Before you build audience segments for social media agencies

What to take away

Start from behaviour

Demographic labels look tidy in a deck, but they rarely predict what a client will buy. Start with what people already do: which platforms they use and what they respond to.

The ONS publishes data on social media use by UK adults and households. Use it to check whether a segment is large enough to justify a campaign before you write any creative.

For England, filter national figures against regional population splits. A segment that looks healthy nationally can be thin once you isolate one city or county.

Three evidence sources

Primary evidence comes from client-side data: exportable analytics, enquiry logs and past campaign results. That shows who already converts rather than who might.

Secondary evidence covers published research and platform documentation. Google's guidance on campaign settings explains the policy constraints that shape which audiences you can lawfully target on its advertising surfaces.

Third comes structured interviews. Speak to six to ten existing customers per segment. Record what triggered the purchase, not what they like about the brand.

Lawful collection rules

Audience data is personal data the moment it identifies a living person. You need a documented lawful basis, a retention period and a route for people to object.

The ICO's children's information guidance sets expectations for services likely to be accessed by children. If any part of your audience skews young, age assurance moves from optional to essential.

Keep a provenance note for each segment. When a client asks how a list was built, a written record beats reconstructing it later.

Our guide to audience research for social media agencies in England sets out the wider research sequence this scoring step sits inside.

Score before you brief

Run every candidate segment through the same rubric. Score each criterion from one to five, then total. A workable threshold is 15 out of 20, but agree your own number with the client first.

Score size and reachability first. A segment with no route to it fails however well it scores elsewhere.

If two segments compete for the same budget, the higher total wins. Document the reasoning so it survives staff changes.

Test on a small budget

Run a limited test before committing the full budget. Split spend across two segments and compare cost per qualified enquiry, not cost per click.

Set a minimum sample before you judge. A week of low spend rarely separates two similar segments.

Where a test underperforms, check the creative before discarding the segment. Poorly matched messaging is a common false negative.

Review and refresh

Set a quarterly review date. Check whether the ONS figures have been updated and whether platform policies have changed.

Pair this with competitor research for social media agencies. Rival positioning often explains why a segment stops responding.

Retire segments that stop converting. Keeping a dead segment in the deck wastes briefing time and muddies the recommendation.

Common questions

How many segments should an agency work with?

Three to five is usually manageable for one client. More spreads testing budgets too thinly to reach a confident result.

Can I use client customer lists for targeting?

Only with a lawful basis and appropriate notices. Consent is the safest route where the list is used for advertising to individuals.

What if the client has no audience data at all?

Start with official statistics and a small primary research round. Six interviews plus published data will support a first hypothesis.

How often should segments be rebuilt?

Quarterly is a sensible default. Rebuild sooner if a platform changes its targeting rules or the client shifts sector.