It's a question we’ve all faced: what to do when you’ve written machine learning code to help Nasa determine the optimal configuration of components for a manned mission to Mars?
Use it to sell advertising, obviously.
Such is the story of DataXu, one of the pioneers of Real Time Bidding. Founded way back in 2007, it is now capable of analysing and optimising buying decisions on over 100bn digital advertising opportunities every day.
I spoke with Chris Le May, senior vice president and managing director of Europe and emerging markets, about the changing face of programmatic.
Life on mars
We’re sat in DataXu’s London office, where I ask Le May just how a piece of Nasa technology translated into advertising. “At our core we are a decisioning system, basically. And that decisioning system’s job is across all of our clients and all of their campaigns – across all of the different media channels which we integrate – to make decisions about which piece of advertising inventory is most appropriate for that brand, and that campaign, at that moment in time, based on who the audience is.”
Pausing as I realise the sheer absurdity of asking someone to decode Nasa decisioning technology mid sentence, Le May takes up the mantle with a smile, explaining it to me in the same way Dr Bill Simmons, one of the founders of DataXu, explained it to him: “coming from MIT, he’s a very smart chap – and rather than belittle me he was very gracious. We were sat in a meeting room with tables and chairs like we are now, and he said to me: ‘imagine that the question that we are trying to answer is: what is the optimal configuration of the furniture in this room, for best use?’
“He said we could write a piece of code that would crunch an almost infinite set of configurations of the tables and chairs in this room, and it would take a very long time, and it wouldn’t be definitive. Or, we can write something that is capable of understanding that any configuration that involves the tables or chairs being upside down in any position is sub-optimal – don’t consider them. Anything that involves the table or chairs being nailed to the wall or ceiling is sub-optimal – don’t consider them. And what you’ve actually done there, is you have exponentially reduced the number of things you need to worry about considering. And this is basically what the core kernel code was when they were working for Nasa to help them plan this Mars mission. This is where it came from.”
Clearly there were no tables and chairs in space to nail to walls. But there were 3bn possible component variations that the code narrowed down to 1,600 in a matter of hours. “The underlying principle behind what we do is to reduce the number of variables that we need to consider.”
Such is the case with progammatic television (PTV), a nascent field DataXu is investing heavily in. The idea is that, in the same way we are targeted highly personalised ads online, television advertising will become the same. There’s a similarity between Mars exploration and PTV: lots of talk and speculation, the technology exists, and it will happen eventually – but it’s a step into the relative unknown.
“Our industry tends to do this with whatever the latest buzzword, phrase or acronym is,” says Le May. “Everyone quickly realises ‘wow, I ought to claim that I can do that, shouldn’t I?’ A lot of people have a lot of talk, but the hype versus the actual scale of activity is probably a complete mismatch at the moment.”
Despite this, DataXu has been running PTV campaigns, predominantly in the US, for about a year. It has recently expanded its partnership with Acxiom, to license its third-party data for advanced TV and analytics solutions, as well as receving a £10m investment from Sky to advance its role in shaping the market as it progresses.
Many a 2017 outlook piece said this year would be The Year of Programmatic Television, but is it?
“Well, how long did we talk about the Year of Mobile before mobile finally got here? It depends what somebody means. If they’re talking about it becoming the norm that the majority of TV is bought in this way, then no, that’s not this year at all.”