ioi said: I don't really know what you are asking.
VG Chartz takes data and applies some method to it which it feels scales it accurately to represent the whole market consistently across all data points.
NPD takes data, much more data, and has to estimate for retailers (Walmart primarily) that it does not poll directly.
Both are totally independent ways at trying to arrive at something and both differ subtly in what data they actually present - in that VGC is for the whole of the Americas whilst the NPD data you are using is for the USA only.
Why would anybody expect the two to agree perfectly? Look at Wii and PSP sales this week from Media Create and Famitsu - Media Create have Wii 6k ahead and Famitsu have PSP ahead - both by about the same margins that you are pointing out here (as a percentage of the values they are putting out). They both have large coverages in Japan and have been going for many years.
If you take NPDs data as being accurate to the nearest, say 5% (generous I feel personally when they are having to estimate for an unknown 40% of the market) and our data as accurate to 10% (which would be immensely impressive when directly getting around 1-2% of market data) then our figures are well within those ranges of each other.
I think a better understanding of statistics, sampling, the difficulty in doing what we do and some acknowledgement of the fact that most of the time we are comfortably within a 5%/10% margin would go some way to explaining any issues that people have with differences in data. |
When VGC consistently under track / over tracks the same consoles, it seems your sample weighting needs to be adjusted.
But ioi, with such a small sample (1%), don't you think VGChartz should use any source available to improve your numbers? The problem I see is that you guys estimate every week (awesome job btw) but you also keep historical data and total sales. This historical/total data could benefit from NPD adjustment, just like you do for others with ChartTrak data...