Audience targeting on Meta ain’t what it used to be.
Charities have spent years refining how they reach supporters – layering interests, narrowing demographics, building detailed pen profiles and testing countless audience combinations. The assumption was that the more precise your targeting, the better your results.
That was broadly true. But it isn’t anymore.
Meta’s advertising system has fundamentally changed. The platform has been moving away from granular, advertiser-controlled targeting for some time – with the loss of Audience Insights, the disappearance of targetable interests, and the removal of interest exclusions – but the introduction of Andromeda, Meta’s updated ad retrieval and matching system, marks a more significant shift. The levers that used to drive performance are weakening. New ones matter more now.
Audience building hasn’t disappeared. It’s still central to campaign performance. But what makes an audience strategy good has changed significantly. Here are five things charities need to reconsider.
1. Stop over-defining your audiences
There’s a deeply ingrained instinct in marketing to build tightly controlled audiences – stacked interests, narrow age ranges, carefully considered exclusions. Charity marketing is no different. It feels rigorous, and for a long time it was a good approach.
The problem is that Andromeda is specifically designed to work differently. The more you constrain an audience, the less room you leave the algorithm to find the right people. Overly rigid targeting limits the learning pool, which limits performance.
The shift here isn’t about being less strategic – it’s about understanding that the system now performs better with more freedom, not less. That has real implications for how audience structures should be built and how much weight should be placed on demographic assumptions.
2. Think in audience types, not just audiences
Meta gives advertisers three distinct audience types to work with: saved audiences built on interests and demographics, custom audiences based on direct engagement with your charity, and lookalike audiences modelled from seed groups.
Historically, many charities have leaned heavily on saved audiences and underused the other two.
But that balance needs to shift. Each audience type plays a different role in feeding the algorithm with useful signals – and a campaign structure that relies on only one type is leaving significant performance on the table.
Getting this right isn’t just a matter of switching a few settings. It requires thinking clearly about what each audience type is being asked to do, and whether the seed data behind lookalikes is actually strong enough to build from.
3. Move from targeting identity to using evidence of behaviour
The classic audience-building model was rooted in identity: target a certain type of person, on the assumption they’ll behave in a certain way. Target environmentalists. Target people interested in mental health. Target people who follow the kinds of organisations your donors might follow.
The issue isn’t just that interest signals are weakening – it’s that stated interest has always been a weaker predictor of action than actual behaviour. Someone following a well-known conservation charity doesn’t tell you much about whether they’ll donate to yours.
Andromeda leans much more heavily on observed behaviour – what people have actually done, not what they’ve indicated they might do. That means the quality of your behavioural signals, and how well you’re capturing and using them, now has a direct bearing on how well your campaigns perform.
4. Treat first-party data as infrastructure, not an afterthought
For most charities, first-party data is both the most valuable asset in their Meta campaigns and the most underused. CRM data often sits largely untouched. Website tracking is fragmented or incomplete. Data flows into the platform inconsistently, or not at all.
Before, this was manageable because interest targeting could compensate. Now, it can’t. The quality of the data you’re feeding into Meta directly affects how well the system can optimise.
Weak signals produce weak results, regardless of how good your creative is.
Getting this right involves more than uploading a contact list occasionally. It requires thinking seriously about tracking architecture, data hygiene, and how your CRM and ad platform are actually connected – and whether what you’re sending Meta is genuinely useful to it.
5. Reframe what a good audience strategy is trying to achieve
The old model used audience groups partly as a targeting mechanism and partly as a research tool – by testing different audience segments, you could build a picture of who your donors were and refine your approach over time.
That model is breaking down. Interest signals are weaker, small targeted groups are less effective, and the granularity that made this kind of testing possible is disappearing.
A good audience strategy now looks quite different. It’s less about finding the right segment and more about creating the conditions for Meta’s machine learning to do its job well. That means thinking carefully about data quality, signal strength, campaign structure and creative variety – and understanding how each of those things interact.
Good data infrastructure means better Meta performance
The charities that will perform best on Meta over the next few years won’t necessarily be the ones with the best audience research. They’ll be the ones with the best data infrastructure, the clearest understanding of how the platform now works, and the campaign structures built to take advantage of it.
If you’re not sure whether your current approach is set up for that, it’s worth finding out.
How we can help
We can audit your Meta ads account or draw on our 12 years experience in running Meta campaigns for charities by planning your audiences and running your fundraising and awareness raising campaigns for you.
Drop us a line to set up a call to find out more.
