Geoffrey James has a very interesting post where he points out the top 8 reasons why sales forecasting don’t work. In a the nutshell, manual sales forecasts are quite subjective, and the more experts you pill up (especially if they happen to have different agenda), the more random the forecasts become.
In order to avoid the pitfall of manual forecasting, G. James suggest to hire a mathematician to built a computerized forecasting model. Yet, this approach has a major drawback: you need to recruit an excellent mathematician (or more precisely a data miner for that matter).
Indeed, in order to do that, you must overcome two large obstacles
distinguish the good mathematician from the not-so-good mathematician.
motivate the mathematician to join your company.
Paying an extra salary might be an issue also, but most probably, it’s a small issue compared to those two problems. In my experience, non-tech-oriented large companies usually fails dramatically at those two issues. This is not very surprising considering the tremendous difficulties encountered by attractive tech-oriented companies to recruit talented people.
Thus, if you do not happen to be a large tech-oriented company, Lokad might represents a low-cost outsourced mathematician: much cheaper and hassle-free.