Data mining of social network data

Targeted advertising is all the rage - instead of being bombarded with random advertisements, your behaviour on the Web can determine the kinds of advertisements you see. As Phorm is rapidly finding out this is a contentious area. However, the controversy surrounding targeted advertising is a simply the tip of the iceberg.

Targeted analysis is a relatively broad church. It analyses data about how a person interacts with some entity (the Web, a particular company, a combination of the two) builds up a profile of the person and then targets specific advertisements to them in order to influence their buying behaviour. The best known example of this is from the company Phorm, and as we've reported before on Server Management, it's not a universally popular technology.

Targeted marketing of this sort is clearly a much more powerful tool than randomly placing advertisements but it still a very blunt tool because we (the people in all of this) know what advertisements are for. We know they are simply attempts to influence us. So most of us actively try not to be influenced by them; although we are, of course, influenced whether we like it or not. (As a side line, my Father worked for the BBC and a great many years ago commissioned a survey on the efficiency of television advertisements. He delighted in one of the responses to the question "If you drink Guinness, are you influenced by the company's advertisements?" which was "I don't drink it because of the advertising, I drink Guinness because it is good for me.") So, targeted advertising is more effective than random but it is just the start.

The future

In analytical terms the business world realised some time ago that it was possible to gain a far better understanding of their customer’s behaviour by grouping them into clusters with similar traits. But humans are more than just ciphers that can be grouped; we are social animals; we don't just interact with companies, we also interact with each other. We influence our families, friends, peers and colleagues and in our turn are influenced by them. And we are influenced far, far more by our peers than by advertising. My guess is that you are far more likely to believe this article if it was recommended to you by someone whose opinion you rate.

In a sense this effect has been know for centuries and the business world has tried to manipulate it. Politicians are expensively wined and dined, as are journalists (but happily, of course, the latter group remains totally unaffected….). These are all attempts to influence influencers. But that is clearly at the macro level. If the business world could understand those subtle social interactions at the individual level it would have an extremely powerful tool, not simply to understand customer behaviour, but to manipulate it.

Suppose, for example, that we could identify those individuals who are peer leaders; then it might prove much more cost-effective to direct our marketing campaign against them specifically. But the devil is, of course, in the detail. Influencers and those influenced have complex relationships. Parents usually have a very heavy influence on the school that their teenage children attend, possibly rather less on the clothes that those teenagers wear at weekends. Your boss will have significant influence on your career, very little on your choice of holiday destination.

So, even focusing in on a specific trait (perhaps choice of holiday destination) how could you possibly identify the influencers within or for a particular customer group when human interaction is so subtle and undocumented? The answer is, of course, that you can't......unless the way in which human social interaction changes significantly; which is exactly what has happened with Social Networking. The Facebooks and Bebos of this world are essentially large databases that record how people interact. Every carelessly thrown sheep, each idly refused 'friendship' is a transaction in a database. Finding those people with many friends is a simple matter of counting. Following the interactions is more complex but it is still simply a matter of running the right kinds of analysis and we now have very, very powerful tools for analysing transactional databases.

So, a whole new world of analytical capabilities has just opened up. Of course, I’m not suggesting that Facebook is a mirror of 'normal' human interaction. I have never hurled a heifer in real life and, if I did, I'd probably be locked away. So I happily accept that Facebook is a whole new form of interaction; one that certainly borrows from 'real life' but no more than that. But it is true that, for many people, Facebook has become a significant part of their everyday social interaction; and analysing that part is orders of magnitude easier than analysing the rest. So, inevitably, it will be done. We can either see this as a higher and higher level of invasion into our private lives or as a marketing opportunity. (My view varies, depending upon the side of the fence that I occupy at any one time.) It doesn't matter. History tells us that if it can be analysed, it will be. Welcome to the Brave New World.