| ioi said: The fact that you are playing on the "X - 3X" range means that you aren't fully understanding the idea of the confidence element of it and that it is about most of the data falling within a smaller range. There has been times when NPD data has been more than 10 times off "real" data, but obviously this only happens on very rare occasions - the majority of the time they will fall within 5-10%. Unless you track 100% of the market you can never say with certainty that your data will always fall in a set error band (i.e a discrete margin) since you simply don't know how many units are being sold in the proportion of the market you don't get data for. |
I fully understood. I know that there are several levels and ranges of confidence and that most of the values will fall within a shorter range than the 50%-150% I was talking about. I'm sorry if my example wasn't the best to illustrate what I was trying to say. I was supposing to work with 95% confidence (most of the companies from the different industries work with 95% or 99%, no one finds 85% or 70% reasonable) in order to understand what could I expect from the weekly numbers I see. My conclusion was: I can expect very little and the way the numbers are presented doesn't tell me that at all. My suggestion is: create a radio button at the top in which evey user can sellect the confidence he/she wants (95%, 85%, 70%) and display the numbers as rounded intervals. For example, a game like COD Ghosts for X360, which sold 550854, would appear like this "275k - 826k" for the 95% option, "367k - 734k" for the 85% option and "459k - 643k" for the 70% option. This way it would be clear for anyone how reliable the ranking actually is.
| ioi said: The best way to understand how unlikely in the case of VGChartz is to do a real backstudy with out data and see for yourself what kind of typical ranges you can see. The obvious issue here is the lack of available data from other sources to compare to, the care that needs to be taken over comparing the right date ranges / regions etc and, of course, the fact that they will also carry their own error margins, but there should be enough out there to give you an idea. |
Thank you for the tip but the range of perspectives from you and other users here in this thread is enough for me to have a pretty decent idea of what the numbers represent. I'm ok with what I get from this site and the information I have now to really understand that. I just think you could improve the way VG Chartz communicates and the way you interact with your own userbase (don't forget they are your ultimate clients).
Prediction made in 14/01/2014 for 31/12/2020: PS4: 100M XOne: 70M WiiU: 25M
Prediction made in 01/04/2016 for 31/12/2020: PS4: 100M XOne: 50M WiiU: 18M
Prediction made in 15/04/2017 for 31/12/2020: PS4: 90M XOne: 40M WiiU: 15M Switch: 20M
Prediction made in 24/03/2018 for 31/12/2020: PS4: 110M XOne: 50M WiiU: 14M Switch: 65M







