Why Better Ad Targeting Starts with Better Audience Data
As automation takes on more of the ad-buying process, audience data gaps can leave marketers reaching people far removed from their original campaign brief.
In an ideal world, every campaign starts with a clear view of the audience it wants to reach. But by the time the ads are served, the intended audience can often be quite different, undermining campaign effectiveness and weakening performance.
Automation has been a boon in many respects, but without care and attention, it can also amplify a series of unintended consequences.
The problem is the gap between the original audience brief and the final impression. In the brief, the audience may look spot on in the description or methodology, but it may not accurately describe the people within it.
And even at the data provider stage, demographic data is accurate in only 55% of cases, with a further 40% loss in accuracy during onboarding. Audience data gets matched and re-matched, target groups are simplified into broad segments, and relevancy is diluted.
With automation set to become more integral to the ad buying and optimisation process – Dentsu predicts algorithms will account for 75% of all ad spend by 2028 – more campaigns will be optimised toward whatever audience definition they are given. If that definition starts drifting, marketers can find themselves a long way from their intended audience before they realise what is happening.
The difficulty is that this audience slippage often goes unnoticed. Most marketers judge campaigns on reach, CPM and clicks, metrics that don’t necessarily reveal whether the right people were reached.
PrimeAudience research found that 93% of UK marketers are confident in their targeting, yet 92% plan on changing their strategy next year. This raises a bigger question: if marketers are so confident in their targeting, why are so many planning to change their approach?
Failing to electrify consumers
What does this look like in the real world? Say a car brand wants to launch an electric SUV targeted at affluent consumers who are actively researching the market, and it’s particularly interested in consumers comparing models on safety, real-world range, charging, and total cost of ownership.
This is a good brief: it names the category, the precise points in the customer journey, and highlights the factors influencing purchase decisions.
But then the brief meets the buying platform’s broad, loosely defined audience categories, where specific behaviours and motivations are lost in vague, catch-all segments. Someone interested in EVs, someone actively comparing models, and someone researching running costs could all be grouped under the same broad labels: EV audiences in general, vehicle buyers, and SUV customers. These categories say little about what people are actually looking for, what matters to them or where they are in the buying journey.
Then, as audience data moves through the system during onboarding, it is matched to the identifiers used to buy ads, potentially diminishing its accuracy by the aforementioned 40%. It’s not a question of whether third-party data is good or bad, but rather about the distance between how the audience is initially described and what ends up in the campaign.
If the marketer can’t see where the slippage is happening, they can’t act. Closing the gap between intention and outcome means giving buying systems a fuller picture, not just of audience segment but also of consumer intent and the context in which each ad appears.
History plus context
We know historical audience profiles can tell us much about a consumer. Past purchases, website visits, or demographic profiles are useful signals, but they don’t tell the whole story.
However, in our electric SUV example, someone who has just read a detailed review of that specific type of vehicle has shown a clear and current interest in knowing which brands are selling what. It’s a strong sign of audience relevance in the moment.
Contextual signals don’t replace first- or third-party data, but the campaign’s objectives will determine how those elements are balanced. Relevance isn’t an either/or; it comes in degrees and can depend on the campaign and the context.
Effective targeting, therefore, depends on weighing multiple signals together rather than relying on a single audience label.
Keep the detail, match the speed
All of this can easily happen in manual processes. When automation gets involved, it happens faster and at scale.
Algorithms will always optimise towards the definitions and signals they’re given, and if the audience has already lost accuracy during onboarding, the automated system is going to reinforce that weaker targeting definition.
AI can help, but only when used hand in hand with a marketer’s skill. LLMs can interpret a detailed brief and pick out the elements that matter most, such as buying stage or the range anxiety highlighted in the SUV example. These insights can then translate into actionable signals for an ad-buying system. This helps preserve more of the original audience brief throughout the buying process.
But marketers must always be firm on their intent. AI can interpret that intent – but it can’t create one that wasn’t there in the first place. Marketers need to ensure the AI-powered audience still matches the initial intent by the time the impression is delivered; for that, the system needs the right understanding of the audience from the word ‘go’.
Ultimately, the marketer’s role remains the same: define the brief, review the output and ensure the campaign stays on course. As automation takes on more of the execution, preserving the detail behind the original audience intent will become increasingly important.
ALSO READ: Mixpanel’s CEO Jen Taylor on Turning Technology into Measurable Business Growth
