NPD to stop giving numbers to public?

[kn]

NPD uses a statistical process. Period. There is no other way unless all game consoles and software go through one person's hands and they count every title and every piece of hardware that they sell. I'm sure NPD's statistics are quite sound as they have both a large sample size and enough history to correct for unforseen anomalies. That said, it is a statistical estimation just the same. I would be willing to bet that NPD is NEVER better than +- 5% of actual sales in any month-long period for software. They may do better on hardware, though. Wal-mart is a big question mark in the numbers as they apparently don't report to NPD so NPD truly has to estimate those numbers... Those that take NPD as "actual" aren't thinking clearly about what they really do.

For any of you that actually understand statistical sampling, I would be interested (we'll never know, of course because that is their secret sauce) in what NPDs confidence intervals are. For the laymen it is basically a number expressed as a percentage that indicates how confident the person doing the analysis is that their results are "correct". I.e. we are "90% confident that the sales of guitar hero 3 for the PS3 are between 90-110K for October." Some people don't want 90%. They want 95 or 99%. As you push the percent confident up, it requires a larger sample size. a 100% confidence level would, of course, required 1:1 sampling which isn't possible.

So, in sum, NPD (and VGC for that matter) take a sample, analyze it, apply some statistics to arrive at a mean and then shoot for a confidence interval they can live with. NPD is reporting their mean number above of 100K (they leave out the +- 10K of course).