What is retargeting and how does it work?

Retargeting is showing ads only to people who have already interacted with you, most often by visiting your website. It works because you put a small piece of code called a pixel or tag on your pages. When someone loads the page, that code quietly reports the visit to an ad platform such as Meta or Google. Inside a walled garden the platform already knows exactly who that person is, because the same company owns your tag and the feed they scroll, so it can serve your ad to that specific account later with no guesswork. On the open display web the match is far weaker: it relies on third-party cookies and a process called cookie syncing, where ad tech companies swap ID tables so they can agree that two different cookies belong to the same browser.

Does retargeting actually work, or does it just take credit for sales that would have happened anyway?

It moves some people, and it still costs more than it brings in. Randomised field experiments do find a real effect: Sahni, Narayanan and Kalyanam measured 14.6% more users returning to a retailer's site within four weeks when retargeting was switched on. But the number Meta and Google report is far larger than the effect they caused, because the ads are shown to the people most likely to buy anyway. In a typical holdout test, a campaign that claims 8,000 sales turns out to have caused only about 2,200 of them, or 27.5%. Divide the reported cost per customer by that and a $22 customer really cost $80, which is more than the same platform charges you to reach a total stranger. The only way to know your own number is to run a holdout: withhold ads from a random slice of the audience and compare.

What are the limitations of retargeting?

The biggest limit is structural: your retargeting audience can never be bigger than the traffic you already have, so it multiplies demand and never creates it. On top of that, privacy defaults delete most of the audience before you can reach it. Roughly 29.5% of internet users run an ad blocker, which stops the pixel firing at all. Cookie consent acceptance averages 42% to 47% in consent-required regions. Safari caps JavaScript-set cookies at 7 days, and at 24 hours when the visitor arrives from an ad click, which quietly guts any 30 or 90 day audience. On iOS apps, retargeting is largely broken without App Tracking Transparency consent, which only 15% to 30% of users grant. In practice a site with 10,000 monthly visitors can typically reach only about 2,600 of them.

How much does retargeting cost and is it cheaper than normal ads?

The reported cost is much cheaper and the real cost is not. Retargeting CPMs run about $5 to $15 and clicks about $0.50 to $1.20, and the cost per customer Meta and Google report is commonly up to 50% below what they charge for a stranger. But that reported figure has two problems. First, a typical holdout finds the ads really caused only 27.5% of those sales, so a campaign reporting the benchmark $22 cost per customer is really costing $80, which is more than the same platform charges to reach a total stranger: $42 to $48 on Meta and $49 on Google Search. Second, you cannot scale into it: the audience only exists because you paid for the traffic that created it, so moving budget out of advertising to strangers shrinks the very list you are retargeting.

Does retargeting work better if the visitors came to your website organically rather than from an ad?

It removes half the objection and leaves the other half standing. If the traffic arrived from search, word of mouth or a forum post, the audience genuinely costs nothing and spending on it takes nothing away from your other advertising, so the usual warning that retargeting cannibalises the very budget that creates its audience does not apply. What does not change is the price. The 27.5% of sales that a holdout typically credits to retargeting is low because the audience is selected on intent, and someone who found you by searching is at least as likely to have bought anyway as someone who clicked an ad. A $22 reported customer is still an $80 real one, against $42 to $48 for a stranger on the same platform. So your price decides it: at $80 a customer you need a customer worth roughly $240, and the median subscription app earns $21.37 per payer in year one. The one genuine advantage is that a free audience makes the holdout free too, so you can measure your own number instead of guessing. Retargeting people who read an article about your subject rather than people who looked at your product is a weaker case again: the share of sales the ads really cause should be better than 27.5%, because those readers were never shopping, but they convert far worse for exactly the same reason, and they are worth more as a seed for finding similar strangers than as a retargeting audience.

Does Google Display retargeting track people through Chrome?

No. The tracking is done by the Google tag you install on your own pages, or by a GA4 audience linked to Google Ads. Chrome does not report anyone's browsing to your ad account, and there is no browser setting that switches retargeting on. If the tag is not on the page, the visit does not exist as far as Google is concerned. Chrome does matter for a different reason: serving those ads back out across the Google Display Network relies on third-party cookies, which Safari and Firefox block outright and which Chrome retained after reversing its plan to remove them. So open-web display retargeting still works on Chrome and largely does not work elsewhere, but that is a limit on how many people you can reach rather than the mechanism that recognises them. Google Display is also the most expensive of the three main retargeting options once you correct for the sales the ads really caused: about $138 per customer against $95 on Google Search and $80 on Meta.

How much traffic do you need before retargeting is worth setting up?

Meta and Google will both technically run an audience of 100 people, and Google standardised its minimum to 100 active users across Search, Display and YouTube. But delivery is erratic below roughly 1,000 matched people. Because privacy defaults typically cost you about three quarters of your visitors, a 1,000-person reachable audience needs something like 4,000 monthly visitors. Below that, retargeting is not wrong so much as a rounding error, and the effort is better spent on the traffic itself. The exception is that you should install the tag on day one regardless, not for retargeting but because without it Meta and Google cannot tell you which ad produced a sale, cannot aim a campaign at buyers rather than clickers, and cannot build an audience of strangers resembling your existing customers.

What Is Retargeting and Does It Actually Work? How It Works, What It Costs and Its Hard Limits

You looked at a pair of shoes once and now those shoes follow you across the internet for a fortnight. An ad that follows you after you visit a site is called retargeting. Advertisers are told retargeting is the cheapest thing in marketing: 10 times the click rate, half the cost per customer, free to set up. All three of those retargeting claims are true, but all three are misleading.

The short version:

The verdict: retargeting does not work, because it costs more than it brings in. Meta says a retargeted customer costs $22, but only about a quarter of those buyers needed the ad, so the real price is $80. Meta's ordinary ads find you a customer who never visited your site for $42 to $48.
Install the free Meta and Google tracking tag on your site anyway, for a reason other than retargeting. On your order-confirmation page the tag is the only way Meta and Google find out a sale happened. That sale report lets them tell you which ad worked and find strangers who resemble your existing customers.
Retargeting is not a new audience. Retargeting only shows ads again to people who already visited your site.
If your visitors arrived free, from search or word of mouth, retargeting them takes no money from finding new customers. Retargeting free visitors is worth a test, but an $80 retargeted customer still costs more than a $45 customer from ordinary ads, so your product's price decides whether it pays.
A tracking tag on your page tells Meta or Google which logged-in account visited your site, so inside Meta's and Google's own apps there is no guessing who to show the ad to.
A retargeting audience is free to set up. Both Google and Meta now run a retargeting audience of as few as 100 people, though ads reach an audience of under 1,000 people erratically.
The retargeting numbers Meta and Google report to you are real. Retargeting ads get around 10x the click rate of ordinary display ads shown to strangers. They convert roughly 3.8% against 2.2% for ads that are not retargeted. Meta and Google report up to 50% lower cost per customer.
Ad blockers, cookie consent and Safari's privacy rules remove most of your retargeting audience. 29.5% of internet users run ad blockers. Only 42% to 47% of visitors accept cookies where consent is asked. Safari caps cookies at 7 days (24 hours from an ad click). Together, those rules leave a site with 10,000 visitors a month able to reach about 2,600.
App retargeting on iPhones is largely broken without permission from Apple's App Tracking Transparency prompt, which only 15% to 30% of iPhone users grant.

Retargeting buys a second conversation with traffic you own

Ordinary advertising pays to reach strangers who have never heard of you. Retargeting buys a list of people who already did something: visited a page, watched a video, added to a cart or installed an app.

A retargeting audience is a list, not a description. Normal targeting says "women aged 25 to 40 interested in fitness". Retargeting says "the 812 people who opened my pricing page in the last 30 days".
A retargeting list can only ever be as big as the traffic you already have.

Type of retargeting	What puts someone on the list	Where the ad then appears	How well it works next to other retargeting	Is it even available to a small product
Site retargeting	Loading any page carrying your tag	Any platform whose pixel you installed, plus the open display network	Most volume and the steadiest delivery of the seven retargeting types	Yes, this is the default retargeting type
Cart or checkout abandonment	Reaching a specific deep page and not converting	The same places as site retargeting, but with a much sharper message	The best-converting slice of retargeting, though its results are still the platform's own reported numbers	Only if you have a multi-step checkout
Customer list (CRM) retargeting	You upload hashed emails or phone numbers	Anywhere the platform can match the account	Depends on match rate, the share of uploaded emails matched to accounts, typically 60% to 80%	Yes, and a customer list keeps working when cookies are blocked
Engagement retargeting	Watching your video or opening your lead form	Inside that platform only	Weaker intent to buy than a site visit shows, but far more people	Yes, and it needs no website at all
Dynamic product retargeting	Viewing a specific product, which is then shown back to you	Display and social, as an auto-generated product ad	Worse than generic ads on average, per published research	No, needs a product catalogue feed
App retargeting	Installing or opening your app	Other apps, via the mobile advertising ID	Largely broken on iOS without tracking consent	Rarely, and almost never on iOS
Search retargeting (RLSA)	Visiting your site, then later searching on Google	Google search results, at an adjusted bid	High intent: these people are searching and already know your site	Yes, but only if you already run search ads

The one row worth arguing with

Dynamic retargeting, the shoes-following-you-around kind, is on average the worst performing form. In a field experiment by marketing researchers Anja Lambrecht and Catherine Tucker, published in the Journal of Marketing Research, showing people the exact product they had browsed was less effective than a generic brand ad, except for people whose preferences had clearly moved on.

A tag on your page hands your visitor identity to Meta

You tell Meta who visited your site, and Meta already has that visitor logged in to Facebook or Instagram. A Meta pixel identifies your visitor in four steps:

1. You paste a free snippet of JavaScript, called a pixel or tag, into your site's <head>.Setup

2. A visitor's browser runs that code, which calls Meta's servers with the cookie from their Facebook or Instagram login, so Meta already knows who that is.Identify

3. Meta records "account #4471 visited your /pricing page at 14:02" in your ad audience.Store

4. Next time account #4471 opens Instagram, you can bid to show them your ad, with no guesswork.Serve

Retargeting follows an account, not a browser, so it works across devices. A person can visit your site on a laptop at work and see your retargeting ad on their phone that night.

Outside Meta, Google and TikTok, where people are not logged in, recognising a visitor is much less reliable.

How you get recognised	What the match is actually based on	How reliable it is	What breaks it	Works across devices
Meta pixel, logged-in account	Their Facebook or Instagram login, read as a first-party cookie on Meta's own domain	Deterministic: an exact match, because Meta both runs the tag and holds the login	Ad blockers and consent refusals stop the tag firing	Yes, because the Meta ID belongs to the person, not the device
Google tag, Google account	Their signed-in Google identity, plus the click ID on ad traffic	Deterministic (an exact match) where signed in, weaker where not	Signed-out browsing, consent mode, ad blockers	Yes, where signed in
Hashed email upload (Customer Match)	You upload emails, the platform hashes and matches them to accounts	60% to 80% match rate of uploaded emails, and cookie blocking does not affect it	People using a different email with the platform	Yes, and an uploaded email list survives every other privacy block
Server-side events (Conversions API)	Your own server posts the event to the platform, not the browser	Recovers 20% to 40% of the events the browser tag would have lost, not all of them	Needs real engineering to implement properly	Only where you can identify the user yourself
Third-party cookie and cookie syncing	Ad tech firms swap ID tables so two cookies can be agreed to be one browser	Leaky: every handover between ad tech firms loses matches	Safari and Firefox block it outright, and it works less well in Chrome every year	No, it identifies a browser, not a person
Mobile advertising ID (IDFA, GAID)	A device-level identifier passed between apps	Collapsed on iOS since App Tracking Transparency	15% to 30% opt-in on iOS leaves too few iPhones to retarget	No, it is one device

The same tag on your order page is the only way Meta and Google learn you made a sale

On your order-confirmation page, the tag plus one extra line of code reports "someone paid, $34.99" rather than "someone looked at this".
That confirmation-page report is the only way Meta or Google ever learns you made a sale, because the payment happens on your site or inside Stripe.
Knowing which visitors paid lets Meta tell you which ad worked and aim your next campaign at strangers who resemble your buyers. Neither of those uses of sales data has anything to do with retargeting.
Apps need a software development kit (SDK) instead of a website tag, to report sales to Meta and Google the same way. You add Meta's SDK or Google's Firebase SDK to report events such as the first open and a $34.99 purchase.
App reporting works properly on Android and is crippled on iPhones. Unless a person taps Allow on Apple's tracking prompt, which only 15% to 30% of iPhone users do, app sale reports arrive a day or more late.

Use Camera to Clipboard to move a photo off a phone with no upload

The usual way to move a photo from a phone to a computer is to email it to yourself or put it in a cloud album, which hands it to a company for no reason. Camera to Clipboard, one of my own apps, sends the photo straight from the phone to the computer instead.

Download Camera to Clipboard from the App Store

Download Camera to Clipboard from Google Play

The phone and the computer send the picture to each other over your own local network, so it never reaches anybody else’s servers.
A photo taken inside Camera to Clipboard is on your computer’s clipboard the moment you take it, with no cable and no file to go and find.
Pictures already in the phone’s library can be sent to the computer with Camera to Clipboard, several at once.

Set the tag up anyway, because it makes all your ads work

1. Create a free Meta or Google ad account.
2. Copy the platform's tracking tag into the <head> of every page on your site.
3. Mark the pages that show a visitor is interested in buying, such as pricing, checkout, signup and thank-you.
Step four is to save a retargeting audience, which is just a filter such as "visited in the last 30 days".
Step five is waiting, because a retargeting audience only grows as fast as new visitors arrive on your site.

Start the tag before you need it

A retargeting audience is built only from visits that have already happened, so deciding in November to advertise a December launch means starting with an audience of zero. Set every retargeting audience to the longest membership period the platform allows from day one, because visits that were never recorded cannot be recovered.

Ad platforms differ in the smallest retargeting audience each will show ads to and how long each remembers a visit. Installing a platform's free tag can be worth it even when buying its retargeting ads is not.

Platform	Smallest audience it will run	How long someone stays on the list	What you install	Worth installing the tag
Meta (Facebook, Instagram)	100 people technically, 1,000+ before delivery is stable	Up to 180 days for website visitors	Meta pixel, plus Conversions API if you can	Yes. Meta's tag is what measures and aims your ordinary campaigns, whether or not you ever retarget
Google Ads	100 active users across Search, Display and YouTube (was 1,000 on Search)	Up to 540 days on Display, 180 on Search	Google tag, or GA4 audiences linked to Ads	Yes. Google's tag measures and aims your ordinary Google campaigns, and it feeds GA4 (Google Analytics) as well
Microsoft Ads (Bing)	300 people before it will serve the list at all	180 days, and that is a hard cap you cannot raise	The UET tag	Only if you already run Bing search ads
TikTok	1,000 people for a custom audience	Up to 180 days	TikTok pixel or Events API	Only if you already advertise on TikTok
Pinterest	100 matches on an uploaded customer list; Pinterest publishes no floor for site visitors	Up to 540 days, tied with Google Display for the longest period of any platform in this table	The Pinterest tag	No unless you already advertise on Pinterest
Snapchat	1,000 people, the highest minimum of any self-serve platform in this table	Up to 395 days	The Snap Pixel	No unless you already advertise on Snapchat
LinkedIn	300 people and a high minimum price per impression	Up to 365 days	LinkedIn Insight Tag	Only if you sell B2B and already advertise on LinkedIn
Reddit and X	Around 100 to 300, varies by objective	Typically 180 days	Their own pixel	No. Small audiences and few places to show the ads
Third-party networks (AdRoll, Criteo)	Effectively thousands of visitors to be worth their fee	Varies by contract	Their tag, plus a monthly platform fee	Never. A monthly fee for reach that barely works any more

Two settings matter more than any bid you will place

An audience of pricing-page visitors beats an audience of anyone who visited. Pricing-page visitors and all visitors should get separate budgets and different ad messages.
A 7-day retargeting audience outperforms a 180-day one on almost everything. A long retargeting audience period builds a bigger audience, not better results.
Always exclude existing customers from a retargeting audience. Existing customers are the most likely people to click, so advertising to them makes your retargeting numbers look better while wasting the budget.

It clicks 10 times more and converts twice as well, both real

The statistics quoted in every retargeting pitch are mostly accurate. What goes wrong is what advertisers take those retargeting statistics to mean.

The claim you will be shown	The number	Is it actually true	What it really proves
"Retargeting clicks 10x better than display"	0.7% against 0.07% for standard display	Yes, and newer data sets put the retargeting click rate at 0.9% to 1.2%	That the click rate of standard display ads is catastrophically low, not that retargeting is strong
"It converts far better than cold"	3.8% against 2.2% non-retargeted	Yes, and retargeting converts 3x to 5x as well as ads shown to pure strangers	That people who already visited you are closer to buying, which you knew
"It halves your cost per customer"	Up to 50% lower reported CPA	True as reported, but it is the platforms' own reported figure, not a measured one	Nothing, until the lower cost survives a holdout test that withholds ads from a random group
"Retargeting CPMs are cheaper"	$5 to $15 against $12 to $20 broad cold	Contested: other data sets show audiences of past visitors at the highest cost per 1,000 impressions	Both happen: an ad relevant to past visitors costs less, but a tiny audience pushes the price back up
"Retargeting lifts return visits"	14.6% more users returned within four weeks	Yes, and this one is causal: a randomised field experiment	A real measured effect: retargeting caused more people to come back

The first four retargeting claims compare people who saw retargeting with people who did not, and those two groups were never comparable. Only the return-visits figure comes from an experiment, so it is the only claim that shows retargeting ads caused anything.

The ads reach people who were going to buy anyway

A retargeting audience is chosen because its people already showed interest in buying. Everyone on a retargeting list already came to your site.
Meta or Google then claims the sale for the retargeting ad, even when the buyer would have bought without the ad.
Many platforms also credit a sale to an ad that was only displayed and never clicked.
Only a holdout test measures whether retargeting caused the sale. Randomly withhold ads from part of the audience, then compare how often the group that saw ads and the group that did not went on to buy.

Sales Meta says the campaign produced8,000

Purchases made by the exposed 80% of the audience12,200

Purchases made by the held-out 20%, scaled to the same size10,000

Purchases the advertising actually caused, which is 27.5% of the reported figure2,200

27.5% of the sales Meta and Google credit to retargeting typically survive a randomised holdout test. So every reported retargeting cost per customer has to be divided by about four.

The error runs both ways, so neither dashboard is right

Meta and Google claim too many sales for retargeting. Meta and Google count people who only saw the ad, and assume everyone who bought was persuaded by it.
Google Analytics credits retargeting with too few sales. In one controlled group experiment, Google Analytics credited remarketing with 1.7% of conversions while the measured lift against the control group was 18.5%.
Meta overclaiming and Google Analytics underclaiming can both be true on the same campaign.
The real effect of retargeting varies so widely that only your own test answers it.

How you are measuring	What it actually counts	Which way it gets it wrong	How hard it is to run	Should you believe it
Platform-reported conversions	Anyone who converted after clicking, or often just after seeing, your ad	Claims far too much: Meta and Google measure the results of the ads they sold you	Zero effort, it is the default report in your ad account	Only as a trend line, never as the true number of sales
Last-click analytics (GA4)	Only conversions where the ad click was the final touch	Claims far too little: measured at 11x too low in one experiment	Zero effort, already installed	No, especially where most people see the ad without clicking it
Platform conversion lift study	A randomised no-ads control group run by the platform itself	Roughly right, though the platform designs the test	Moderate: needs minimum spend to qualify	Yes, the most accurate answer for the effort
Your own audience holdout	The difference between a random exposed group and a random excluded one	The actual answer for your account	Moderate: a week of setup, weeks of patience	Yes, this is the most reliable method
Geo holdout	Turning retargeting off entirely in some regions and comparing them with matched regions	Clean and causal, but blunt	Hard: needs scale and matched markets	Yes, if you are big enough to run it
Eyeballing it after switching it on	Every other change that month, mixed in with the effect of the ads	Unbounded in both directions	None	No

Your audience cannot exceed your traffic, minus three quarters

Here is what happens to the retargeting audience of a site with 10,000 monthly visitors, which is more traffic than most independent products have:

Monthly website visitors10,000

Minus the 29.5% who run an ad blocker7,050

Minus consent refusals, at 42% to 47% average acceptance where consent is required3,170

Minus Safari's 7-day cookie cap, on its 24% browser share2,590

Share of your visitors you can actually put an ad in front of~26%

What eats your audience	What it does	How much it costs you	Can you do anything about it
The traffic ceiling itself	You cannot retarget people who never came	Absolute: your traffic caps all of your retargeting	No: the only fix is more traffic
Ad blockers	Block the pixel from loading, so the visit is never recorded	29.5% of users globally, 32.5% in the US, 38% to 40% in the UK	Partly: server-side events recover some blocked visits
Cookie consent	No consent, no tracking, in the EU, UK and increasingly elsewhere	42% to 47% average acceptance, under 25% in some countries	Somewhat: the design of the consent banner changes the acceptance rate a lot
Safari ITP	Caps JavaScript cookies at 7 days, and at 24 hours from an ad click	Guts long windows (audiences kept longer than 7 days) for the 24% of users on Safari	Partly: server-side tagging restores longer lifetimes
Apple's App Tracking Transparency	Removes the device ID that app-to-app retargeting depends on	Only 15% to 30% opt in, so the number of iPhones you can retarget collapsed	Barely: sending people from a web page into the app is the workaround
Third-party cookie decline	Open-web display retargeting depends on cookies Safari and Firefox already block	Chrome kept them (third-party cookies) after Google reversed course, but retargeting through them keeps getting less reliable	Yes: retarget inside Meta and Google, where people are logged in
The platform minimum	An audience under the platform's minimum size will not be shown ads	100 people to run at all, ~1,000 to run reliably	Yes: keep people on the list for longer, or combine audiences

A mostly US audience leaves far more people to retarget, because US sites mostly do not need consent banners. A site with 10,000 mostly US visitors a month leaves roughly 5,000 people you can reach, rather than 2,600.
Hashed email lists are the one retargeting audience that survives ad blockers, consent refusals and Safari's cookie limits, because a matched account does not depend on cookies, browsers or ad blockers.
Retargeting needs roughly 4,000 monthly visitors to deliver reliably in a consent-heavy market, working backwards from the 1,000-person audience the platforms need for steady delivery.

Use Repeat Recorder to rehearse a demo video instead of buying ads

Under roughly 4,000 monthly visitors there is nobody to retarget, so what matters is making strangers turn up, usually with a two minute video of the product working. Repeat Recorder, another of my own apps, records you and plays each take straight back without you touching the phone.

Download Repeat Recorder from the App Store

Download Repeat Recorder from Google Play

Repeat Recorder plays each take back automatically and then starts recording again, so you can run the same opening line twenty times without picking the phone up.
Meta and Google charge $42 to $48 for each new customer from ordinary ads, while a better demo video, made in an afternoon, is yours to keep.
In Repeat Recorder, takes worth keeping go into favourites and into folders you name yourself, so two readings of the same sentence sit side by side.

A small audience forces you past the creepiness cliff

Showing the same person an ad too often is almost always an arithmetic accident rather than a decision.

Most of a retargeting ad's value comes from the first two times a person sees it. Roughly 80% of an ad's weekly impact occurs in the first two views.
After about 5 to 7 views of the same ad, past visitors start to dislike it. After 10 or more views, measured negative feeling toward the ad rises by about 16%.

Your reachable retargeting audience, from the chain above2,590

A modest monthly retargeting budget$200

At a $10 CPM, that budget buys impressions20,000

Times each person sees your ad this month7.7

The counter-intuitive fix

If people see your retargeting ad too many times a week, the answer is usually to spend less, not to change the ad. Cap each person at around 3 to 5 views per week and put the rest of the budget into finding new people.

Its cheap cost per customer is partly an accounting illusion

What you are comparing	Advertising to strangers	Retargeting	Why the gap exists
Cost per 1,000 impressions	$12 to $20 for broad targeting of strangers	$5 to $15, but small audiences of past visitors can cost more	Relevant ads earn a lower price, but advertisers competing for a tiny audience push it back up
Cost per click	Around $0.30 to $1.00 on Meta	$0.50 to $1.20, dearer per click but far more clicks per impression	A higher click rate offsets the dearer click, so the two end up close
Conversion rate of that traffic	0.7% to 1.5%, typical for strangers	3.8%, up to 3x to 5x the rate for strangers	Selection: these people already chose to visit you
Cost per customer Meta and Google report	$42 to $48 on Meta, $49 on Google Search	Up to 50% lower, about $22 on Meta	Meta and Google count sales the retargeting ads did not cause
Cost per customer after a holdout	Roughly what it said. Strangers were not already trying to buy	About $80, which is more than a stranger costs	Divide the reported figure by the share the ads really caused
Can you scale it by spending more	Yes, until you exhaust the market	No: extra budget buys repetition, not reach	The audience is fixed by last month's traffic, not by your budget
What happens if you cut it	Your retargeting list shrinks next month	Nothing else is affected	Retargeting is built from the visitors that advertising to strangers brings in, never a substitute for it

Monthly retargeting spend$2,200

Customers Meta reports, at the benchmark $22 each100

How many of those 100 the ads actually caused, at the 27.5% a holdout typically finds27.5

What one real customer actually cost you, which is $2,200 divided by 27.5$80.00

$80 is what a "$22 per customer" retargeting campaign really costs, because most of those buyers never needed the ad. That is more than the same platform charges for a total stranger, which is $42 to $48 on Meta and $49 on Google Search.

The most expensive misunderstanding

A cheap reported cost per customer is not a discount you can scale into. The retargeting list only exists because money spent on strangers created it, so moving that money leaves nobody to retarget. Advertisers typically budget 20% to 30% of ad spend for retargeting, but at $80 a retargeted customer against $42 to $48 for a stranger, the better answer is zero.

Work out what retargeting would cost you specifically

Three questions decide whether retargeting pays for you: how many people can you reach, how often will they see your ad, and what does one real customer cost. Set your own numbers in the calculator, or use its presets.

A free audience is additive but still not cheaper

Many visitors arrive free, from search, forums or friends.

Retargeting free visitors takes no money away from finding new people. Free traffic keeps arriving whether or not you advertise, so retargeting it takes nothing from other advertising.
Retargeting free visitors still costs more per real customer than ordinary ads do. Arriving free does not make people less likely to have bought anyway, so a retargeted customer still really costs $80, against $42 to $48 for a total stranger.
Your product's price, not where your visitors came from, decides whether retargeting pays. At $80 a customer, the usual rule that a customer should return three times their cost means each must be worth about $240, while the median subscription app earns $21.37 per payer in a year.
Retargeting free visitors gives you something paid visitors never do: a holdout test that costs nothing. Withhold retargeting ads from a random 10% to 20% of those visitors to measure how many sales your ads really cause.

A subject page converts worse but proves more

Take an accent training app that publishes an article on how accents form, read by people who came to learn, not to buy.

Retargeting ads shown to article readers should cause more than 27.5% of the sales they are credited with, which is good news. A reader of an explainer article is not shopping at all, so there are far fewer sales for the ad to take undeserved credit for.
Retargeted article readers will buy at a far lower rate, which is bad news. The 3.8% retargeting conversion rate was measured on people who looked at products.
Nobody has measured whether article readers' higher share of caused sales outweighs their lower buying rate. Because nobody has measured it, only your own holdout test can tell you whether retargeting article readers pays.
Do not show article readers ads for a product they never looked at. Marketing researchers Lambrecht and Tucker found specific product ads only beat generic brand ads for people whose preferences had already formed.

Free article readers are worth more as email signups and a seed for similar strangers

The three ways a visitor reaches a retargeting list, side by side:

How the visitor got to you	What the audience costs you	Does spending here shrink your other advertising	How many of the sales the ads really cause	Should you spend money on it
They clicked one of your ads	The price of the click, which is most of your ad budget	Yes. Moving ad money into retargeting makes next month's list smaller	About 27.5%. Visitors who clicked an ad were already shopping	No. $80 a customer against $42 to $48 for a stranger
They found you in search and looked at your product	Nothing, and search traffic keeps arriving	No. Search traffic comes whether or not you advertise	Still about 27.5%. Arriving free does not make them less likely to buy anyway	Worth testing, because finding out costs nothing
They read your article on the subject and never saw your product	Nothing	No	Better than 27.5%. Article readers were not shopping, so fewer of their sales would have happened without the ad	No. Article readers buy far less often, and they are worth more as a seed for finding similar strangers

The platforms' minimum audience size rules out most article readers. An article read by 2,000 people a month leaves roughly 520 readers you can reach, well under the 1,000 needed for steady delivery of ads.
Use article readers as a seed for Meta or Google to find strangers who resemble them. Give Meta or Google your list of readers, especially those who paid, so it can find similar people across the whole internet.
Ask article readers for their email address while they are on the page. Emailing those readers then costs nothing, forever, instead of $80 a customer.

Where not to start

Google Display ads are the most expensive place to retarget. Corrected for the sales the ads really caused, a customer costs about $138 on Google Display against $95 on Google Search and $80 on Meta, compared with every other channel.

Seven of the ten reasons to retarget are a straight no

The retargeting industry names ten situations when it tells you to run retargeting.

Your situation	Should you spend money on it	Why	What to do instead
Pre-launch, no traffic yet	No. There is nobody to retarget	The list is built from past visits and you have none	Install the tag anyway, for conversion tracking rather than for retargeting
Under 1,000 visitors a month	No. Not even close	Roughly 260 reachable people are too few for the platforms to show ads steadily	Spend the same money and effort on getting more traffic
5,000 to 50,000 visitors, considered purchase	Only as a test, with a holdout running from day one	Big enough to deliver, but a real customer still costs more than a stranger does	Cap the spend at what you can afford to write off, and measure it properly
Ecommerce with cart abandonment	The strongest case here, worth testing	Reported $13 a customer, or about $47 after dividing by the 27.5% a holdout typically finds, the only situation in this table near the $42 to $48 a stranger costs	A 7-day cart audience only, and hold out a random slice to see what it really did
Free iOS app with in-app purchases	No.	App-to-app retargeting needs tracking consent that 70% to 85% of people refuse	Nothing. Put the money into finding new people
B2B SaaS with a long sales cycle	No, and LinkedIn charges a premium on top	Long decisions do favour reminders, but a LinkedIn impression is the dearest in the table	Email your list of leads yourself. Email costs nothing and reaches those leads properly
Cheap impulse product, instant decision	No.	If shoppers did not buy a cheap impulse product on their first visit, they mostly never will	Improve that first visit to your site instead
You have an email list of past signups	Email them instead. It is free	Uploading the list to Meta means paying to reach people you can already reach for nothing	Email the list directly. Only upload the list to Meta if email has already failed
Organic visitors who looked at your product	Worth testing, because the test itself is free	The audience is free and takes nothing from your other advertising, but a caused customer still costs more than a stranger	Hold out a random slice from day one, and seed a lookalike audience from those visitors either way
Organic readers of your content, who never saw your product	No. Use them as a seed instead	Content readers were never shopping, so they buy far less often than the shoppers behind the $22 benchmark	Take their email on the page, and hand them to Meta or Google to find similar strangers

The setup that wastes the least money if you run it

1. Install the free tracking tag today, even at zero traffic.
2. Set every retargeting audience to the longest period allowed. That is 180 days on Meta and 540 days on Google Display.
3. Email your customer list yourself before you upload it to any ad platform.
4. Build exactly three retargeting audiences. Visitors to deep pages such as pricing, kept for 30 days, all visitors kept for 180 days, and people who already converted, as an exclusion.
5. Cap each person at 3 to 5 ad views a week, and set the budget to what that many views costs.
6. Keep at least 80% of the ad budget on finding new people.
7. Do not use dynamic product ads by default.
8. Run a holdout test before you believe any retargeting result. Exclude a random 10% to 20% of the audience for a month, and switch retargeting off if it does worse than your ordinary advertising.
9. Divide the reported cost per customer by the share your own test says the ads caused.

Use Camera to Clipboard to photograph the product your ads show

If most of the money goes on finding new people, the picture in the ad does the persuading, and a plain image of the product usually beats an automated feed of the thing somebody walked away from. Camera to Clipboard, one of my own apps, is a phone camera that pastes.