Google Ads Remarketing: Types, Setup and Audiences

Посібник з ремаркетингу

Remarketing means showing ads to people who have already been on your site, watched your videos or given you their contact details. The audience is warm, so conversion rates are usually higher and the cost per lead lower than in campaigns aimed at cold traffic.

This article was first published in 2023 and described building audiences in the old Google Analytics, which no longer exists. We have rewritten it from scratch and expanded it: it is now a full breakdown of every type of remarketing in Google Ads — dynamic, search, video, app and customer data — with current audience limits and what to use in place of similar audiences.


1. Remarketing and retargeting: is there a difference

In practice, no. "Remarketing" is Google’s term; "retargeting" came from the rest of the ad platforms. Both mean showing ads again to people who have already interacted with the business.

What makes remarketing different from ordinary advertising is not the ad format but who you choose:

  • in search, you target the query;
  • in display, interests and topics;
  • in remarketing, an action the person did — or failed to do — on your site.

That is why remarketing does not work without two things: audience collection that is properly set up, and traffic for those audiences to accumulate from. If 200 people visit the site a month, remarketing will not produce meaningful volume — you need traffic first.


2. How it works technically

Three components:

  1. the Google Ads tag (or the Google tag via GTM) on every page of the site. It records visits and builds segments;
  2. audiences — in Google Ads under Audience Manager, or in GA4. In the second case the GA4 property has to be linked to the Google Ads account, and the audience itself has to be available for export, otherwise it will not appear in Google Ads. Universal Analytics audiences are gone along with the service itself, so any instruction that mentions "Audience Definitions" in the Analytics admin is out of date;
  3. user consent. For visitors from the EEA and the UK you need Consent Mode v2, otherwise the segments stay empty.

Timing matters too. A new list takes 48–72 hours to populate, so creating audiences the day before a campaign launch is pointless. And a segment that has not been used for targeting for 540 days is closed automatically by Google.

As for cookies: in April 2025 Google dropped its plans to remove third-party cookies from Chrome and kept the current approach. In practice remarketing works as it did before, but it is worth building on first-party data — tags, conversions, customer lists — because that does not depend on browser decisions.


3. Types of remarketing in Google Ads

Type Who you reach Where it shows
Standard Site visitors — generic banners and responsive ads Display Network, YouTube, Gmail
Dynamic The specific products a person viewed Display, Performance Max, Demand Gen
Search (RLSA) Previous visitors, when they search on Google again Search and search partners
Video Channel and individual video viewers, subscribers YouTube and video inventory
App App users, based on their in-app actions Apps and mobile sites
Customer data (Customer Match) People from your own contact database Search, Shopping tab, YouTube, Gmail, Display

 

Below, what actually affects results in each of them.


4. Dynamic remarketing

This is the format where a person sees the exact product they looked at. For a store it delivers more than any other kind of remarketing, because you do not have to guess the message — the ad is assembled from the feed.

What you need:

  • a feed. For e-commerce, a product feed in Merchant Center. For everything else (real estate, travel, cars, education) — a feed in the business data section;
  • a tag with parameters. The plain tag is not enough: along with the event it has to pass the product identifier and the business type. If the ID does not match the ID in the feed, the ad cannot be assembled — this is the most common reason dynamic remarketing "doesn’t work";
  • current feed data — price, availability, links, images. Merchant Center suspensions and feed problems stop more sales than bidding does.

The parameters themselves deserve a separate note, because there are a lot of outdated instructions around. The old set along the lines of ecomm_prodid and ecomm_pagetype is no longer described in Google’s current documentation: data is now passed through Google tag events with parameters.

What you pass How it works now
Product identifier The id parameter; the value has to match the Merchant Center feed ID exactly
Business type The google_business_vertical parameter: retail, flights, hotel_rental, education, real_estate, custom
Stage of interaction Not a separate parameter but standard events: view_item, add_to_cart, purchase and others
Vertical specifics For travel, destination and dates; for education, location; for real estate, the listing ID

 

Where to run it:

  • a Display campaign with a feed — the classic option, with control over audiences and placements;
  • Performance Max — uses the feed automatically, but you cannot separate remarketing from new customer acquisition into different campaigns inside it;
  • Demand Gen — works with a product feed, and business data feeds are supported for non-retail verticals.

Shopping and Performance Max setup is part of our Google Shopping and Performance Max services.


5. Search remarketing (RLSA)

Here ads appear in search results rather than across the network — but only to people who have already been on your site. The logic is different: the person types the query themselves, so you are not chasing them, you are winning the auction a second time.

Two typical strategies:

  • higher bids for existing audiences inside your normal campaigns — the same keywords, but more expensive for people who already know you;
  • separate campaigns with broad queries that are too expensive to buy for cold traffic. For someone who has already viewed the product, that query behaves differently.

In practice RLSA is also the cheapest way to test broad and generic queries: the audience filter keeps the waste down while you find out which of them convert at all.


6. Video and YouTube remarketing

If the channel is linked to the Google Ads account, audiences are collected automatically. You can segment by action:

  • viewed any video on the channel, or specific videos;
  • viewed a video as an ad;
  • subscribed to the channel;
  • liked, added to "Watch later", shared;
  • visited the channel page.

This is the only kind of remarketing that works with people who have never been on your site. A practical scenario for services: explain the offer in a video, collect the people who watched it through, then bring them back with search or display ads carrying a specific offer.

A separate case is app remarketing. Audiences here are built not on site visits but on in-app events: install, sign-up, add to cart, purchase. It requires in-app conversion tracking via an SDK or Firebase, and ads show in other apps and on mobile sites. The logic is the same: one segment for people who installed but never registered, another for people who bought a long time ago.


7. First-party data: Customer Match

You upload a contact database (emails, phone numbers, addresses), Google matches it against user accounts and shows ads to the people it can find.

  • available in Search, the Shopping tab, YouTube, Gmail and Display;
  • for the list to stay active it has to contain at least 100 contacts added or updated within the last 540 days;
  • maximum membership duration is 540 days, so the database needs regular updating — ideally a CRM sync;
  • for EEA users you need valid consent signals;
  • the match rate is never 100%: some contacts simply will not be found.

The most useful scenarios are not "sell to everyone again" but precise ones: repeat purchases timed to the consumption cycle, upsells to what was already bought, excluding existing customers from acquisition campaigns, and separate messaging for people who have not bought in a long time.


8. How to build segments

You need the basic "all visitors" list, but on its own it produces the most expensive impressions: it includes everyone who left after three seconds. What works in practice is splitting by depth of interest:

  • by page: category, specific product, pricing page, delivery;
  • by stage: added to cart, started checkout, never reached the thank-you page;
  • by behaviour: time on site, number of pages, repeat visits;
  • by conversion: people who already bought — for upsells and for exclusions;
  • combinations: reached the checkout page but not the thank-you page. This is the classic one, and the cheapest conversions you will get.

On exclusions specifically: audiences of existing customers, wholesale buyers, job applicants and contractors should be excluded from campaigns. Without that, part of the budget goes on people you already have.

Analytics has to be in order before any of this: key events configured, not duplicated, and passing value. That is part of our web analytics service.


9. Limits: audience size, duration, frequency

Parameter Value
Minimum audience size 100 active users in the last 30 days — for Display, Search and YouTube. Search and YouTube used to require 1,000; the threshold was lowered
Maximum membership duration 540 days
Time for a new list to populate 48–72 hours
Unused segment Closed automatically after 540 days without targeting
Google-provided segments A user inactive for more than 18 months is removed from the segment

 

Membership duration should match the purchase cycle: for food delivery 7 days is a lot, for real estate 540 is not much. A simple benchmark: take the average time from first visit to order and double it.

Frequency capping is set at campaign level and exists so you do not wear the audience out. The symptom you can see in the reports: plenty of impressions, almost no clicks, and a small audience — people have already seen the ad dozens of times. In campaigns using Smart Bidding frequency is managed by the algorithm, but in Display campaigns it is worth keeping under control.


10. What replaced similar audiences

Google removed similar audiences in 2023: from 1 May they could no longer be created, from 1 August they were removed from campaigns. What to use instead:

  • Lookalike segments in Demand Gen. You provide a seed list and the system finds similar users. One important detail: since March 2026 the reach setting — narrow, balanced, broad — works as a signal for the algorithm rather than a hard similarity limit. Which means the main lever you control is now the quality of the seed list;
  • optimized targeting in Display and Demand Gen campaigns: the system goes beyond the audiences you specified, guided by conversions. Your segments work as a signal, not a hard filter;
  • audience signals in Performance Max. Remarketing lists there are not targeting but a hint to the algorithm about where to look first;
  • Customer Match as the foundation: a buyer list is the best seed segment both for Lookalike and for exclusions.

11. Optimisation and measurement

What actually changes results:

  • separate ads per segment. Someone who abandoned a cart and someone who was reading the blog need different messages;
  • the page you bring them back to. If the ad promises a discount, the discount has to be on the page. Otherwise the person leaves a second time, and for good;
  • checking overlaps. One user can sit in three lists at once — then your campaigns compete with each other for them;
  • excluding converters from acquisition campaigns;
  • bids. In remarketing a higher bid than in cold campaigns is often worth it, because warm audiences convert better. How that works with the strategies themselves: Google Ads bidding strategies.

How to evaluate it:

  • look at the number and cost of key events rather than CTR;
  • remember that remarketing often closes a conversion another channel started: conversion paths are more informative than last click;
  • compare periods with remarketing on and off — the simplest way to see its real contribution;
  • plan seasonal peaks in advance: lists are built before the season, not during it — see how to prepare for the sales season.

12. Common mistakes

  • One "all visitors" list for every campaign. Expensive and imprecise;
  • Default membership duration with no regard for the purchase cycle;
  • No exclusions: ads chase people who have already bought;
  • The tag ID does not match the feed ID — dynamic ads cannot be assembled;
  • Remarketing instead of acquisition. Warm audiences are always finite: without new traffic they run out;
  • No frequency cap in Display campaigns;
  • Consent not configured for EEA traffic — the lists stay empty while you look for the cause in the campaigns;
  • Judging by impressions and CTR instead of leads and revenue.

13. In short

  • Remarketing and retargeting are the same thing; the difference is only in the platforms’ terminology.
  • Audiences are now built in Google Ads and GA4; they no longer exist in the old Analytics.
  • The minimum for targeting is 100 active users over 30 days, maximum membership duration is 540 days, and a new list takes 48–72 hours to populate.
  • Dynamic remarketing needs a feed and a tag with parameters; mismatched IDs are the main reason it fails.
  • Similar audiences have been gone since 2023; in their place, Lookalike in Demand Gen, optimized targeting and audience signals.
  • Without exclusions and without new traffic, remarketing runs out quickly.

If you need remarketing set up, or want to find out why existing campaigns are not producing leads, that is part of what we do — see Google Ads setup.

Get in touch!

Get More Value

You will get the best content from us to help your business grow.
Your request has been sent.

    Subscribe

    Advance with us — apply now

    Get a free consultation and business promotion assessment
    Your request has been sent.

      Send