Free 30-minute strategy session — no obligation  |  Call +1 720-712-8615
Retargeting Agency

Retargeting Takes Credit For Sales You Already Had.

Geo-holdout studies across more than fifty brands found median retargeting incrementality of about 28%. Roughly seventy cents of every retargeting dollar funds purchases that were happening anyway. We still think you should run it — we just think you should know which seventy cents.

Median incrementality~28%Recast geo-holdout studies
Reported vs real ROAS40–70%Typical overstatement
Our approachHoldoutBuilt in from month one
Plain English

What is retargeting?

Retargeting, also called remarketing, is advertising shown specifically to people who have already interacted with your business — visited your website, viewed a product, abandoned a cart or opened an email. A tracking pixel records the visit, adds that person to an audience, and ads follow them across other websites, social platforms and connected TV until they convert or the audience window expires.

Retargeting companies sell it as the best-performing channel in almost every advertising account, and by every metric the platforms report, it is. Retargeting display converts at roughly double the rate of standard display. Cost per acquisition is usually the lowest in the account. Return on ad spend frequently looks spectacular.

And most of that is an illusion. Not fraud, not incompetence — an artefact of measurement. Retargeting audiences are, by definition, people who already showed interest. They visited your site. They looked at the product. Many of them were going to come back regardless. When one of them does come back and buy, the retargeting platform records a conversion and takes the credit, because the ad was the last thing it saw.

That is the single most important thing to understand about this channel, and it is why a remarketing agency page opens by telling you the numbers are inflated. We would rather have that conversation now than in month four.

Tool 01

The test that settles it,
in about forty seconds

A holdout test is simple enough to explain in a sentence. Take your retargeting audience, deliberately withhold ads from a random slice of it, and compare what the two groups do. The difference is the only thing your advertising actually caused. Everything above that line was going to happen anyway.

Press run and watch it play out on 1,000 returning visitors, with a 10% holdout.

Ready

Exposed group — 900 people

These visitors are shown retargeting ads as normal.
Converted0

Holdout group — 100 people

These visitors are deliberately shown nothing at all.
Converted0
Exposed conversion rate
Holdout conversion rate
Incremental lift
Press run. Both groups will convert. That is the entire point, and it is the thing platform reporting can never show you, because a platform cannot report on people it deliberately did not advertise to.

A demonstration of the method using illustrative figures in the range published research reports for retargeting. Your own numbers will differ — the point is the shape of the result, and that you cannot know your shape without running the test.

Tool 02

What your reported ROAS
is probably worth

Until you run a holdout you do not know your incrementality rate. But you can see what different rates would mean, and published research narrows the range considerably: analyses of retargeting campaigns commonly find incrementality between 20% and 40%, with one set of geo-holdout studies across more than fifty brands reporting a median near 28%.

Set your own numbers and an assumed incrementality rate below. The left column is what your dashboard says. The right is what is probably true.

What the platform reports

Conversions claimed 120
Revenue claimed $31,200
Cost per conversion $33
Reported ROAS7.80x

What is probably true

Conversions actually caused 34
Revenue actually caused $8,736
True cost per conversion $119
Incremental ROAS2.18x

Incrementality ranges drawn from published 2026 analyses including Recast geo-holdout studies, Northbeam research and Haus lift experiments. The 28% default is a reported median, not a prediction about your account. The only way to know yours is to measure it.

Why this is not an argument against retargeting

An incremental ROAS of two is a good business. It is simply a very different business from an incremental ROAS of eight, and every decision downstream depends on knowing which one you have. Budget allocation, bid caps, how hard you scale, whether to shift money toward prospecting — all of it changes.

The failure mode is not running retargeting. It is running retargeting on reported numbers, concluding it is your best channel because the dashboard says so, and moving budget out of the channels that actually create demand and into the one that harvests it. Research on this is consistent: prospecting campaigns typically show incrementality of 60 to 80%, roughly double retargeting's. The channel that looks worse on your dashboard is frequently doing more of the real work.

Tool 03

Do you even have
enough people to retarget?

This is the question that should be asked before anyone builds a campaign, and it usually is not. Retargeting needs a pool large enough to deliver against, and the pool is much smaller than raw traffic suggests. Ad blocking removes some. Cookie restrictions and opt-outs remove more. And your audience window quietly expires people every single day.

Usable audience5,200
Refreshed per week1,213
VerdictViable

A planning model. It applies your untrackable share to visitors accruing over the window you set, which approximates how audience membership actually behaves. Real pool sizes depend on returning-visitor overlap, platform-specific match rates and how aggressively you exclude converters.

The window trade-off nobody explains

WindowWhat it meansUse when
1–7 daysSmallest pool, highest intent. These people are still actively considering.Short purchase cycles, cart abandonment, high traffic volume
14–30 daysThe usual default. Reasonable balance of size and intent.Most businesses, most of the time
60–90 daysLarger pool, considerably weaker intent. Frequency falls as the pool grows.Considered purchases, long sales cycles, low traffic
180 daysMostly people who have forgotten you exist. Cheap impressions, thin returns.Rarely. Usually a symptom of not having enough traffic

If you find yourself stretching to 180 days to fill an audience, the honest answer is that you do not have a retargeting problem — you have a traffic problem, and the money belongs in paid search or social until you do.

Scope

What a retargeting agency
should actually be doing

Setting up a retargeting campaign takes about twenty minutes. Everything below is what separates a retargeting agency whose work compounds from retargeting companies that quietly bill you for demand you already had.

Holdout measurement

A permanent control group so incrementality is known rather than assumed. This is the whole argument and almost nobody runs it.

Converter exclusion

Removing people who already bought. Astonishingly common to find accounts advertising to existing customers at full price.

Audience segmentation

A cart abandoner and a blog reader are not the same person and should not see the same ad at the same bid.

Frequency capping

Past a certain point additional impressions stop persuading and start irritating, while still costing full price.

Window tuning

Matching the audience window to your actual sales cycle rather than accepting a platform default.

Tracking integrity

Pixel, server-side events and match quality. Covered in depth under Meta Pixel setup.

Creative sequencing

Showing a different message at impression twelve than at impression one, because the same ad forty times persuades nobody.

Cross-channel delivery

Retargeting runs across display, social, search and connected TV from one audience strategy.

Landing page continuity

Sending someone back to a generic homepage wastes the click. See conversion optimization.

Free incrementality review

Send us your retargeting numbers. We will tell you what your reported ROAS is likely worth, and design a holdout test you can run yourself in about an hour.

Get my review
How we work

Running retargeting
so the numbers mean something

  1. Establish the holdout before anything else. A permanent 5 to 10% control group, set up on day one. Once campaigns are running it becomes politically difficult to introduce, which is precisely why most accounts never have one.
  2. Verify the tracking. Pixel firing correctly, server-side events in place, match quality checked. Bad audience data produces bad audiences and there is no fixing it downstream.
  3. Exclude everyone who already converted. The first thing we look for and one of the most common findings. Paying to advertise to existing customers is pure waste and it is invisible on a ROAS report because those people buy again.
  4. Segment by behaviour, not by page. Cart abandoners, product viewers, pricing page visitors and casual readers get different messages, different bids and different windows.
  5. Cap frequency deliberately. Uncapped retargeting finds heavy browsers and shows them the same ad relentlessly. It looks efficient in aggregate and does considerable brand damage.
  6. Sequence the creative. Impression one and impression twelve should not be the same message. If someone has seen an ad eleven times without acting, repeating it a twelfth time is not a strategy.
  7. Read the holdout after four weeks. Published guidance suggests a control group generating at least a couple of hundred conversions before the result means anything. We will say when it does not yet.
  8. Rebalance on incremental numbers. If incremental ROAS comes in below your break-even, cap frequency, narrow the audience or cut budget and move it to demand generation. The test only pays for itself if you act on it.
The other cost

The point where retargeting
starts working against you

Everybody has experienced badly run retargeting. You looked at a pair of boots once and they followed you for six weeks across every website you visited, long after you bought them somewhere else. That experience is what the phrase "creepy advertising" describes, and it is not caused by the technology. It is caused by three settings nobody adjusted.

What went wrongThe fix
No frequency capLimit impressions per person per week. There is a point past which additional exposures persuade nobody and simply accumulate resentment.
No converter exclusionRemove people the moment they buy. Following a customer with ads for the thing they just purchased is the most visible possible sign that nobody is minding the account.
Window far too longSomeone who visited five months ago and never returned has made their decision. Let them go.
One creative, foreverThe same image forty times reads as automation rather than communication. Sequence it or cap it.

This matters commercially, not just ethically. Reported conversions cannot see brand damage, so an account can be simultaneously producing an excellent ROAS and steadily annoying a portion of your market. It is one more reason the reported number should not be the only number you look at.

Sector fit

Where retargeting
genuinely earns its place

Incrementality varies enormously by how long the buying decision takes. The longer and more considered the purchase, the more genuine work retargeting does.

Questions we get asked

Retargeting
questions, answered

Free incrementality review

Send your retargeting numbers. We will tell you what they are likely worth and design a holdout you can run yourself.

Your information is never sold or shared. We respond within one business day.