Ask a brand who its audience is and you will nearly always get the same answer:
👉 People who buy what we sell.
👉 People who browse the category.
👉 People who look like our existing customers.
Here are the three main problems with that approach:
- There are only so many people that shop in your aisle.
- It is contested, because every competitor has defined their audience exactly the same way and is bidding against you for it.
- A good portion of them already buy your product, which means a meaningful share of the budget is spent reaching people who were going to buy anyway.
The interesting audiences are usually one category over but historically they haven’t been easy to find.
Why interest targeting cannot find them
Interest targeting sorts people by what they seem to like. Follow enough whisky pages and you are filed under whisky. Read about homeware and you are filed under homeware.
That categorising is the problem. It keeps people in the box they already appear to belong to, so it can never tell you that someone buying one thing is about to buy something completely different.
First-party transaction data can. It records what people actually bought, in every category they bought it, so it shows you real overlaps instead of assumed ones. That is the entire mechanism, and it only works where first-party purchase behaviour is the input.
So which category should you look at?
Across Flow Platform campaigns, the same four answers keep appearing.
A purchase they have already made creates the need for yours.
Some purchases arrive in a predictable order. Buy the first, and the second becomes near-inevitable within a short window. If you can see the first, you do not have to wait for someone to start shopping for the second.
@home needed to sell homeware. The obvious audience was people shopping for homeware.
Instead the campaign targeted active property buyers and tenants browsing properties at specific price points. The signal had nothing to do with homeware and everything to do with what comes next. People in the middle of moving buy homeware, in volume, in a compressed window, under time pressure. And almost nobody else was bidding for them.
That campaign returned a 9.78 return on ad spend, with a higher click-through rate and a significantly lower cost per click than the retargeting it replaced.
Look for sequence when your product is bought in response to a life event or a preceding purchase rather than on a routine cycle.
People plan occasion-led purchasing and it’s predictable.
Product categories are how retailers arrange shelves. Occasions are how people actually shop. One occasion pulls in products from all over the store, and anyone planning for it leaves a trail across every one of them.
Maison Martell was promoting its marquee at Durban July 2026 and needed to drive traffic to the ticketing page. Durban July is a fashion event as much as a horse race. People dress up, book restaurants and buy tickets, often weeks before anyone pours a drink.
So the campaign used three audiences: fashion shoppers, diners using a local booking app and users of a ticket purchasing app. Not one of them was specifically an alcohol purchasing audience.
The campaign delivered > 70 000 clicks, 81% above target, at a cost per click 41% below what was planned. Unique click-through rate hit 5.49% against a 4% target, and 88% of clicks came from different people, so this was real reach rather than the same crowd clicking twice.
Look for occasions when your product is bought for an event, and people start planning that event long before they think about buying from you.
They have already shown they will spend what your product costs.
For premium brands, the problem is almost never interest. It is how few people can afford the thing. Stick to your own category at that price and there may not be enough buyers left to run a campaign at all.
Johnnie Walker Blue Label faced exactly this challenge. Nobody stumbles across an ultra-premium whisky online, and the pool of people who buy at that price is tiny.
So they built a campaign audience around spend, not taste: verified wine and spirits shoppers, proven premium gift buyers, and people who buy often in other premium categories. The qualifying behaviour was not whisky. It was spending at that level, wherever they happened to spend it.
Festive revenue grew 39% year-on-year, and the campaign brought in 92% of all traffic to the product page over the peak.
Look for value when price point rather than category interest is what disqualifies most of the market.
Your category audience is largely made up of existing customers. Growth is everyone else.
The best audience to grow your base is not always the obvious ones. What do we mean by this? Consider this test Reckitt ran with Flow and Woolworths:
Reckitt set out to grow Finish dishwasher products at South African retailer Woolworths. The obvious audience was cleaning supply shoppers, which largely covers the brand’s existing buyers.
Instead, the campaign targeted Woolworths Food shoppers. Almost every household doing a full grocery shop is also washing their dishes, so that segment contains the current Finish buyer and the potential one alike. The existing base, and the market beyond it.
The campaign reached more than 200 000 Woolworths shoppers and beat the food shopper-conversion benchmark by 118% on impressions and 212% on link clicks. It also drove 93% of all traffic to the product pages on average. Without it, those pages would have gone largely undiscovered during the campaign period.
The shoppers you haven't won are shopping at the same store.
They're just shopping in a different aisle.
It does not have to be a retail audience
Adjacency stems from transaction data rather than retail itself. Any business that records real transactions holds signals about product and shopper categories it does not sell in, which is why commerce media reaches well beyond retailers.
Pokerbet needed to acquire customers in online betting, a category where advertising is both competitive and constrained. The audience came from Flow audience partner Zapper, a payments platform rather than a retailer, built on real payment behaviour across the places people actually spend.
Nothing in that data is gambling data. It predicted the category anyway. Pokerbet achieved a 156% return on investment on its first campaign.
How to find your own
Work backwards from the purchase:
What happens in someone's life in the weeks before they buy from you? A move, an event, a season, a routine. Each leaves a purchase trail somewhere. It is the first of the three questions worth answering before a budget is set, and the one most plans skip.
Write the hypothesis first:
"People who buy X are likely to buy from us because Y." If you cannot finish that sentence with a reason, you have a coincidence, not a signal.
Test it against real purchase data:
Instinct about adjacency is often wrong, and only what people actually bought will settle it. Interest targeting cannot do this step for you.
Check the audience is big enough for the budget.:
Too much spend against too small a pool just shows the same people the same ad repeatedly. Keep frequency between three and four, and if the pool cannot hold that, widen the signal or reduce the budget.
The audience nobody else has bid on
None of these campaigns was won with a bigger budget. Each was won weeks earlier, by asking a harder question about who the buyer actually is and then checking the answer against real purchase data rather than instinct.
The obvious audience is obvious to your competitors too. That is what makes it expensive, and it is why so little of your growth is sitting in it.
The audience worth finding is the one nobody else has thought to bid on.
Ready to turn your targeting into a growth strategy? Flow’s Audience Marketplace is home to hundreds of verified, first-party data audiences across categories from South Africa’s leading retailers and platforms.
GET STARTED 
Let the Flow team help you put together first-party data audience strategy
Frequently asked questions
What is adjacent-category targeting?
Adjacent-category targeting is the practice of reaching people based on verified purchases in a category other than the one you sell in, where that purchase behaviour reliably predicts a purchase in yours. A homeware brand reaching people who are moving house, or a dishwasher-tablet brand reaching grocery shoppers, are both examples. It works because purchase behaviour crosses categories even though audience data is usually organised within them.
How is this different from interest-based targeting?
Interest targeting infers what someone might want from declared affinity: page follows, content consumption, profile signals. Those signals are organised by category, so they keep people inside the category they appear to belong to. Adjacent-category targeting uses verified purchase data, which records what people actually bought across categories and can therefore show overlaps that interest data cannot see.
How do I know which adjacent category to target?
Four patterns cover most cases. Sequence, where a purchase they have made creates the need for yours. Occasion, where your product belongs to a moment that other purchases mark out earlier. Value, where proven spend at a price point matters more than category interest. And growth, where your category audience is mostly existing customers and the bigger group next door holds everyone you haven’t won yet. Start by working out which of the four describes your product.
Does adjacent-category targeting only work for retail brands?
No. Any business holding verified transaction data can produce audiences that predict purchases outside its own category. Payment platforms, booking services and marketplaces all hold this kind of signal, and it works the same way retail purchase data does.
Is an adjacent audience not less relevant than a category audience?
Usually the opposite. A category audience mixes people who are interested with people who are ready, and a good portion of it consists of customers you already have. A well-chosen adjacent audience is defined by a behaviour that predicts imminent purchase, which is a stronger qualifier than declared interest in the category.
