GoFuckYourself.com - Adult Webmaster Forum - View Single Post - Business Thread... Am I the only stats freak in the industry?

jayeff · 04-26-2006, 12:26 PM

I studied statistics (at the London School of Economics) and started working life as a chartered accountant: thus I have a lot of faith in statistics. But I am also aware of their limitations.

Statistics which only assign a low probability to an event, or which have to include very broad parameters to be more definite, are almost useless for everyday business. The value of statistics is especially limited in this industry because we fail to sell so much more often that we are successful, therefore most of us don't have big enough numbers to do very much with at all.

Bigger sponsors are an exception, but even they can only dig a little way before their numbers will also become suspect. And the problem which even they face, is how to accurately interpret the figures they can collate.

A simple example, you have three tours and you want to know which one works best. It seems obvious: which generates the most sales per visitor? Not so.

Let's say that initially you had one tour, so for however long, all your affiliates pointed their surfers at that tour. Thousands of neglected pages are scattered around the 'Net, still sending visitors to it. Your more recent tours are only being used by newer affiliates and those who keep their sites updated, all assuming new is better. Mixed in will be a small number of affiliates who have actually tested which tours work best for them. Etc. The point is that the three tours do not get the same quality of traffic and therefore a simple calculation is not good enough. You are not really asking which converts surfers best, but which one adds most to the intrinsic conversion potential of the surfers it receives.

Framing questions correctly is where the problems start. Next comes recognizing which figures need to be modified before they will be useful. Then finally, as in the "simple" example above, there is the often insurmountable problem of how to generate those modifiers.

I suspect that between the nature of this industry's statistics and the likelihood that few people understand statistical theory, we probably end up with more wrong decisions than right ones being made as a result of people looking too closely at their numbers.

04-26-2006, 12:26 PM
jayeff Confirmed User Join Date: May 2001 Posts: 2,944	I studied statistics (at the London School of Economics) and started working life as a chartered accountant: thus I have a lot of faith in statistics. But I am also aware of their limitations. Statistics which only assign a low probability to an event, or which have to include very broad parameters to be more definite, are almost useless for everyday business. The value of statistics is especially limited in this industry because we fail to sell so much more often that we are successful, therefore most of us don't have big enough numbers to do very much with at all. Bigger sponsors are an exception, but even they can only dig a little way before their numbers will also become suspect. And the problem which even they face, is how to accurately interpret the figures they can collate. A simple example, you have three tours and you want to know which one works best. It seems obvious: which generates the most sales per visitor? Not so. Let's say that initially you had one tour, so for however long, all your affiliates pointed their surfers at that tour. Thousands of neglected pages are scattered around the 'Net, still sending visitors to it. Your more recent tours are only being used by newer affiliates and those who keep their sites updated, all assuming new is better. Mixed in will be a small number of affiliates who have actually tested which tours work best for them. Etc. The point is that the three tours do not get the same quality of traffic and therefore a simple calculation is not good enough. You are not really asking which converts surfers best, but which one adds most to the intrinsic conversion potential of the surfers it receives. Framing questions correctly is where the problems start. Next comes recognizing which figures need to be modified before they will be useful. Then finally, as in the "simple" example above, there is the often insurmountable problem of how to generate those modifiers. I suspect that between the nature of this industry's statistics and the likelihood that few people understand statistical theory, we probably end up with more wrong decisions than right ones being made as a result of people looking too closely at their numbers.