Targeting is the part of online advertising that sparks the most debate in class: should you target narrowly or let the algorithm decide? The answer depends on the objective, the budget and the data you have. This article gives you the reference points to reason it through, with exercises to redo in a simulator.
Why targeting matters
A perfect ad shown to the wrong people sells nothing. Conversely, an average ad shown to a relevant audience can be enough. Targeting influences three things:
- cost: an audience that many other advertisers want costs more to reach;
- relevance: a well-matched message earns more clicks and conversions;
- learning speed: an audience that is too small prevents the algorithm from gathering enough data.
The three main audience families
1. Core audience (or “cold” audience)
You define it with criteria:
- location: country, region, city, radius around an address;
- age and gender;
- interests: what people follow, like or browse;
- behaviors: for example online shoppers or frequent travelers.
This is the classic entry point for a student. It suits introducing an offer to people who do not know you yet. See Set up location and interest targeting.
2. Custom audience (or “warm” audience)
It is built from your own data:
- website visitors (thanks to the pixel);
- people who engaged with your Page or Instagram account;
- video viewers;
- an uploaded customer list.
These people already know you: they usually convert better and at lower cost. This is the basis of retargeting. See Create retargeting audiences and Import a customer list.
3. Lookalike audience
From a source audience (your best customers, for example), Meta finds people who resemble them. The higher the quality of the source, the better the lookalike. A source of a few dozen people gives fragile results; a source of several hundred or thousands of real customers is far sturdier. See Create lookalike audiences.
| Family | Starting point | Typical use |
|---|---|---|
| Core | Demographic criteria and interests | Discovery, launch |
| Custom | Your own data | Re-engagement, loyalty |
| Lookalike | A sample of your best customers | Acquisition at scale |
Broad or precise: which to choose?
Platforms now encourage broad audiences, because their algorithms can find the right people from creative and conversion signals. But “broad” does not mean “without thought.”
Broad targeting works best when:
- conversion tracking is properly installed;
- the budget produces enough results per week;
- the creative clearly speaks to one audience (the message itself does the sorting).
Precise targeting is still useful when:
- the offer is very specific (a trade, a rare practice);
- the territory is limited (a neighborhood shop);
- conversion data is missing, as at the very start.
Audience size
An audience that is too small saturates: the same people see the ad too often. Frequency (average views per person) climbs, so does the cost, and fatigue sets in. An audience that is too large, on the other hand, can dilute the message when the budget is low. There is no universal number: you check frequency and cost per result, then adjust.
Exclusions: the step people forget
Excluding means not paying to reach people who have already converted or who are irrelevant:
- exclude existing customers from an acquisition campaign;
- exclude recent buyers from a retargeting campaign;
- exclude visitors from the last X days when another campaign already covers them.
Target by stage of the journey
The right targeting depends on how far a person is from buying. This logic is detailed in The marketing funnel applied to advertising.
| Stage | Audience | Campaign objective |
|---|---|---|
| Discovery | Broad or interests | Awareness, Traffic |
| Consideration | Visitors, video viewers | Traffic, Engagement, Leads |
| Decision | Abandoned carts, contacts | Sales |
The five classic student mistakes
- Stacking too many criteria: the more filters you add, the smaller the audience gets.
- Confusing interest with intent: liking running does not mean wanting to buy shoes today.
- Forgetting the language and country of the offer: delivery limited to France cannot be targeted Europe-wide.
- Not testing: a single audience allows no comparison. See Test with an A/B test and The A/B testing guide for ads.
- Changing the audience too often, which restarts learning.
Frequently asked questions
What is the best audience on Facebook Ads? There is no best one in absolute terms: it depends on the objective, the budget and the data. You test several audiences and compare cost per result.
Is a lookalike audience always better? No. It depends on the quality of the source. A source that is too small or poorly qualified yields an unreliable audience.
Are interests still useful? Yes, especially at the start or for a niche offer, but they are increasingly supplemented by the algorithm’s own signals.
Do you train students in digital marketing? Request access.