Blogs / Is Retargeting Making Sales or Taking Credit for Them?

Is Retargeting Making Sales or Taking Credit for Them?

Nov 16, 20267 min read
Pulkit Khurana

Founder, SproutMe

A line drawing of a glass bell jar, illustrating how retargeting incrementality separates sales an ad caused from sales it merely claimed.

Your retargeting campaign posts the best ROAS in the account, so every budget review ends with a request to scale it. Yet everyone it reaches had already browsed your site or filled a basket, and the platform claims their purchase whether the ad changed anything or not.

Retargeting does make some sales, but its reported results overstate them almost by design, because it targets people who were already on their way to buying. How much of its credit is real varies widely by account, audience and attribution window. The only reliable way to find your number is a holdout test: switch retargeting off for a comparable group and compare total sales.

Why does retargeting look so profitable?

Retargeting audiences are selected for one property: they have already shown intent. A product-page visitor or a cart abandoner is far more likely to buy than a cold prospect, with or without an ad. A high conversion rate among people who saw the ad therefore tells you about the audience, not about the ad. That is correlation doing an impression of causation.

Attribution then compounds the problem. Last-click credits whatever sits at the end of the journey, and retargeting is built to sit there. Take a buyer who found you through organic search, read reviews and compared prices. They get one final nudge, and the whole sale lands on the nudge.

View-through windows add more credit. A banner that was served and scrolled past still claims the order if the purchase falls inside the window. Each platform also counts the same order as its own, so adding up their reports can show more sales than you made.

None of this makes retargeting worthless. It means the reported figure answers "which converting customers saw this ad?" rather than "which customers converted because of it?" The gap between those two answers is what retargeting incrementality measures, and it is what you are actually paying for.

This is one instance of the wider problem of working out which channel is actually bringing you customers. It is usually the instance where the gap is widest, because the audience is chosen on the very behavior that predicts a sale.

How much of the credit is real?

There is no reliable benchmark, and you should be wary of anyone who offers one. Published retargeting holdout results are mostly single-account anecdotes from measurement vendors and agencies, with no method, sample or date attached. They point the same way: a large share of reported retargeting conversions would have happened anyway. None of them gives you a figure you could plan a budget against.

What does hold up is the structure of the problem. Platform-reported conversions and caused conversions are different numbers, and the gap between them varies by channel, campaign type and audience. That variation is why a blanket discount, such as "assume retargeting is worth half what Meta says," is as arbitrary as trusting the platform outright. You would be swapping one unmeasured number for another.

A useful rule of thumb follows from the mechanism. The closer the audience sits to the purchase, the higher the reported return and the smaller the share of it that is incremental.

  • Close to purchase. A campaign chasing people who abandoned checkout an hour ago will post spectacular ROAS and probably caused little of it.
  • Further away. A campaign reaching people who read one blog post three weeks ago will look mediocre in the dashboard, but it may be doing more real work.

When does retargeting actually create sales?

Retargeting earns its keep when there is a real chance the buyer would otherwise drift away. It adds little when the purchase is already in motion. These conditions tend to tip it toward creating demand:

  • Considered purchases. When decisions take weeks and involve comparison, a reminder can bring back someone who would otherwise have bought elsewhere.
  • Cooler segments. Visitors who bounced from a single page are closer to prospects than buyers, so an ad has more to change.
  • Disciplined exclusions. Excluding recent purchasers and people mid-checkout stops the campaign from claiming orders that were already happening.
  • Tight windows. A short click window and little or no view-through credit do not make the ads more effective. They do strip out passive claims, so the reported number sits closer to the real one.

The reverse conditions predict capture:

  • impulse-priced products
  • warm audiences pulled from checkout pages
  • long view-through windows
  • high frequency against a small pool

If your setup looks more like this second list, expect the holdout to disappoint.

How do you prove which one yours is?

Run a holdout. Withhold retargeting from a comparable group, keep it running for everyone else, and compare total sales between the two. The difference is what retargeting caused. Everything else it reported was going to happen anyway.

Two designs do the job:

  • Platform conversion lift study. Meta, Google and TikTok all offer one. It randomly splits eligible users into a group that can see your ads and a control group that cannot. It is the cleanest design within one platform. However, the platform runs it on data you cannot audit, and it cannot see whether a lost sale came back through another channel.
  • Geographic holdout. You pause retargeting in regions matched for size and buyer behavior, usually for two to four weeks. Then you compare conversions using your own order system. It is noisier, but it measures total sales, which is the number your finance team cares about.

Read the result plainly. If sales in the paused group barely dip, retargeting was mostly capturing existing demand. If they drop sharply, it was creating it. Matching regions, sizing the test and reading noisy results is a subject of its own, covered in how to test what a marketing channel really adds.

Before you commit to a test, a cheap check tells you whether one is urgent. Add up the conversions each platform claims and compare the total with the orders you actually received. If the claims clearly exceed reality, something is being double-counted, and retargeting is the usual suspect.

SproutMe Companion answers that question in one place. It reads spend and conversions from Google Ads and Meta Ads alongside orders from Shopify.

What should you change after the test?

If the holdout shows mostly capture, shrink retargeting rather than killing it. Keep the segments that showed lift and exclude recent buyers. Then move the freed budget toward prospecting, which feeds the pool that retargeting later harvests.

If it shows real creation, keep it. From then on, plan with the measured lift rather than the platform's ROAS, so the next budget conversation starts from a number you can defend.

Either way, test again. Incrementality moves when your prospecting mix, pricing or seasonality changes, so a result from last year is a hypothesis, not a fact.

Conclusion

Retargeting sits at the point in the journey where credit is easiest to claim and hardest to earn. That is why its reported numbers are the least trustworthy in most accounts. It can still create sales, especially on considered purchases and cooler segments, but you cannot tell from the dashboard. A holdout is the only thing that separates the two, and the result should set your budget instead of the platform's ROAS.

Once the lift result is in, SproutMe Plan models how the freed budget would perform across Google, Meta, TikTok and LinkedIn before anything is committed.

Frequently Asked Questions

Rarely. Even when a holdout shows that most reported conversions would have happened anyway, some segments usually show genuine lift. Those tend to be cooler visitors or buyers in long consideration cycles. Cut the segments that capture existing demand, exclude recent purchasers, and keep the slice that changes behavior. Switching everything off throws away that slice along with the waste.

Not on their own. Shapley and Markov models spread credit across touchpoints by estimating what happens when one touchpoint is removed. They still work from observed journeys and an attribution window you choose, so a different window gives a different answer. They also cannot see offline influences like word of mouth. Treat their output as one reasonable perspective, not proof of causation.

A mix model estimates each channel's contribution from aggregate spend and sales over time, and open-source options exist, such as Meta's Robyn and Google's Meridian. On its own, a model can mistake correlation for effect, especially for a channel that tracks demand as closely as retargeting does. Calibrate it with holdout results so its estimates are anchored to measured lift.

Grow smarter with AI marketing tips

Join our newsletter to get practical insights, automation ideas, and performance tips straight to your inbox.

Get a complimentary audit to uncover AI opportunities hidden in your data.

Put these strategies to work