Personally, I think that ioi does a great job as well. The numbers here should be taken for trends, either continuing or being broken, not as the gospel of the actual number truly sold.
Every time you use a sample to predict the whole by definition there are going to be outliers, that is what the confidence level means. 95% within X% of the number forces a 5% change of greater than or less than X%.
And I would agrue that retail sales predicting are HARDER than a lot of sampling predictions. Why? Because it is so bottom-up fed. That is, some stores have problems and dont report their sales to corporate until the sales data has already been sent. Or the weather was so bad in a region that no one was nuts enough to go shopping. Or a new computer system goes online and has bugs in it, causing the data to have to be revised a week later.
If your sample data was polling any of those cases, then most likely it WILL become an outlier situation. But, if the later data shows it as such, and it gets corrected, LIKE IT DID, then the initial numbers don't matter!! Sure, PS3 dropped below 20% again, but in a few weeks or so, guess what. It will be back over 20%. How do I know? I don't, but the numbers here show a trend of PS3 % going up and I'm willing to bet that it will continue.
Torturing the numbers. Hear them scream.
