Quote:
Originally Posted by Hrafn
As do you. You are inferring, without any apparent basis, that: - Amazon could not infer from your browsing and buying habits that you are relatively price indifferent. Or that they even care about the price elasticity of single customers, as opposed to large demographic groups.
- That every single mistaken click will in some way totally invalidate Amazon's data. I am quite sure that Amazon fully expects individual browsing data to be very 'noisy' from a statistical point of view, and therefore only interesting and useful in large sample sets.
- That asking buyers questions will reveal more about their buying habits than about the buying habits they think they should have. Finding out what actually makes them put down their money is almost certainly more useful.
|
And now your assuming that I'm assuming.
It's no assumption at all. We have enough history to show that pure stats isn't the end all/be all that some make out it is. Sure, it's a useful tool, but once again, it's not generic widgets that we are talking about. People have been trying to use pure statistics to predict since the 60's and before, with little success.
This reminds me of the sabermetrics debates which rage through the baseball fan community. Sabermetrics has gone from a useful tool that some people have had success using (but not very many) to the end all, be all way to evaluate baseball players without having ever seen them play. Stats are a useful tool that can point you in a direction. It's not the end all/be all that replaces all other tools.
The fundamental problem is that there is a host of factors that people evaluate when they choose a book. Price is just one. Without a model that takes those factors into account, it's very unlikely that you can reliably predict what will happen, just using that model.