There was a fascinating TV programme last night that set out how policing is being influenced and improved as a result of using predictive algorithms. The show set out how the Los Angeles PD in one particular area have been using the algorithm to effectively show up in the right place at the right time to anticipate criminal activity and to deal decisively with it or deter it altogether. The numbers are impressive, the reduction in the level of crime is hard to ignore. So much so that the use of ‘predictive policing’ is now being rolled out all over cities in America.
The use of predictive algorithms has come about because academics at UCLA spotted the potential to marry what they had learned from earthquake aftershocks with the massive database held by the LAPD.
The use of powerful algorithms to use past behaviour to attempt to predict future behaviour doesn’t just extend to criminal behaviour. Obviously marketers are hugely interested in being able to anticipate demand, planners would love to be able to anticipate where to build roads, drains and other infrastructure and so on.
The one that I was thinking about though was managers. It struck me that a lot of problems in business these days are not new problems. They often involve people making bad decisions in an eminently predictable way. “I knew when he spoke to them like that they would refuse to co-opererate”. “I knew the minute the email went out that people would get the wrong idea”. Many of the stupid things that happen in business can probably be predicted because they closely mirror mistakes made previously.
So if the algorithm is vitally important because it can help to find meaning in the data, what about the data? Well that’s surely the problem when it comes to learning from managerial mistakes. I’m not sure that managerial mistakes are actually captured in data form anywhere, in any company. Why is that?
If people are our most important asset (do I hear you yawn?) then surely we should be learning about the way we manage and treat our people to find out if we are being effective. We should have data, data which we can mine to find out how to predict the problems and avoid making them in the future.
Obvious, isn’t it?
Justin's 'Just One Thing..' 