The true cost of unreliable audience data

Hi everyone, and welcome to edition 128 of The SEG3 Report! This week I’m diving into NumberEight’s report on audience data, and what it means for folks trying to attract new fans, audiences, and consumers.

Plus, our speed reads:

Let’s get into it.

A couple of months back, a good friend of mine came over to cat sit for a long weekend. He and I have sometimes overlapping, sometimes very different, taste in TV and movies. He texted me at one point to apologise for ruining my algorithm by watching a specific show. 

After he left, and my partner and I had settled back into our routine, I realised that he had been left logged into one of our streaming accounts (a free one, by the way!) and I had to text him the same apology. 

And yet, fundamentally, no one on the other side of those streamers, collecting audience data, had any idea that I wasn’t my friend, or that he wasn’t me. Or my partner! All they knew was that someone named Gabriella Geisinger was suddenly, and briefly, into BBC Proms: Bond and Beyond. And sorry to disappoint, but I am not.

This is effectively the main takeaway of a new report from NumberEight, and for folks who rely on audience and fan data to build their businesses, sell their products, or strategise for IP expansion, it is crucial reading.

Test one is CTV, which comes from a data broker across 7,013 shows and more than 3 million viewers. The TV IDs identify a household, not a person, so the report checked what the data says about each genre against published research on who watches what.

The data returned the same audience for everything. Every genre, from horror to romance, came back 53% to 55% female. Horror landed within a point of romance, and the male-skewing genres showed no male skew at all.

Age was just as flat, with animation, news and war all looking like one older-leaning audience. A planner buying "male 18-34" against this data would be working from a profile that can't tell a war film from a soap.

Test two is mobile and where NumberEight's model is applied. Using campaign delivery data, the report compared its own ID-less model with an IP-based model from a leading US data provider, across five games. The IP-based model returned close to 50/50 every time. For Fashion Battle it said 51% female, while NumberEight's model said 82%.

The two tests point to the same weakness in the way this data is collected. A household or IP-based ID test can only describe the household, so a family of four gets one profile and every show inherits it. NumberEight's approach skips the household and reads the behaviour and context instead, which is why its numbers change from game to game,

Here's why this slips past so many plans. A 50/50 split looks believable for gaming, because the industry really is close to it: ESA's 2026 figures have 53% of American men and 46% of women playing.

Nobody blinks at the number because it’s right on average, but it’s less correct when it comes to specific titles. (Duh, we hear you say. But it’s a fair reminder!)

These titles are where audiences separate, and that - obviously - matters to anyone trying to capture new fans. In the same ESA research, puzzle is the top genre on mobile, driven mainly by older players, while action, shooter and arcade lead on console. A brand that treats "mobile gamers" as one audience will miss that a puzzle player on the bus and a shooter player at home aren't the same customer. 

→ Remember, although a number can pass a sense-check because it mirrors the population, it’s still important to run it against a specific title and really drill down into the data.

There’s an impulse to see self-declared data as a fix to this, but it comes with its own problems.

One mobile gaming publisher in the report found that 5% of its users had given their age as 99 or 100 (lol). Where it’s less ‘lol’, is in research commissioned by Ofcom in 2022, which found a third of 8-17-year-olds with a social media profile had signed up with an adult age; as we enter an era of social media bans for minors, this becomes key for folks trying to capture younger audiences and stay on the right side of regulations.

A prime example of this is what I mentioned in edition 127: that the NFL's view was that a fan it hasn't captured by 18 won't give it another chance, so leaning on a birthday field someone typed in a hurry feels shaky. 

And, anyway, labels don’t always stand up, even when people are honest. In edition 83, Livewire and PMCI Play's research found that 64% of players don't strongly identify with the word “gamer”. This means that trying to parse an audience by this behaviour as self-identified may not be accurate, and given how impactful in-game presence can be, this is an important audience to get right. 

It’s funny to me in a way because this butts up against something we know to be true - that fans want to wear, show, and shout about the things they love. However, that’s very different to feeling like your data is being captured, sometimes against your will. How many of us have filled in bogus names and emails into forms to get on the free WiFi at a museum or a bar? 

Voluntarily displaying your pride in a piece of IP is very different to feeling like your personal information and behavioural habits are being scraped, and balancing that line is delicate. 

Truthset, which rates the quality of audience data, estimates that roughly 40% of open-market programmatic CTV spend is wasted on inaccurate data, about $7.4 billion in 2026.

Still, demographic buying is still everywhere: Gracenote data reported by eMarketer has 29.5% of CTV advertisers using demographic targeting.

So, the concrete question to ask is a comparative one. If you look at audience data for two different genres side by side, and - for example - your romance series comes back with the same age and gender split as a horror one, you're not looking at data that’s going to help you find new fans, audiences, or consumers.

It’s also important to leverage the insight you already have. Sports, entertainment and gaming brands have something most advertisers don't: fans whose behaviour is visible in the game, the app and at fan events. In edition 110, we looked at why in-game placements work — because they arrive inside a moment where someone’s attention is already switched on. It’s more feasible to plan for an audience whose behaviours you already understand, and whose needs you can anticipate. And it’s data that’s of value to people who want to collaborate across the Venn diagram of sports, entertainment, gaming, and consumer. 

Also, there’s another slightly obvious option: plan for the room as a whole. If you buy CTV data, you can pretty safely assume more than one person is watching. So what you have to do is build a creative that works for a whole household, regardless of whose profile is logged in. Which, granted, is a tough creative ask, which is why partnering with the right team is essential (and why the human touch is invaluable).

I do keep thinking about the Bond Proms, and the idea that somewhere some tiny datapoint says that’s something I’m interested in. It’s funny to me, but for someone trying to actively engage an audience that is into Bond Proms, they’re fighting a losing battle. 

It illustrates quite plainly that it’s worth thinking critically about what an audience of "women 25-34" or "sports fans” or “someone who watches Bond Proms” actually is: a real audience, or a household average that would look identical whatever was on screen? 

There is no quick fix, no form that is going to deliver the exact demographic of people watching a specific show, or playing a specific game, which is why collaboration across these silos of sports, entertainment, gaming, lifestyle and more is so important. The more we understand about how fans operate, the better experiences can be built, which is healthy for the industry as a whole. 

Celebrating Bluey’s hometown of Brisbane, Citymapper is also launching in the city for the first time

TL;DR

Why you should care

Like the example above, this collaboration works because of a shared core. In this case, it’s built on families spending time together. From Citymapper’s perspective, travelling with young children can turn even a simple trip into a major operation (as my friends with children often tell me!). The kid-friendly routes, with shorter walking times, step-free access and pram-friendly stations, are genuinely useful for app users, regardless of their feelings about Bluey. The integration then brings Bluey to new fans, new families. 

The actual integration features custom navigation icons, video clips built into journey planning and games via Bluey.tv and this all coexists with the app’s existing mechanics. The character isn't there to sell you anything; it’s there to make a stressful moment slightly easier, which is precisely the kind of role Bluey plays in a family's day.

This matters because audiences are quick to spot the difference. As Shelley Macintyre, CEO of BBC Studios' Brands and Licensing, told us on The Speakeasy, "people sniff commercial cynicism very quickly." A transport app is not an obvious home for a cartoon dog, but the utility and the specific campaign language are what make it work.

The fit with the show's premise is the second point. Play The Bluey Way is BBC Studios' campaign built on the idea that play can be found in everyday life. A commute is about as everyday as it gets. Shelley described the core of the brand as "the emotional connection that is born from shared play between adults and children", and a journey is time an adult and a child already spend together. Rather than asking families to set aside time for a branded experience, this exists for them in a moment they’re already in.

This strategy also harkens back to the digital, physical, product experience loop we’ve discussed at length before. The murals and play activations are the physical moments people photograph and share. But long after these activations end, people will keep using Citymapper and watching Bluey, meaning the three layers continue to inform each other. 

Working on anything cool, or have a press release you would like us to cover? Send it in for the chance for it to be covered in next week’s edition!

That’s all for now, everyone - thanks again for reading the latest edition of The SEG3 Report. If you found it of interest, please do consider sharing with a colleague or friend who’d enjoy it too!