Onsite vs Offsite Retail Media: Which Should You Fund First?

Every retail media conversation eventually splits into two halves. Onsite is the advertising you sell against your own site, app, and store. Offsite is the advertising you buy elsewhere using your shopper data. They look like two products in the same catalogue. They are actually two different businesses with different economics, different buyers, and different failure modes.

Get the sequencing wrong and you either leave margin on the table or you scale a low-margin reseller business before you have proven you can measure anything.

The economics, honestly

Onsite is close to pure margin. You already own the page. The incremental cost of serving a sponsored listing is effectively an ad server call. Whatever the brand pays, minus platform licence fees and the sales team, drops through. Typical contribution margins sit somewhere north of 70 percent, often much higher.

Offsite is a media reseller business. You take the brand’s money, buy inventory on Meta, YouTube, connected TV, or the open exchange, and keep a spread. The spread is real but thin: after media cost, platform fees, and data activation costs, you are often working with 20 to 35 percent. You are also carrying the operational risk of running campaigns on someone else’s platform with someone else’s delivery rules.

Onsite makes money. Offsite makes scale. That is the trade in one line.

The ceiling problem

If onsite is so profitable, why does anyone bother with offsite?

Because onsite has a hard ceiling and it arrives faster than most teams expect. Your onsite revenue is bounded by three multiplied numbers: sessions, ad slots per session, and price per slot. You can grow sessions slowly. You can raise price until demand thins. But ad slots per session is where teams get greedy, and it is the one lever that quietly destroys the underlying business.

Add a fourth sponsored result to a ten-result page and revenue goes up this quarter. Conversion rate goes down. Shopper trust goes down. Organic discovery of your own private label goes down. Six quarters later you have a monetised site that converts worse than the competition and you cannot work out why.

So the sequence is not “onsite forever”. It is: max out onsite within a load ceiling you set deliberately, then use offsite to sell beyond it.

A practical decision framework

Run your situation through these five questions before allocating a euro.

1. Is your onsite auction actually competitive?

Onsite pricing only works if there are enough bidders per keyword and category. If your top-100 search terms have fewer than three active advertisers each, your auction is not clearing at market price, and you are selling premium inventory at floor. Fix auction density before you build anything new. Density comes from supplier onboarding and self-serve access, not from more inventory.

2. Do you have the identity graph offsite requires?

Offsite requires matching your shopper segments to a platform’s users at an acceptable match rate. Below roughly 40 to 50 percent match, your addressable audience shrinks so much that the campaign either underdelivers or the effective CPM becomes indefensible. Test the match rate with your two largest offsite partners before you sell a single offsite package.

3. What is the buyer actually asking for?

Onsite buyers want conversion, share of search, and defence of their branded terms. Offsite buyers want reach, new-to-brand shoppers, and upper-funnel coverage that onsite cannot deliver because onsite only reaches people already shopping. If your suppliers are asking “how do I find shoppers who are not already in the category”, onsite cannot answer that, no matter how you price it.

4. Can you measure it the same way in both places?

This is the one that trips up most networks. Onsite measurement is closed loop by default: exposure and transaction live in the same system. Offsite measurement requires an exposure log coming back from a platform, joined to your transaction data, usually inside a clean room, with all the match-rate and lag problems that implies. If your offsite measurement is materially weaker than your onsite measurement, brands will spot it, and they will discount their offsite renewal accordingly. Whatever you do, do not paper over the gap with attributed ROAS. Read how to measure retail media incrementality and hold both channels to the same standard.

5. Do you have the operating capacity?

Offsite adds campaign management, platform certifications, creative specs, and delivery troubleshooting to a team that may currently just be approving keywords. Every network that launched offsite understaffed spent the following year apologising for delivery.

The sequencing that works

For most retailers, the order looks like this:

Phase 1 (months 0 to 9): onsite search only. Sponsored listings in search and category results. Self-serve platform for the long tail, managed service for the top 30 suppliers. Publish a rate card. Establish an ad-load ceiling and defend it against your own sales team.

Phase 2 (months 6 to 18): onsite display. Homepage, category banners, brand stores. This unlocks brand budget rather than only performance budget, and it gives you something to sell when search is sold out in a category.

Phase 3 (months 12 to 24): offsite, starting narrow. Pick one or two platforms. Run a proper measurement design from day one, ideally a geo or audience holdout rather than platform-reported conversions. Prove the incremental case in one category before you productise it.

Phase 4 (month 18 onward): in-store and full-funnel packaging. In-store digital is the hardest to measure and the most capital-intensive. It belongs after you have measurement credibility, not before.

The overlaps are deliberate. These are not gates; they are staggered starts.

The two failure modes

Failure mode A: onsite maximalism. The network hits its number every quarter by adding ad slots and raising floors. Nobody measures site conversion rate as a counterweight. By year three, the core retail business is subsidising the media business without anyone writing it down. The fix is governance: an ad-load ceiling and a conversion-rate guardrail owned by someone who does not carry the media revenue target.

Failure mode B: premature offsite. The network launches offsite to hit a revenue number, discovers the margin is thin and the measurement is contested, and burns supplier trust on a product it could not support. The fix is patience and a narrower launch.

Both failures are strategy failures rather than execution failures. They come from taking on more surface area than the operating model can carry, which is the same disease described in deciding what not to do.

A quick sanity check

If you want one number to orient around: for a retailer with a healthy onsite business, offsite typically represents 15 to 30 percent of network revenue at maturity, but a much smaller share of network profit. If offsite is a majority of your revenue and you have not deliberately chosen to be a data-activation business, something has drifted.

Onsite pays the bills. Offsite raises the ceiling. Fund them in that order, measure them to the same standard, and never let the revenue target set your ad load.

The pan-European definitions of onsite, offsite and in-store, the three buying models, and the measurement gap that makes one network impossible to compare with the next, sit in The Third Wave of Digital Marketing. If you are still at the definition stage, start with what a retail media network actually is.