Brandtech's Third Date

7:58 on a Thursday evening, the table by the window, that’s where he said to meet.

You already know a great deal, actually. He will be early, as he is always ten minutes early, you have been told. He will order an IPA, not a cocktail. He works three streets away, in a job you have read enough about to ask at least one decent question. You know the last place he went on holiday and that he did not much like it. Every fact is correct. Every fact is about the past.

The door opens. He is early.

In about forty seconds you are going to have to say something. The first thing you say. The second. The first thing you ask.

The problem is that nothing you have gathered makes you interesting.

It only makes you punctual.

Which is, as it happens, a fair description of the agency data business.

Data can answer two kinds of question. Who is he, where is he, and when? And what do I say when he sits down?

For fifteen years the industry has spent billions answering the first kind. A year ago this month I wrote that BrandTech was the Goldilocks company and that the thing to watch was how it moved upstream of the last mile.

This week Brandtech launched a 500-person data practice, but what is interesting is the question the practice was built to answer.

Not where is he. What do I say when he sits down.

A data practice learns the question it is paid to answer. Media is where holding-company margin lives: the trading, the scale deals, the sheer cost to a client of moving a billion dollars of spend somewhere else. So when the groups built their data businesses, they built them next to media, and the data learned to serve it.

Look at what was bought. Publicis paid $4.4 billion for Epsilon in 2019. IPG paid $2.3 billion for Acxiom the year before. Dentsu bought Merkle. WPP built Choreograph. Call it $7 billion and change, and nearly all of it aimed at one thing: resolution. Knowing a person precisely enough to buy him more cheaply than the agency next door. Where is he, and when.

The result was the most punctual date in the history of the industry. Right bar, right time, right drink. And then forty seconds of silence, because nobody had been paid to think about the sentence.

Three in ten American adults have used a dating app. The apps are extraordinary at the top of the funnel: the swipe, the match, the message and some back and forth before a first date.

And then they stop. By the apps’ own surveys, somewhere between a fifth and a third of first dates produce a second. Pew found that 12% of users ever reached a committed relationship. The figures are self-reported and American, but the shape of them is not in doubt. The platforms made the first date abundant and the third date no easier at all.

That is not a failure of intent. It is margin gravity again. Tinder is paid for visibility: fewer paying users last year, each charged more to be seen. So Tinder optimised the swipe. Hinge, same parent company, built its product around the second date and grew direct revenue 26% last year while Tinder’s fell. Same group, different P&L, different question.

The dating funnel is the marketing funnel with better lighting:

Everything above the line is what $7 billion of agency data was bought to do. Everything below it is creativity, and nobody’s P&L was sitting next to it.

Until, perhaps, this week.

Brandtech Data, announced this week, pools fifty-five, DP6 and Jellyfish: more than 500 specialists, 22 countries, north of $100 million in revenue, Will Heins as chief executive and Pierre Harand as president, both keeping their day jobs. The three brands stay. There was no reorganisation, and Harand says fifty-five went from ten countries to twenty-two overnight simply by being enabled to work with its siblings. On its own that is a respectable practice launch and not much more.

What makes it more is what Brandtech does not have. No linear television. No print. No trading desk, no scale deal, no legacy media margin to defend. Its media is biddable, transparent and thin. What it does have is Pencil, which has made ads at industrial scale, and OLIVER, which makes them inside the client’s own walls, and Jellyfish, which carries them through social and creators. Pool the data in that group and it has nowhere to go but the conversation.

And that is what Harand told the French trade press. The English-language trade press only ran the release (seeing it perhaps as a bit ‘meh’), and the only interview I have found is the one Minted’s editor conducted, in French. Weigh what follows as a single source. Brand data, he said, will go into Pencil to shape what gets made, and measurement will come out the other side to say whether it worked. Generation to evaluation, one loop.

His diagnosis of the last three years is the best line anyone has offered this week. He calls it pilot purgatory: a proliferation of AI tests run with no measurement protocol and no success criteria agreed in advance, so that nobody can prove the return and no finance director will fund the scale-up. The remedy is unglamorous. Define the result, the metric and the protocol before the pilot starts.

One caution, because the metaphor cuts both ways. There is a kind of date who has read everything. Your employer, your ex’s name, your half-marathon time, and opens with all three. Data you cannot turn into something worth saying is not an asset. It is slightly creepy. The industry’s version is the forty-slide data review that spends two hours on reach and match rates and four minutes on the three executions on slide thirty-eight. More knowing is not the answer. Knowing what to do with it is.

Suppose the first date goes well. You were on time, you were interesting, there is a second. By the third, the thing that matters is no longer where he is or what he drinks. It is what you do with what he told you on the second. The sister in Lyon. The job he is thinking of leaving. Whether tonight you have something new to say about either, or start again from nothing.

Media data cannot do this. It was never built to. Its idea of memory is a frequency cap: we have met three times in a fortnight. Met, and nothing else. It retargets, which is the date who keeps appearing at your bar after you have said no. It sequences, which is telling the second story after the first regardless of how the first went. A diary with every date filled in and every page blank. After a year of it the brand walks in, sits down, and asks what you do for a living.

Which is the real reason Brandtech’s Pencil platform is sitting inside so many global marketers, and why that matters more this week than it did last. It is not that it makes ads quickly, though it does, by the million. It is that an orchestration layer is where the last conversation shapes the next one: what was made, for whom, what it said, what landed, what did not, and the next thing it makes begins from there rather than from nothing. Remembering is the easy part. Saying something better because you remembered is the work.

That is my reading of what the layer is for, not a feature list, and Brandtech may describe it differently. But it is the only reading under which a $100 million data practice and an AI production engine belong in the same sentence. Media data remembers that it met you. It has no idea what to say next.

The big groups will not sit still. As media margin thins, the reason to point data at media thins with it, and I will make a call: inside eighteen months one of the four will launch a creative-data practice and present it as a first. It will not be. Brandtech got there by not having the option, which is what scarcity does to a company. It makes you work out what your data is actually for.

A year ago I asked how Brandtech would move upstream. Upstream, it turns out, means this: data in the service of what you say next.

Data gets you to the bar on time. Creativity gets you asked back.

Your agencies can get you a first date with almost anyone. Which of them has ever been judged on the third?

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry’s table: senior client, global agency leader, technology founder. Over 4,300 senior marketing leaders read it twice a week. Worth a second date, every Monday and Thursday: brandflow-marketing.com.