Retargeting is advertising to people who have already interacted with your business. It sounds like the safest money in advertising, and that is exactly why it is so often overrated. This article explains what a retargeting audience is built from, when a separate retargeting layer is justified, when it is redundant, and why its numbers look better than they are.
What retargeting is, and why it is no longer the cheap win
Retargeting means showing ads to people who have already been on your website, watched a video, engaged with your Facebook Page or Instagram account, or are on your customer list. The logic is easy to follow: these people already know you, so persuading them should be easier and cheaper.
Two things complicate that logic today. First, in Meta’s audience model an included audience is a signal, not a boundary. If you put a retargeting audience into an ordinary ad set, the system starts there but expands to other people as well. Pure retargeting needs a different setting, which we come to later in the article. How the audience model works as a whole is set out in our article on Facebook ads audiences.
Second, retargeting audiences have become smaller, because a share of visitors do not give cookie consent and a share of browser events never reach Meta. An audience that was large enough for a separate ad set a few years ago often is not any more. At the scale of Latvia, where the total number of visitors is small to begin with, that is the rule rather than the exception.
What a retargeting audience is built from: sources and prerequisites
A retargeting audience is a custom audience, and custom audiences have different sources with different prerequisites. The essential distinction is between sources that need tracking on your website and sources that work entirely inside Meta’s platform.
| Source | What it needs | What it gives you, and what it does not |
|---|---|---|
| Website visitors | A working Pixel or Conversions API and the visitor’s consent | The most precise intent signal (specific pages, basket, events), but it sees only those who consented and whose events arrived |
| Customer list | Your own data (emails, phone numbers) that Meta matches to profiles | Real buyers, independent of tracking, but only part of the list will match |
| Video viewers | Nothing outside Meta | A broad and stable audience, but a view is a weak intent signal |
| Instagram profile and Facebook Page engagement | Nothing outside Meta | Stable and independent of consent, but engagement is not purchase intent |
| Lead forms | Nothing outside Meta | People who opened or submitted a form; narrow but strong |
The first practical conclusion follows from the table: if your website tracking does not work, website retargeting is impossible no matter how well everything else is set up. How to check that tracking works is covered in our article on the Meta Pixel and Conversions API. The second conclusion: Meta’s own sources work even when website data is incomplete, which often makes them a more stable base than a visitor audience.
Every audience has a time window: how long a person stays in it after the action. For website audiences it is usually up to 180 days, and the exact window is shown when you create the audience. A short window gives a warmer but smaller audience; a long window gives a larger one with weaker intent.
If no single source is large enough on its own, sources can be combined into one warm audience: website visitors together with video viewers and profile engagement. That is exactly where practitioner advice has moved in 2026: fewer tiny retargeting audiences, more combined warm pools in which the system has something to learn from.
When retargeting is justified: a use-case tree
Retargeting is not one tactic but several, each with a different goal and a different message. If you show warm audiences the same ad as cold ones, retargeting loses most of its point.
| Situation | Goal | Message | Minimum condition |
|---|---|---|---|
| Added to basket, did not buy | Complete the purchase | Remove a specific obstacle: delivery, returns, payment | A working purchase event and enough basket adds |
| Viewed the offer, took no action | Bring them back to the decision | A concrete reason to act now, not a repeated introduction | Website tracking and enough visitors |
| Existing customers | Repeat or additional purchase | The next step, not the first; no introduction | A customer list or purchase events |
| Watched a video, never visited the site | Move them to the website or a form | A concrete offer for people who have already seen the story | Enough viewers; tracking not required |
The unifying principle: a retargeting ad speaks to someone who already knows you, so there is no room in it for an introduction. There is room for the next step and for removing a specific obstacle.
When retargeting is not worth it
This is the section most retargeting articles do not have, and it is the one that saves the most money.
The audience is too small. In ACCESSIO’s practice a separate retargeting ad set is not created when the audience looks too narrow for the system to optimise within it properly, and especially when the product is not expensive: for an inexpensive product with a small flow of visitors, a retargeting ad set can spend more on learning than it brings in. Independent sources confirm it: finely split warm audiences do not leave the learning phase, and the advice is to combine them into larger pools, not to split them further.
The data has not accumulated yet. Retargeting is an accumulation tactic: it needs prospecting to have been delivering visitors and events for a while. In ACCESSIO’s practice a simpler structure, one broad ad set with existing buyers excluded, is the better choice until nuanced retargeting data has built up over a longer period. Adding a separate retargeting layer in the first month is usually premature.
The buying cycle is short. If a person decides within hours, they have either bought already or already decided not to. Retargeting a week later only annoys them.
The system is reaching warm people anyway. In a broad ad set with Advantage+, the system often reaches people who have already engaged by itself, because they are precisely the ones with the highest predicted likelihood of acting. A separate retargeting ad set then competes with your own broad ad set for the same people.
Frequency is already high. If the warm audience sees the ad many times a week, an extra retargeting layer only increases that. An ad someone sees for the tenth time is no more persuasive than the fifth.
How to set it up so it really reaches only them
If the decision goes in favour of a separate retargeting layer, three things determine whether it is actually retargeting.
First: the audience goes in as a boundary, not a signal. A custom audience included in an ordinary ad set is only a suggestion to the system. For only those people to see the ad, the ad set needs the “Further limit your audience” button, which switches to a setup with stricter boundaries. Without it, a “retargeting campaign” is a broad campaign with a warm start.
Second: exclusions. People who have already bought must be excluded from purchase retargeting, otherwise you are paying to show ads to people who no longer need them. And if there are several retargeting layers, each one excludes the people further down: basket abandoners do not see the visitor ad, buyers see neither. Exclusion is a Control, so it always holds.
Third: a different message. A retargeting ad that is the same ad shown again is the weakest form of retargeting. The message matches the situation in the table above: obstacle removal for basket abandoners, a concrete reason to act for those who viewed, the next step for customers.
And a fourth, practical one: watch Frequency. A retargeting audience is small, so even a modest budget saturates it quickly. If one person is seeing the ad several times a day, either reduce the budget, widen the time window or change the creative; continuing to show the same ad even more often is the most expensive of all the options.
What retargeting does not prove
In platform reports retargeting almost always looks like the best campaign: a lower cost per result, a higher ROAS. Part of that is real, but part is a feature of the measurement that needs to be understood before decisions are made.
A retargeting audience is, by definition, people who were already close to buying. Some of them would have bought without the ad. The platform credits those purchases to the ad because the person saw it, and as a result retargeting gets credit for sales it did not cause but merely accompanied. That is attribution, not causation, and we set out that boundary in detail in our article on Meta Ads metrics.
The practical test is simple, if not convenient: pause retargeting for a while and watch your total sales data, not the platform figures. If total revenue barely moves, the retargeting was mostly claiming credit. If it falls, it was genuinely working. Large organisations do this with controlled holdout tests; for a small business, pausing and taking an honest look at the total figure is enough.
And one more thing retargeting does not prove: that the problem was the audience. If people were on your website and did not act, the offer or the page is more often to blame than the fact that they had not yet seen the ad a second time.
A checklist before creating a separate retargeting layer
Before you build a separate retargeting ad set, answer these questions. If the answer to any of them is no, solve that first.
- Does tracking work, and does the event you will build the audience from actually arrive in Events Manager?
- Is the audience large enough for the system to optimise within it, not merely to deliver?
- Is prospecting replenishing this audience regularly, or is it a one-off accumulation?
- Does the retargeting ad have a different message from the main ad?
- Are existing buyers and the later stages of the funnel excluded?
- Is the audience set as a boundary, not a signal?
- Do you know which total sales figures you will use to judge whether retargeting added anything?
In summary
Retargeting is justified when tracking works, the audience is large enough and regularly replenished, the message differs from the cold ad, and there is a clear way to judge it by total data. If any of those is missing, a simpler structure, one broad ad set with buyers excluded, is usually better. And platform figures always flatter retargeting, so they are the start of the assessment, not the end of it.
If you would like an experienced provider to take on retargeting decisions and the management of your advertising, see ACCESSIO’s Facebook and Instagram ads management service.
