Remarketing: what it is, how it works and how to measure it

You open your Meta Ads report and the remarketing campaign shows an 11x ROAS. It is the best line in the account by a wide margin. The uncomfortable question is a different one: how many of those sales would have happened anyway, with no ad at all?

Remarketing is one of the most widely used tactics in digital marketing and, at the same time, one of the most overstated in reports. It works. The catch is that it talks to people who were already close to buying, and the ad platform tends to take credit for a decision the customer had already made.

What remarketing is and how it works, in one paragraph: remarketing means showing ads to people who have already interacted with your brand, such as visiting your website, opening your app, watching a video or joining your customer list. A pixel or tag flags that visitor, they are added to an audience list, and they start seeing your ads on other sites, social networks and apps.

How remarketing works, step by step

Behind the ad that “follows” you after you look at a pair of sneakers sits a simple six-step mechanism.

1. The visit. Someone lands on a page of your site, a product page or the cart.
2. The tag. The Meta Pixel, the Google Ads tag (through GA4 or the Google tag) or the Conversions API records the event and ties it to an identifier.
3. The list. That identifier joins an audience with a rule you define, for example “viewed a product in the last 14 days and did not buy.”
4. The ad. When that person browses Instagram, YouTube or a display network site, the auction knows they are on your list.
5. The return. They click (or simply see the ad and come back later through search or by typing your URL).
6. The conversion, or not. If they buy, they leave the non-buyer list. If not, they keep being targeted until the membership window expires.

Two practical details change the outcome. The first is minimum size: according to Google’s official remarketing list documentation, ads only serve once a list has at least 100 active visitors in the last 30 days. The second is tagging quality. If the pixel drops events, the list ends up smaller and noisier than it should be, which is why it pays to understand the difference between the Conversions API and the Meta Pixel before blaming the campaign.

Are remarketing and retargeting the same thing?

In day-to-day practice, yes. “Remarketing” is the label Google popularized; “retargeting” is the term Meta and most of the industry use. Some marketers reserve “remarketing” for email re-engagement and “retargeting” for paid ads, but that split is not a standard and rarely changes a decision. What matters is where the audience comes from.

What types of remarketing are there?

Type	Who joins the list	When it makes sense
Standard (website)	Visitors to specific pages	Sites with decent traffic and a short journey
Dynamic (catalog)	Product viewers; the ad shows the item they saw	E-commerce with many SKUs
Customer list	Emails and phone numbers from your base (Customer Match, custom audiences)	Repeat purchases, upsell, win-back
Engagement and video	People who watched videos or engaged with your profile	Brands with strong content and little site traffic
App	Users who installed and stopped using the app	Subscription apps or marketplaces

Customer lists deserve special attention. They rely on data you collect with consent, not on cookies, so they are the part of remarketing least exposed to browser changes. If this is new to you, the post on first-party data explains why that base became an advertiser’s most valuable asset.

Does remarketing still work without third-party cookies?

It does, and the uncertainty has shrunk. In April 2025, Google announced on the official Privacy Sandbox blog that it will not roll out the standalone prompt that would have asked Chrome users to accept or reject third-party cookies. Chrome keeps its current controls in settings.

That doesn’t mean nothing changed. Safari and Firefox have blocked third-party cookies for years, iOS limits cross-app tracking, and privacy laws such as Brazil’s LGPD and Europe’s GDPR require a legal basis to process personal data. Anyone relying solely on a browser pixel is watching lists shrink. The fix has been the same everywhere: server-side events, logged-in users and customer lists.

Why remarketing reports almost always skew high

This is the part most guides skip. Remarketing targets people who have already shown intent. The Baymard Institute, averaging 50 studies in a list updated in September 2025, puts cart abandonment at roughly 70%. Some of those shoppers come back on their own, through search, through email or because payday arrived.

When they return and buy, the platform checks its attribution window, sees an ad click or view, and books the sale. The report can’t tell “bought because of the ad” from “bought and happened to see the ad first.” The result is an inflated ROAS, and it gets worse the wider your view-through window. Our post on conversion attribution breaks down how each model hands out that credit.

In practice, this leads to a very common mistake: the company sees remarketing with a sky-high ROAS, moves budget out of prospecting and into remarketing. A few months later the list shrinks (because nobody new enters at the top) and total sales drop, even though remarketing still “performs” on the dashboard.

How to measure whether remarketing drives extra sales

The right metric is incrementality: how many sales exist because the ad ran, compared with a world where it didn’t. The only reliable way to measure it is with a control group.

The holdout test in practice

You randomly carve out part of the list (10% to 20%) that gets no ads. Everyone else is served as usual. After a long enough period, you compare the conversion rate of both groups. Meta and Google offer conversion lift studies that run this split for you, with minimum spend requirements; without them, you can build a homemade version by excluding a segment of your customer list.

Here is a simulation with round numbers. A fashion retailer has 20,000 people on its “viewed a product, didn’t buy” list and spends $9,000 a month.

• Exposed group: 18,000 people, 540 purchases, a 3.0% rate.
• Control group: 2,000 people, 50 purchases, a 2.5% rate.
• Incremental sales: (3.0% − 2.5%) × 18,000 = 90 purchases.
• Lift: 0.5 ÷ 2.5 = 20%.

Now the shock. Using a 7-day click and 1-day view window, the platform attributed 420 purchases to the campaign. With a $250 average order value, the dashboard shows an 11.7 ROAS (420 × 250 ÷ 9,000). The incremental ROAS is 2.5 (90 × 250 ÷ 9,000). CPA jumps from $21 to $100.

Note that 2.5 can still be a good result, depending on your margin. Your break-even ROAS is what tells you whether the campaign pays for itself. The question was never “does remarketing work?” but “what is it really worth, and how much should I put into it?”

When you don’t have budget for a formal test

You can get close with three cheaper signals:

• Compare windows. Check how many conversions come from views alone (no click). If they are the majority, be suspicious.
• Pause by region. Turn remarketing off in two or three states or regions for four weeks and compare total sales against regions where it kept running.
• Look at time to purchase. If most attributed conversions happen within 24 hours of the first visit, many of them were probably already decided.

How to run remarketing without wasting budget

A few rules that, in our experience, separate profitable campaigns from ones that only look profitable:

• Segment by intent, not just by visit. Someone who abandoned a $600 cart deserves a different ad (and a different bid) than someone who glanced at the homepage for five seconds.
• Exclude recent buyers. Sounds obvious, yet it is the most frequent mistake in audited accounts. Showing the same sofa to someone who just bought it wastes money and annoys the customer.
• Cap frequency. Above 8 to 10 impressions per person per week, returns usually fall and brand fatigue rises. The number varies by industry, so test yours.
• Use short windows for quick-decision products. For food delivery or fashion, 7 to 14 days. For real estate, courses or B2B, 30 to 90 days makes more sense.
• Don’t let remarketing eat prospecting. A common rule of thumb keeps remarketing between 10% and 30% of media spend. Above that, check whether the list is still growing.

Frequently asked questions

What is the difference between remarketing and retargeting?

In practice, nothing meaningful. Both describe ads shown to people who already interacted with the brand. “Remarketing” is Google’s term; “retargeting” is more common at Meta and across the industry.

Is remarketing legal under privacy laws like GDPR and LGPD?

Yes, as long as there is a legal basis for processing the data, usually consent collected through the cookie banner, and your privacy policy discloses the use. Customer lists require that emails and phone numbers were collected legitimately.

How long does someone stay on a remarketing list?

You decide. In Google Ads and Meta, the window is set when you create the audience. The ideal length follows your purchase cycle: short for cheap products, long for considered purchases.

What is the minimum audience size for remarketing?

In Google Ads, the official documentation requires at least 100 active visitors in the last 30 days for ads to serve. With very small lists, cost per impression rises and the algorithm learns slowly.

How do I know if my remarketing is working?

Don’t rely on dashboard ROAS alone. Run a control-group test (holdout or conversion lift) and compare the conversion rate of people who saw the ad with those who didn’t. The gap is the real result.

The remarketing worth paying for is the one you can prove

Well-run remarketing is profitable, and nobody should switch it off out of fear. But it has to be judged by what it adds, not by what the platform manages to credit to itself. That means joining media, website and actual sales data, which usually live in three different tools. With Sherlok, you connect Meta Ads, Google Ads, GA4 and your CRM and ask plain-language questions like “how much of remarketing revenue came from people who already had an open cart?” The answer arrives ready, with no spreadsheet in between.