card applies.

IAB UK's category definition (accessed 2025) recognises on-site, off-site and in-store environments. These settings must not be averaged: an on-site sponsored result, an off-site audience activation and a store screen have different delivery records and affected people.

Amazon's Sponsored Products sit inside Amazon's own search results. An off-site buy made through The Trade Desk or Meta Collaborative Ads reaches the shopper on other publishers. It uses retailer data the retailer agreed to share.

How the named models differ in practice

Amazon Ads runs a self-serve CPC auction with no negotiated floor. Tesco Media and Insight Platform sells booked placements against Clubcard data. Sainsbury's Nectar360 combines on-site media with off-site audience charges. Boots Media Group sells placements against Advantage Card data.

Criteo operates a managed network within retailer authority, bundling media, labour and technology. CitrusAd provides white-label technology the retailer runs. The Trade Desk and Meta Collaborative Ads use approved retailer-data routes off-site. NIQ and Circana sell measurement panels beside these standards. The practical difference is who holds the contract, what event triggers payment and who can export the evidence.

Keep claims and measurement bounded

IAB Europe's Commerce Media Measurement Standards V2.1 (accessed 2025) includes defined sales measures and incrementality methods. It is a European industry standard. It is not proof that a supplier implemented the method or that a campaign caused sales.

A delivery event, product view, order and captured payment are different states. So are a return and a retained sale. Each model should report the same eligible population and window before a like-for-like comparison is attempted. Advertiser attribution remains separate from a causal estimate. NIQ and Circana sell retail measurement panels beside these standards.

Keep regulatory gates outside the price score

The CAP Code (accessed 2025) applies to relevant non-broadcast marketing communications and places duties on advertisers, agencies and media. A managed supplier such as Criteo does not absorb every advertising decision just by carrying the placement.

Where online tracking is used, the ICO's storage and access guidance (accessed 2025) requires an examination of the actual technology, purpose and route under PECR and, where relevant, UK GDPR. A clean-room match through LiveRamp or a retailer's own platform still needs a lawful basis.

Data access and statistical usefulness are independent questions. The CMA's unfair commercial practices guidance (accessed 2025) provides the current consumer-law context for business-to-consumer practices. Product presentation and price gates cannot be offset by cheaper media.

Run one controlled comparison

Give each model the same fictional campaign record: one approved product, one creative, one placement window and a refused tracking choice. Add a correction, a return and a final export. Record which party can perform each action, what evidence it leaves and whether another reviewer can reproduce the result.

Worked example, illustrative only. A fictional campaign has a £12,000 media budget, one approved product, one creative and a four-week window. Eligible population: 1 million shoppers. Run it through two named models.

Amazon Ads Sponsored Products: assume a £0.60 CPC. The £12,000 budget buys 20,000 clicks. At a 2% conversion rate that is 400 orders. With 40 returns, retained sales are 360. At £40 average order value, retained revenue is £14,400. Cost per retained sale is £33.33.

Tesco Media and Insight Platform: assume a £15,000 booked placement fee. It delivers 600,000 impressions and 6,000 clicks. At a 5% conversion rate that is 300 orders. With 30 returns, retained sales are 270. Retained revenue at £40 average order value is £10,800. Cost per retained sale is £55.56.

These figures are illustrative, not rate cards. The models also differ on evidence fields, so the numbers do not rank them.

Production serving stays disabled. A model that cannot export its definitions is not comparable just because its dashboard has a total. Keep failures visible.

If a managed network bundles media, labour and technology, ask for separate units or mark the commercial field UNKNOWN. Do the same when an off-site platform cannot show which retailer permission supports its data use.

Choose only after every model has the same case, evidence fields, payable-event definition, internal-labour boundary and exit test. If a proposed model cannot fill them, retain the current route and record the gap.

Checklist before you act

Common questions

What should be compared across retail media business models?

The comparison covers operating structures on the same units, evidence duties and exit conditions. It does not rank models by market prevalence.

Which named models are compared on the same exchange?

The table names Tesco Media and Insight Platform, Sainsbury's Nectar360, Boots Media Group and Amazon Ads, among others. Criteo, CitrusAd, The Trade Desk and Meta Collaborative Ads cover the supplier-side and off-site routes. Each row fills the same fields: who does the work, the billable unit, the buyer evidence and the exit question.

Why can on-site, off-site and in-store placements not be averaged?

IAB UK's category definition recognises on-site, off-site and in-store environments. The article states these settings must not be averaged because an on-site sponsored result, an off-site audience activation and a store screen have different delivery records and affected people.

What should happen when a proposed model cannot fill the evidence fields?

If a model cannot fill the same evidence fields, retain the current route and record the gap. The full test is in the comparative summary above.