The Ecomm Analyst

Mostly not, in the sense operators usually mean. Retargeting ROAS is real in that the orders exist and the revenue is genuine. It is misleading in that most of those orders would have happened without the ad. You are paying to show a banner to somebody who already had your product in their cart, and then counting their purchase as something the banner produced.

Retargeting almost always posts the highest ROAS in the account. It is also the line item I trust least, and those two facts are directly connected.

Why the number is structurally inflated

Retargeting audiences are built from people who already visited your site, viewed a product, or added to cart. That is a group selected for intent. They were further along than a cold audience before a single retargeting dollar was spent.

So the campaign inherits a conversion rate it did not create. Then last click attribution hands it credit for the purchase, because a retargeting ad is very often the final touch before someone comes back and buys. High intent audience plus last touch position equals a spectacular ROAS number that reflects audience selection rather than persuasion.

View through makes it worse again. If the campaign claims credit for impressions that were never clicked, it will claim a large share of people who were coming back anyway. Serving an impression to someone who has your product in their cart and then billing yourself for their return is close to a machine for manufacturing good looking ROAS.

The test that settles it

No attribution model will resolve this, because every model is dividing observed credit and the question here is causal. You need a holdout.

The cheapest honest version is a geo split. Pick a set of regions that make up a meaningful share of orders, turn retargeting off there, leave everything else running, and compare total regional revenue against a matched control set over three or four weeks. You are not looking at the retargeting ROAS during the test. You are looking at whether total revenue in the test regions fell.

I walked through the mechanics of setting one of these up in running a cheap geo holdout to sanity check a channel. It takes an afternoon to configure and it will teach you more than a quarter of dashboard analysis.

What usually happens is that revenue in the holdout regions drops by far less than retargeting claimed to be generating. I have seen accounts where retargeting reported an 8 ROAS and a holdout suggested the incremental contribution was a small fraction of that. I have also seen retargeting hold up better than expected on brands with long consideration cycles and expensive products. The result varies. What does not vary is that platform reported retargeting ROAS is an upper bound rather than an estimate.

This does not mean turn retargeting off

Some of it is doing work. Cart abandoners with a real reason to hesitate, longer consideration categories, and anything where a customer left to comparison shop are all cases where a reminder genuinely changes an outcome.

The practical move is to shrink the audience and the window rather than kill the campaign. Retargeting people who visited in the last three days is defensible. Retargeting everyone who touched the site in the last 180 days mostly means paying to advertise to your existing customers and to people who were coming back regardless. Cap frequency while you are in there. Nobody was persuaded by the fourteenth impression.

How I keep an eye on it

Two numbers. First, retargeting spend as a share of total ad spend. If it drifts above roughly fifteen percent without a deliberate reason, something has gone wrong with how budget is being allocated toward whatever looks best on a dashboard. Second, new customer share of orders. If retargeting ROAS is climbing while the proportion of first time buyers falls, the account is getting more efficient at selling to people who already knew about you, which is not growth.

I use ThoughtMetric, which sponsors this blog, to see both of those in one place, mainly because it separates new from returning customers at the order level rather than making me infer it from platform reports. Triple Whale and Polar Analytics will surface similar cuts. Whatever you use, the requirement is that it reads orders from the store rather than accepting the campaign’s own account of what it accomplished.

Common questions

Why is retargeting ROAS always so high?

The audience is selected for purchase intent and the ads sit close to the moment of purchase, so last touch credit lands there. The number reflects who is in the audience more than what the ads did.

What is a realistic incremental ROAS for retargeting?

It is brand specific and the only way to know yours is to test. Expect it to be well below the reported figure, and expect the gap to widen the longer your retargeting window is.

Should I exclude existing customers from retargeting?

From acquisition focused retargeting, generally yes, since paying to reacquire customers you can reach through email or SMS is expensive. Retention campaigns are a separate budget with separate goals.