Consumer trends and Amazon – an interesting question for readers
I learned something today about Amazon’s algorithm structure. In some ways, I knew this already – because in a previous life I attempted to develop a “people who bought this also bought this” function on my company’s website.
This was a great a fun thing to do. Until I realised that you cannot pre-guess what people buy. We found that the algorithm created illogical product comparisons – so we abandoned the idea. Do you really want people to know that customers who bought a book on Thermodynamic Cylinder Head Design also bought the Sexual Fantasies of the British Housewife?
Does it serve a purpose? Well, it might do if you wanted to enter the world of blackmail of course. But seriously, would it be worthwhile to create a database of deviant thermodynamicists which is of suspect commercial value? Probably not. Unless you are in a totally unfocused publishing company which, like the Irish village shop, specialises in coffins and birthday cakes… And there are many publishers like that; niche-less and bible black.
Having said that, of course, what if there was commonality between metallurgists and the wanderings of the bored housewife? Or lawyers and high art? Or military historians and chicklit? Why shouldn’t there be? Who are we to judge what customers choose to buy? If one metallurgist wants to know what housewives think about, we cannot morally impose on the customer base an aribitrary censorship of fantasy. After all, who said that the metallurgist was a man? Or the housewife a woman?
Data is not about patterns that can be manipulated. But behaviour that can be understood.
But, turn this on its head. What if you were able to influence the list of the books that other people bought alongside your title? If this were possible, could you then use purchases of those – more popular – books to ensure that your book appeared as a targeted also-of-interest title when people bought the other book?
Now, data serves a new purpose. The data reveals how people behave. How they decide. What makes them decide. You can’t make them do things but you can respond to their behaviour to elicit a reasonably favourable outcome. How many buy x when you place them in condition y?
Truth be told, I don’t know the answer to this yet. But I’d like to hear the views of those who have worked in this area. It may indeed be a costly marketing process whereby instead of spending marketing dollars on email campaigns and direct mail, you spend it on buying loads of books just to create a logical algorithmic outcome.
If anyone out there has a view – please get in touch. If there are stronger brands out there in the market than you, surely it makes sense to leverage their existing position? After all, people do this all the time in events and sport (although there it’s called sponsorship).
I can foresee a whole new world out there of algorithm busters (just like the SEO geeks who have tried to out-Google Google). Of course, there is a question of ethics but my feeling is that such an approach is a new way of spending marketing money to achieve the same aim (even if it does send money into the hands of other publishers in the process – which may of course be a good thing).
However, and you might expect me to day this, the final purchase (irrespective of who recommends it) will always depend on the VALUE of the product being sold.
If new ways of marketing demand new ways of operating, it’s not how it’s done but what is eventually bought that matters. If there is no truth save the lonely search for money at any cost, we have returned to the horror days of the ghastly salesman with his spivvy moustache, sharp suit and two-tone correspondent shoes.
But if, by using new methods of marketing, we can advance the dissemination of worwhile knowledge, then that, surely, is a wonderful thing.