Remarketing is advertising shown to people who have already been on your website. Someone opens a page, doesn't buy and leaves: their browser goes into a list, and from that moment they can see your ads again on Google, on YouTube or on one of the sites in the Display Network.
Everyone covers that part. What most guides leave out is the maths: how many visitors you need before a remarketing list can actually serve ads, and what happened to third-party cookies between 2025 and 2026, given that plenty of articles still take it for granted that they have gone.
How it works, technically
Three steps, always the same.
- There is a tag on the site (the Google tag, the Meta pixel or the LinkedIn Insight Tag) that records who visits which pages.
- The tag feeds a segment: "people who viewed the pricing page in the last 30 days", "people who added to basket and didn't pay", "people who watched 50% of a YouTube video".
- That segment becomes the audience of a campaign, which stops looking for new people and simply shows up again in front of those.
That is the whole difference from a normal campaign: you are not buying attention from scratch, you are buying a second pass. This is why the cost per click of a remarketing campaign is almost always lower than that of a search campaign on the same keyword.
What really happened to third-party cookies
Some guides published in 2026 open by saying that "the phase-out of third-party cookies in Chrome was completed in 2025" and that remarketing therefore has to be rebuilt on first-party data. That is not what happened. Chrome still has third-party cookies, and what was shut down was the project meant to replace them.
| Date | Google announcement |
|---|---|
| 22 July 2024 | Google gives up on removing third-party cookies. In their place it announces "a new experience in Chrome" that lets users choose. |
| 22 April 2025 | That choice is dropped too: Chrome keeps its current setting, with no dedicated prompt. Third-party cookies stay where they are. |
| 17 October 2025 | Google retires ten Privacy Sandbox technologies, including Protected Audience, Topics and Attribution Reporting. Stated reason: low adoption. |
| Chrome 144 (December 2025) | The APIs are deprecated in the browser. |
| Chrome 153 | Removal of the Protected Audience code, postponed twice from Chrome 150. |
Protected Audience deserves a closer look. It was the API designed to do exactly this, remarketing without cookies: the browser kept interest-group membership locally and the auction ran on the device. Anthony Chavez, who leads Privacy Sandbox at Google, announced its retirement in October 2025. It was shut down before the cookies it was meant to replace.
The practical point for anyone running campaigns in the UK: in 2026 remarketing on Google Ads and Meta works as it always has, and the tags on your site are first-party anyway. What limits you is not the browser. It is consent.
The real constraint in the UK is consent
Under PECR, advertising and tracking cookies need consent, and the Data (Use and Access) Act 2025 did not change that: its new exceptions cover things like statistics, and the ICO guidance is explicit that the statistical exception cannot be used for online advertising. The same guidance says the consent mechanism must make refusing as easy as accepting, that a user who moves on without clicking any option has not consented, and that six months is a suitable interval before asking again.
On Google's side, since March 2024 two explicit parameters are required in the EEA and the UK before a person can be used for ad personalisation: ad_user_data and ad_personalization, both set to granted. If they are denied or not specified, that person does not enter your lists, as Google states in the Customer Match documentation.
How many people say yes? The 2026 Didomi benchmark, built on hundreds of millions of banner interactions collected in 2025, puts the British Isles at 87.3%, against 75.1% for Western Europe and 89.3% for Eastern Europe. Roughly one visitor in eight, then, will never enter a remarketing list however well you have installed the tag.
How many visitors you need before the list can serve
Google sets a technical threshold: a segment becomes eligible to serve on the Search Network, Display and YouTube only if it contains at least 100 active visitors or users in the last 30 days. The threshold was lowered from 1,000 to 100 for lists uploaded or updated from 1 February 2024, and it is still the number that decides whether your campaign runs or sits there with the message "audience too small".
Put the two together: to get 100 people into a list in the UK you need about 115 visitors in thirty days on that page. In the table below I have done the maths on real UK cost-per-click figures for six sectors. It only counts paid traffic, so it is a floor: organic, direct and social traffic push the number up.
| Sector and keyword | CPC | Monthly budget | Clicks | In the list after consent | List can serve |
|---|---|---|---|---|---|
| Restaurant, "restaurants london" | £0.84 | £250 | 298 | 260 | Yes, with room to segment |
| Estate agent, "house valuation" | £3.50 | £500 | 143 | 125 | Yes, just |
| Catering, "catering london" | £2.81 | £500 | 178 | 155 | Yes, but a single list |
| Dental practice, "dentists london" | £7.98 | £500 | 63 | 55 | No |
| Solicitor, "divorce solicitor london" | £35.90 | £750 | 21 | 18 | No |
| Solicitor, "probate solicitor london" | £19.13 | £750 | 39 | 34 | No |
The rows to look at are the last two. Same London law firm, same budget: neither list reaches the threshold, and the probate list fills almost twice as fast as the divorce list because the click costs about half as much. To get 115 paid visitors a month on "divorce solicitor london" you would need around £4,130; on probate, around £2,200. The arithmetic changes, not the strategy, and this explains why the same advice ("do remarketing") works very well for one business and never gets going for the one next door.
If the maths gives you fewer than 115 visitors a month on the page you care about, the problem is not remarketing. It is the traffic upstream, and it needs solving first, through organic search or by widening the campaigns. We went through thresholds and minimum budgets in detail in when Google Ads is worth it.
The four list types in Google Ads
Google calls them data segments, and it helps to keep them separate because they behave differently.
- Website visitors. Built by the tag. They are the quickest to set up and the most exposed to consent: if the banner is rejected, that visit doesn't exist.
- App users. Same logic, from the SDK.
- Customer Match. You upload your contacts' email addresses or phone numbers, already hashed. It is the only list that does not depend on who visits the site this month, and for a business with a well-populated CRM it is often the biggest one it has.
- People who interacted with your videos. Views, channel subscriptions, clicks on YouTube cards.
Where these lists end up makes a big difference to the result. On the Search Network they become RLSA: they don't show a new ad, they raise the bid when someone who has already visited the site searches again. It is the most underrated format and, for service businesses, the one with the best return. On Display and Demand Gen they become campaigns in their own right. In Google Ads for ecommerce we showed how, with a Merchant Center feed connected, dynamic remarketing rebuilds the ad with the exact product the person had looked at. In Performance Max, on the other hand, the list is a signal rather than targeting: it tells the algorithm where to look, but Google is free to go elsewhere.
How long to keep someone in a list
The technical ceiling is 540 days: past that point Google removes the person from the segment, even if you set a longer duration. Below that ceiling it is your call, and the right variable is not the sector but the time between first contact and decision.
An abandoned basket on a clothing ecommerce site is recovered within a few days or not at all. A house valuation is different: in our analysis of Google Ads for estate agents, the owner who opens the valuation page and doesn't fill in the form often decides weeks later, and a 7-day window loses them along the way. The same goes for professional advice, where more than a month passes between the first search and the phone call.
Rule of thumb: start from the average gap between first visit and conversion that you see in Google Analytics, and set the duration to about double that. Then look at the click-through rate by list age: where it collapses is where to cut.
Sectors where remarketing is prohibited
Few guides cover this part, and it costs suspended accounts. Google's personalised advertising policies prohibit advertiser-created audiences on a specific list of categories: health and contraception, negative financial status, political affiliation, religious belief, sexual orientation, identification as transgender, ethnicity, trade union membership, alcohol, gambling and games, abuse and trauma, relationship difficulties, commission of a crime.
In terms of professions: a psychology practice cannot build an audience from visitors to its anxiety page, as we explained in Google Ads for psychologists. A multi-specialty clinic cannot create lists on its treatment pages, a case covered in Google Ads for medical practices and clinics. A dental practice is outside the ban on its general pages and falls within it for implants and other invasive procedures.
What stays allowed in these cases is Google's predefined segments, which the advertiser does not build, and remarketing on the neutral pages of the site: home, contact, about us.
How to tell whether it is really driving conversions
Remarketing has a measurement problem no other campaign has: by definition it targets people who already know you, so some of those conversions would have happened anyway. Looking at the campaign ROAS and declaring it a winner is the fastest way to overestimate it.
Two corrections, in order of difficulty. The first: view-through conversions, those credited to people who saw the ad without clicking it, have a default window of 30 days. Bring it down to one day, remove that column from your headline numbers and see what is left.
The second is an incrementality test: switch remarketing off in some regions for four weeks, keep everything else unchanged and compare total conversions (not those credited to the campaign) between the regions switched off and those left on. It is awkward, it is the only honest answer, and in our experience on Google Ads management projects it is also the moment some Display campaigns stop looking profitable.
Mistakes that burn budget
- Segmenting too early. Six lists of 40 people serve nobody, because none reaches 100. With fewer than 500 visitors a month, keep a single list.
- Not excluding people who have already bought. Without an explicit exclusion, the customer who has just paid keeps seeing the ad for the product already in their home.
- Sending everyone to the homepage. Someone who viewed a product page should go back to that page. We wrote a dedicated guide on landing pages.
- No frequency cap. On Display the cap is set at campaign level and it needs setting: without it, the same person can see the ad dozens of times in a week.
- Using remarketing to paper over a site that doesn't convert. If the page doesn't work on the first visit, it won't work on the third. We listed the most common causes in why your website doesn't convert.
One last piece of context on the UK market, useful for setting expectations: according to the Office for National Statistics, internet sales accounted for 27.3% of all UK retail sales in August 2026, and the share peaked at 32.4% in November 2025. The audience is there. The bottleneck is how much of it reaches your site and accepts the banner.
Remarketing FAQs
None of substance. Remarketing is the term Google uses in its documentation, retargeting the one common across the rest of the advertising market. Originally remarketing also covered re-engagement by email, but today the two terms are used as synonyms.
The question starts from a wrong assumption. On 22 April 2025 Google announced that Chrome will not remove third-party cookies, and on 17 October 2025 it retired the Privacy Sandbox technologies meant to replace them, Protected Audience included. In 2026 remarketing on Google Ads and Meta works as before: the real limit in the UK is cookie consent, not the browser.
Google requires at least 100 active visitors or users in the last 30 days for a segment to serve on Search, Display and YouTube. Given that about 87.3% of users in the British Isles accept cookies (Didomi benchmark on 2025 data), in the UK you need about 115 visitors in thirty days on that page.
The maximum Google Ads allows is 540 days. Beyond that point the person is removed from the segment automatically, even if you have set a longer duration. The right duration comes from the time between first visit and conversion on your site.
Google's personalised advertising policies prohibit advertiser-created audiences based on health, negative financial status, political affiliation, religious belief, sexual orientation, transgender identity, ethnicity, trade union membership, alcohol, gambling and games, abuse and trauma, relationship difficulties and commission of a crime. In practice this affects medical practices, psychologists and some dental treatments.
It depends on the sector's cost per click. With £250 a month and an £0.84 click (restaurants london) the list reaches about 260 people and remarketing makes sense; with £750 a month and a £35.90 click (divorce solicitor london) it does not even reach Google's minimum threshold. On a budget of a few hundred pounds a month, that money usually does more concentrated on a single search campaign.
Something the industry rarely says: remarketing is not a channel, it is a multiplier of what is already coming in. A search campaign brings new people, remarketing brings them back round. If the first part is small, multiplying it doesn't change the order of magnitude of the result, and budget spent on reaching 80 people again is budget taken away from bringing in 300. So when the question comes up, "when is it worth starting a remarketing campaign", the honest answer isn't about remarketing: it is about how much traffic you are already generating. The cases where we did this maths before starting are collected in our projects, and the decision not to switch it on came more often than you would expect.
