Audience targeting in marketing: the six signals and how to choose between them

Audience targeting in marketing means choosing which people see your message, using what you know or can infer about them. There are six signals to choose from: demographics, interests and behaviour, context, search intent, your own customer data, and lookalikes built from that data. Strong plans combine two or three signals, start with the narrowest signal you have evidence for, and widen only when the data says the narrow audience is spent.

The six signals, side by side

Signal	What it tells you	Where it is used	Best when
Demographic	Age, gender, location, language, income band, job title	Social and display ad platforms, B2B platforms	The product has a clear buyer profile
Interest and behaviour	Topics people follow, content they engage with, past activity	Social and video platforms	You are building awareness in a new group
Contextual	The page or video the ad appears beside	Display, video, publisher deals	You want relevance without personal data
Search intent	What the person is looking for right now	Search ads, SEO	People already know they have the problem
First party data	Who visited, bought, signed up or lapsed	Retargeting, email, customer list uploads	You have a list and consent to use it
Lookalike	People who resemble your best customers	Most large ad platforms	You need more of the people you already win

Start from the buyer, not the menu

The mistake is opening an ad platform and filling every targeting box. Begin with a one page audience definition and take the options from it.

1. Who buys. Name the role or situation, such as "owner of a clinic with two to five staff" or "parent shopping for a first school bag". Use real customers, not an imagined average.
2. What triggers the search. What event makes them look: a deadline, a failure, a birthday, a new law.
3. Where they spend attention. The platforms, publications, communities and search terms they use.
4. What they already know. Do they know your category, your brand, neither?
5. What you can prove. The data you hold today: customer list, site visitors, email subscribers.

Each answer maps to a signal. A buyer defined by a job title points to professional platform targeting. A buyer defined by a trigger event points to search intent. A buyer you already hold in a list points to first party data.

Match the signal to the stage

People who do not know the category

Use interest, behaviour and contextual signals. You are introducing an idea, so reach matters and the click rate will be low. Judge this stage on whether the right people watched or read, not on sales.

People who know the category and are comparing

Use search intent. A person searching a comparison phrase has done the early work. The page they land on has to answer the comparison. Our page on buying targeted traffic explains how to tell a real targeted visit from a package sold under the same name.

People who have visited or bought

Use your own data. Retarget site visitors with a reason to come back, email past buyers, and exclude existing customers from acquisition campaigns so you stop paying to reach people who already bought.

People who resemble your best customers

Build a lookalike from your highest value customers, not from all visitors. The quality of the seed list sets the quality of the lookalike. A seed of loyal repeat buyers produces a different audience from a seed of everyone who ever clicked.

Segment first, then target

Targeting picks who sees the ad. Segmentation decides which groups exist in the first place, and positioning decides what you say to each. If you have not done that work, the targeting menu will not do it for you. Our worked example of segmentation, targeting and positioning shows the three steps filled in, and positioning in marketing covers the statement that each audience should hear.

How to test an audience

• Change one thing at a time. Same creative, same budget, two audiences.
• Give each test enough spend and days to produce a stable result before you judge it.
• Measure the outcome you care about, such as cost per qualified lead, not clicks alone. For the arithmetic, see CPA in marketing.
• Keep a control audience that never changes, so you know whether a drop comes from the audience or from the ad.
• Write the result down. Audiences that worked in March are a starting point for October, not a fact.

Consent, sensitive categories and platform rules

Customer lists and site visitor data are personal data. Use them only where you have consent or another lawful basis under the rules that apply to the people on the list, and keep a record of it. Ad platforms restrict targeting on sensitive topics such as health, finance, housing and employment, and rules differ by platform and country. Read the current policy of each platform before you build an audience in those categories, and expect broad targeting to be the only option in some of them. Contextual targeting is a useful route in these cases because it needs no personal data.

Questions about audience targeting in marketing

What is the difference between audience targeting and segmentation?

Segmentation divides the market into groups that behave alike. Targeting chooses which of those groups you will spend money to reach.

Is narrow or broad targeting better?

Narrow suits small budgets and specific products. Broad suits large budgets and products with wide appeal, where the platform's own system finds buyers from your conversion data. Test both against your own numbers.

How big should an audience be?

Large enough that the ad platform can show your ad at your budget without repeating it to the same people too often. If frequency climbs quickly, the audience is too small for the spend.

How do I target without cookies?

Use contextual placement, search intent, your own logged in or consented data, and email. These do not depend on third party cookies.

How often should I refresh an audience?

Refresh customer lists and exclusions monthly, and review performance weekly during active campaigns.

What an assistant says when a buyer asks it, instead of searching, who to buy from

Every signal above assumes the buyer sees an ad or a search result you can aim at. A growing share of buying questions never reach either. The buyer types "which accounting software suits a ten person firm" into an assistant and gets one written answer with a few names. There is no audience setting for that answer. Your category, your location and your customer type either appear in the pages the assistant reads about you, or the assistant names someone else.

AI Knows Us puts the questions your target audience would ask to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, then reports which businesses each answer names. A free scan at aiknowsus.com shows how your brand appears for the audience you just defined.