Media Perspective: Let's hope adland's algorithms do not cause a flash crash
A view from Russell Davies

Media Perspective: Let's hope adland's algorithms do not cause a flash crash

Last week, we were talking about algorithms, inspired by Kevin Slavin's marvellous presentation at Lift 11. (Google for 20 minutes of brilliant video.)

We were thinking about how pervasive these things are in the modern world - controlling, among other things, vast movements of money, how we find things on the internet and the sound quality of music.

And what's scary is how few people understand what these algorithms can do. Vast amounts of data, extremely high speed and lots of interacting formulae make it impossible for anyone to predict what will happen. Even the Wall Street quants who write these things can't really understand their actions in the real world. This has led to the phenomenon of flash crashes - sudden unexplained activity in the market when the algorithms run away with the trading, losing millions of dollars in value before real people can intervene. Considerable thought goes into preventing these, yet they still happen.

Why should we worry about this? Because more and more of our professional lives are governed by large amounts of data - too much data for purely human interpretation. So, the quants will be stepping into our lives too. And imagine the complexity, all the different algorithms interacting: the maths that decides whether an ad should get made and what it should be like; the maths deciding where it should be placed; how much it should cost; whether it was successful; who it should be targeting and following; whether they bought anything; what it means if they did or didn't; what should happen now; what that does to the value of the media the ad was carried in; what it implies about the value of the product etc etc ad infinitum.

All times many million, all very rapidly. Because all this happens right now, it just tends to happen slowly in different databases, run by different people and based on different assumptions. Some of it happens at high speed but most of it is slowed down by organisational and personal friction. That's good for common sense and calm but, sooner or later, people are going to get excited by the competitive advantages that might come from bundling it all together, putting all the feeds in one place and having a single intelligence overseeing it all. Imagine the advantages you'd get buying and trading inventory at that speed with that flexibility. Feels like it's going to happen sometime, doesn't it? Bound to.

Which raises the delicious and terrifying possibility of a flash crash happening in advertising and media land. Sudden unexplained activity, odd ads happening in odd places, creative work that makes absolutely no sense, media value being destroyed overnight. It seems all too possible. Or, as Mr Slavin suggests, maybe it's already happened. How would we know?