Acquisition

Retargeting

Retargeting shows ads to people who already visited your site, used your app or gave you their contact details, so you can bring back the ones who left before buying.

In short

Retargeting, also called remarketing, is advertising aimed at people who have already interacted with a business: visited its website, opened its app, watched its videos or shared an email. It is used to bring back visitors who left without buying. Field experiments show it can lift visits and purchases, but platform reports overstate the effect, so its real value has to be measured with a holdout group.

Origin: No single inventor; online display advertising practice, practice-based; studied in field experiments from 2013
Level: 201 · Tool
Fits: Startup, Small and mid-size, Scale-up
Time to apply: a day to set up lists and a first campaign; four weeks for a holdout test to read
What you need: a tag or pixel on your site (Google tag, Meta pixel, Yandex Metrica) with goals set for key steps · a consent banner that records who agreed to advertising cookies, where the law requires one · enough traffic: Google needs at least 100 active users in a list before it serves ads · a holdout group or a lift study to measure what the ads add

Retargeting is advertising shown to people who have already interacted with a business: they visited the website, opened the app, watched a video or gave an email address. The business tags those people with a pixel or an uploaded list, then buys ads that reach them on other sites, in social feeds or in search results. Remarketing is the same thing under another name. Yandex’s advertising course says outright that the two terms now mean the same, and Google files the feature under remarketing, recently relabelled “your data”.

There is no single inventor. It grew out of display advertising practice, and the academic record starts with field experiments published from 2013 onward. Ecommerce, travel and subscription businesses use it most, because their customers often look several times before buying.

How retargeting works

A tag on your site records who did what, and the ad platform turns those records into audience lists. Google’s help pages list the main types: website visitors, dynamic remarketing that shows the exact products a person viewed, search lists that raise bids when past visitors search again, YouTube viewers, app users and Customer Match lists built from emails customers gave you.

Each list has two settings that matter. Membership duration is how long a person stays on the list; Google defaults to 30 days, allows up to 540, and recommends a window close to your sales cycle. Minimum size decides whether the list can serve at all: Google needs at least 100 active users in the last 30 days for Display, Search and YouTube. In Yandex Direct the same job is done by retargeting conditions built from Metrica goals, Metrica segments or Yandex Audiences lists.

Does retargeting work?

Randomized experiments say it does, with smaller effects than dashboards suggest. The best evidence comes from field tests where a random part of the audience did not see the ads.

Study	Setting	Finding
Johnson, Lewis and Nubbemeyer, 2017	Online retailer, display retargeting	Site visits up 17.2%, purchases up 10.5%
Sahni, Narayanan and Kalyanam, 2019	Home-improvement seller	14.6% more users returned within four weeks
Lambrecht and Tucker, 2013	Online travel firm	Dynamic product ads less effective than generic ads on average
Hoban and Bucklin, 2015	Financial tools provider	No effect on people who visited but never created an account

The Lambrecht and Tucker result surprises most people. Showing someone the exact hotel they viewed did worse than a generic brand ad, except for people whose browsing suggested they had already narrowed their choice, for example by reading review sites. Bleier and Eisenbeiss (2015) found a related pattern they call overpersonalization: highly personalized banners work best right after a visit and lose effect quickly, while moderately personalized ones last longer.

Why platform reports overstate retargeting

Retargeting targets people who were already likely to buy, so attributed conversions mix the ad’s effect with intent that was there before. Lewis and Rao (2015) call this selection bias “a crippling concern” for observational measurement, after 25 large experiments in which the median confidence interval on return was over 100 percentage points wide. At eBay, Blake, Nosko and Tadelis (2015) found brand keyword ads had no measurable short-term benefit, because those searchers would have arrived anyway. Gordon and colleagues (2019) showed that observational methods at Facebook often failed to recover the effects that experiments measured.

The fix is a control group. Ghost ads, the method Johnson and colleagues developed, logs the people in the control group who would have been shown your ad, then compares them with those who saw it. The difference is the lift.

Two bars side by side. The left bar, Saw the ad, is taller; its lower part matches the right bar, Would have seen the ad, and is labelled Baseline. The extra top part of the left bar is blue and labelled Lift. — Only the part above the control group's baseline is caused by the ads.

Google offers a version of this as Conversion Lift, though not for every account. A simpler route is to exclude a random 10 to 20 percent of each list yourself and compare the two groups after four weeks.

Frequency and timing

The effect of retargeting fades fast after a visit. Sahni and colleagues found that 33% of the first week’s effect came on the first day, and that the impact fell as time since the first visit grew. That argues for lists split by recency, with higher bids for people who left in the last few days.

A chart with Days since visit on the horizontal axis and Effect of one more ad on the vertical axis. A curve falls steeply after day 1 and then flattens; the area under the first days is blue. — Most of the effect comes in the first days after a visit, so recency should drive the bid.

Frequency needs a ceiling too. Google lets you cap impressions per person per day, week or month on Display and Video campaigns. The research does not give one right number. Manchanda and colleagues (2006) found more exposures raised repeat purchase among existing customers, while more distinct creatives had a negative effect. Test your cap the same way you test the campaign.

What privacy changes did to retargeting

Retargeting depends on recognizing the same person across sites and apps, and browsers and phones have made that harder. Safari’s Intelligent Tracking Prevention, launched in 2017, purges tracker cookies for domains a person has not interacted with in 30 days. Since iOS 14.5, apps must ask permission through App Tracking Transparency before tracking people across other companies’ apps and websites.

Aridor and colleagues measured the cost in Management Science: click-through rates on conversion-optimized Meta ads fell 37% after ATT, and firms more dependent on Meta lost 8% to 40% of revenue relative to less exposed firms, with smaller firms hit hardest. Chrome went the other way. In April 2025 Google said it would not roll out a standalone third-party cookie prompt, and in October 2025 it retired Protected Audience, the API Google had built for remarketing without third-party cookies, citing low adoption. First-party lists from logged-in customers and consented emails now carry more of the load.

Healthcare: where retargeting stops

Health is the category where retargeting is most restricted. Google’s personalized advertising policy says audiences an advertiser curates cannot be used when promoting in sensitive categories, and its health rules cover health conditions, chronic-condition treatments, intimate health, invasive procedures including cosmetic procedures and injections, and disabilities. Meta removed health-related detailed targeting options such as “Chemotherapy” in January 2022. Meta also sorts websites into data source categories and can block events such as Lead or Purchase from health sites, as Stape describes; check your category in Events Manager.

The legal risk is real in the US. In 2023 the FTC took action against GoodRx, which had uploaded lists of people who bought heart and blood pressure medication to Facebook to target ads. A federal court in Texas in 2024 vacated part of the HHS guidance on tracking pixels on public hospital pages, but other parts of that guidance and the FTC rules still apply.

For clinics, the workable version is narrow: general brand ads to all site visitors, search ads on treatment queries, and email to patients who consented. The safer plan is to treat retargeting as one measured channel inside the wider marketing funnel, with a holdout group from the first week, the way a Growth Lab test calendar treats any channel.

How to apply Retargeting, step by step

Map who left and where. List the steps on your site where people drop out: product page, cart, booking form, pricing page. Pull the monthly number of people at each step from analytics. Result: three to five candidate audiences with their size.
Build lists by intent and recency. Create one list per step and split each by time since the visit, for example 1 to 3 days, 4 to 14 days and 15 to 30 days. Exclude people who already bought, unless you sell something they will buy again. Result: lists sized and named so the bid can follow intent.
Match the message to the step. Cart abandoners can see the product they left. People who only read a blog post should see a broader brand or category ad, since research shows product-specific ads work best once people have narrowed their choice. Result: one ad set per list with a reason to come back.
Set caps and windows. Cap impressions per person per day or week, and set list duration close to your sales cycle. A food delivery app and a B2B software vendor should not use the same window. Result: a written frequency cap and membership duration for every list.
Hold out a control group. Keep 10 to 20 percent of each list out of the campaign, or run the platform's lift study, and compare conversions between exposed and held-out people over four weeks. Result: incremental conversions and cost per incremental conversion, next to the attributed numbers.
Cut, shift or scale. Drop lists where the holdout converts nearly as well as the exposed group. Move budget to the lists and windows with real lift. Result: a retargeting budget that pays for itself on incremental numbers.

Examples

An online retailer and ghost ads

Garrett Johnson, Randall Lewis and Elmar Nubbemeyer ran a display retargeting campaign for an online retailer using a method they call ghost ads: the platform logs the people in the control group who would have seen the ad, so they can be compared with the people who did. The ads lifted website visits by 17.2% and purchases by 10.5%, according to their 2017 paper in the Journal of Marketing Research. The lift was real, and measurable at a fraction of the cost of older test designs.

A travel site and dynamic ads

Anja Lambrecht and Catherine Tucker tested dynamic retargeting, ads that show the exact hotel a person viewed, against generic brand ads at an online travel firm. On average the dynamic ads were less effective. They stopped underperforming only for people whose browsing suggested they had narrowed their choice, such as visiting review sites. The lesson: product-specific ads suit people close to a decision, not everyone who visited once.

A dental clinic with restricted audiences

Illustrative, no real clinic implied. A clinic gets 3,000 site visits a month, 600 of them to the implants page. Under Google's personalized advertising policy, invasive procedures and treatments for health conditions are sensitive categories, so the clinic should not build an implants-visitor list for Google ads. It retargets all site visitors with a general brand ad about the clinic, runs search ads on implant queries, and follows up by email with patients who gave consent at booking.

When to use it

Use it when a meaningful share of visitors leaves at a step that signals intent, such as a cart, a booking form or a pricing page, and your traffic is large enough for lists above the platform minimum. It fits ecommerce, travel, subscriptions and any purchase people think about for days. It also helps when you can measure lift, either with a holdout or a platform lift study.

When not to use it

Skip it when traffic is so low that lists will not reach the minimum size, when most visitors come back on their own anyway (brand searches, logged-in customers), or when the product sits in a sensitive category such as health conditions, where platform policy bars audiences built from site behaviour. Do not judge it on platform-attributed conversions alone.

Common mistakes

Judging retargeting on view-through or last-click conversions reported by the platform, which credit the ad for purchases many people would have made anyway.
Using one 30-day list for everyone, so a person who left a cart an hour ago and one who read a blog post a month ago get the same bid and the same ad.
Showing the exact product to every visitor, when field experiments find product-specific ads work best for people who have already narrowed their choice.
No frequency cap and no buyer exclusion, so customers keep seeing ads for things they already bought.
Building audiences from health condition pages, which breaks Google and Meta policy and in the US has drawn FTC action against GoodRx.

FAQ

What is the difference between retargeting and remarketing?

In paid media the two words now mean the same thing: ads shown to people who already interacted with you. Yandex says so directly in its advertising course, and Google calls the feature remarketing, now labelled your data. Some marketers keep remarketing for email follow-ups to known customers, but platforms do not draw that line.

How do you set up retargeting in Yandex Direct?

Set goals in Yandex Metrica for the steps that matter, such as adding to cart or submitting a form. In Direct, create a retargeting and audience selection condition from those goals, Metrica segments or Yandex Audiences lists, and attach it to a campaign. Yandex also suggests a minus 100% bid adjustment to exclude recent buyers.

Does retargeting actually work?

Randomized field experiments say yes, with smaller effects than platform reports suggest. A retailer study found 10.5% more purchases, and a home-improvement seller saw 14.6% more visitors return within four weeks. Dynamic product ads did worse than generic ads on average in one travel study. Measure your own lift with a holdout group.

How long should a retargeting list keep people?

About as long as your sales cycle. Google defaults to 30 days and allows up to 540. Effects fade quickly: in one experiment a third of the first week's effect came on the first day after the visit. Split lists by recency so recent visitors get higher bids.

Can clinics and pharmacies use retargeting?

Only in narrow ways. Google treats health conditions, treatments and invasive procedures as sensitive categories where audiences built from your own site data cannot be used. Meta removed health-related targeting options in January 2022 and limits events from health websites. General brand ads, search ads and consented email are the safer routes.

Sources

Last updated Oct 9, 2026

Ilia PushinFounder, PUSHERS & COO Fintech ServiceIlia builds operating systems for growing companies in fintech and healthcare. Since 2021 he has run cross-border payments at ARBI Exchange, a licensed currency exchange in Thailand, including KYC and AML and the move into new jurisdictions.About the author LinkedIn

Related frameworks

More frameworks

Related insights

Three cylinders on a white marble slab, growing from short to tall: two in polished chrome, the tallest in frosted glass, on a light grey background.

Growth systemsArticle

Marketing agency or growth system: when a company outgrows its agency

An agency is the right call while the founder still runs most of the company. Once the business starts copying itself into new locations or markets, it needs a system that sets goals, tests hypotheses and owns the result.

Oct 7, 2026 · 7 min read

Three frosted-glass blocks rising in height like steps, with a small blue sphere on the lowest one, on a light grey background.

Growth systemsArticle

Why healthcare marketing agencies stall: the short, mid and long wins model we use with clinics

Most agencies own the click and nobody owns the front desk. How we structure clinic work in three horizons, the research behind each one, and one year of results.

Sep 26, 2026 · 9 min read

A glass arrow stopped against a white marble block, on a black background.

Growth systemsResearch

Why channel tests stall after month three

Most paid channel tests do not end with a decision. They end when the budget runs out, the attribution model changes, or everyone quietly stops looking at the dashboard.

Sep 23, 2026 · 6 min read

Want Retargeting running inside your company?Request an operations audit