Retargeting is a tax you pay twice

Every retargeting dashboard I have ever opened for a new client shows the best ROAS in the account. That number is not a signal the channel works. It is a signal the audience was rigged from the start.

Why the number lies

Retargeting audiences are built from people who already visited your pricing page, already downloaded the whitepaper, or already sat through a demo. These are, by definition, the people closest to converting in your entire funnel. Showing them an ad and then crediting the ad when they convert is like taking credit for the sun rising because you happened to be standing outside at 6am with a flashlight on.

Last-touch attribution makes this worse. If a buyer's last ad exposure before signing is a retargeting banner, the whole deal gets logged against that banner, even if the actual work was three sales calls, a security review, and a champion who had already decided six weeks earlier. View-through attribution compounds it further, crediting an impression the buyer may not have consciously registered, purely because it loaded on a page they had open in another tab.

The mechanism, stated plainly

Retargeting does not create demand. It is shown exclusively to people who already have demand, which means its measured performance is really the base rate of that audience converting, plus whatever small marginal nudge the ad itself contributed. Most reporting cannot separate the two, so all of the credit goes to the ad. That gap between measured performance and actual incremental lift is the tax, and you pay it twice: once in the wasted media spend, and once in the bad decisions the inflated number talks you into making elsewhere in the budget.

The four jobs actually worth funding

Not all retargeting is the tax. A small number of use cases have a real causal mechanism: the ad is reaching someone at a moment when a nudge plausibly changes their behaviour, not simply riding along with a decision already made.

Notice what is missing: generic 30-, 60-, and 90-day website visitor retargeting, cart-abandonment-style sequences aimed at anyone who opened a pricing page, and always-on brand awareness retargeting dressed up as a lower-funnel tactic. Those are the campaigns eating most of the budget in a typical account, and they are exactly the campaigns with the weakest causal story.

Frequency caps and windows that stop the bleeding

Cap frequency across the stack, not per platform

Most platforms let you cap impressions within their own walled garden, and almost nobody caps across platforms. A single buying committee member can end up seeing your retargeting ad on LinkedIn, in a display network, and inside a native placement, ten or more times in a week, with three different tools each reporting the frequency as low because none of them can see the other two. Set a hard combined cap, even if it means tracking it manually in a spreadsheet once a month, at three to five impressions per person per week.

Shrink the audience window

A 90-day retargeting window includes people who solved their problem somewhere else in month two. Shrink the default window to 14 to 21 days for general site retargeting, and reserve longer windows only for the four funded jobs, where a longer relationship justifies it. Shorter windows also force the audience to stay closer to genuinely recent behaviour, which is the only part of a retargeting audience with any real signal in it.

Running a proper holdout test

1. 1

   Randomly split the audience

   Take the retargeting audience you want to test and randomly exclude 10 to 20 percent of it from any exposure. Use a random split, not a segment split, or you will bias the result.

2. 2

   Suppress the holdout everywhere

   Make sure the excluded group is suppressed across every platform running retargeting, not just the one you are auditing. A leak in one platform contaminates the whole test.

3. 3

   Run it four to six weeks minimum

   B2B sales cycles are long enough that a one-week test tells you nothing. Give the test enough time for the audience's normal conversion behaviour to play out.

4. 4

   Compare pipeline, not clicks

   Measure meetings booked and pipeline created in each group, not impressions or click-through rate. The only question that matters is whether exposure changed the outcome.

5. 5

   Fund what survives, cut what does not

   If the holdout converts within a few points of the exposed group, that audience is not earning its spend. Move the budget to one of the four funded jobs or somewhere else entirely.

What the vendor side will tell you

Ask any programmatic vendor or agency running your retargeting and they will tell you the audience is warm, the intent is real, and the ROAS speaks for itself. All three statements can be true and the spend can still be non-incremental, because warmth and intent describe the audience's pre-existing state, not the ad's contribution to it. A vendor paid on managed spend has no incentive to run a holdout that might shrink the account, so if you are not asking for one, nobody is going to volunteer it.

What to do with the freed budget

Cutting a 40,000 dollar retargeting line and not reallocating it is its own mistake. The point is not to spend less, it is to stop paying for credit that was never earned. Once the tax is gone, move the money toward channels and jobs with a real mechanism: competitive displacement campaigns aimed at accounts in an active evaluation with a named competitor, expansion campaigns inside your existing customer base, and net-new prospecting where the audience has not already self-selected for being close to converting.

How this plays out inside a sales team

Marketing is not the only function that pays the tax. When retargeting inflates credit for deals that were already moving, sales leadership sees a paid channel apparently driving pipeline and starts asking why the sales team is not converting more of the leads it is supposedly generating. The retargeting number becomes ammunition in a conversation that has nothing to do with what actually happened in the deal, and the account executive who did the real work of running four calls and a security review gets none of the credit for it.

I have sat in enough of these meetings to know the pattern. Finance asks why CAC looks efficient on paper while the pipeline review looks thin in reality. Nobody can reconcile the two because the attribution model and the sales team are describing two different worlds. Fixing the retargeting audience does not just save media budget, it removes a source of internal argument that wastes real hours every month.

Watch for the audience that grows itself

A retargeting audience with a rolling 90-day window grows on its own as traffic increases, which means the spend line grows on its own too, without anyone deciding to increase it. Most budget owners notice a retargeting line creeping up and assume it means more demand. It usually means the audience pool simply got bigger, and a bigger low-signal pool produces a bigger low-signal bill. Review the audience size next to the spend every month, not just the spend on its own.

The close

Retargeting is not a bad channel. It is a channel that is almost always measured badly, and bad measurement turns a narrow, useful tactic into a budget line that grows every quarter because nobody can find a reason to cut the best number on the dashboard. Name the four jobs, cap the frequency, shrink the window, and run the holdout. The number you get afterward will be smaller and it will be real, which is a better trade than the other way around.