PDA

View Full Version : Buyers Guide Average and recommendations.


unixham
Sep 18, 2007, 09:40 PM
How are the Buyers guide averages calculated. I ask because I think that the formula used is just a straight linear average based on all the previous generations of a product wieghted equally. I think this is misleading and needs to either be re-weighted, have a moving average (say last two/three generations), or expressed as a series.

Though I cannot recall the last value for iMacs, I do believe it was close to 180 days (as I remember thinking "its just around the corner" during the Stevenote at MacWorld) but that the average is now 'only' 186 days after the most recent 15-month wait. 6mo <<>> 15mo.

This means that the longer term/multi generational products will have the least accurate recommendations, assuming that most recent past performance is a stronger factor than earlier; something that I do believe is true.

This determination of average strongly effects the guides "Recommendation" BUY NOW/BUY/NEUTRAL/WAIT etc.

YMMV.

CalBoy
Sep 21, 2007, 01:51 AM
I partially agree with you. I think that the mean should still be there, but that there should be a seperate mean, only for the last few updates (maybe three). This way, major discrepancies like the iMac can be avoided.

Eraserhead
Sep 29, 2007, 07:59 AM
Maybe a 25%, 25%, 20%, 15% weighting on the previous four generations and then the remaining 15% for any previous generations would be better. Though to be honest I think the whole buyers guide could do with an overhaul as it really isn't very good, new "generations" are sometimes the addition of a new model or a colour change, which isn't ideal.

CalBoy
Sep 29, 2007, 10:16 AM
Maybe a 25%, 25%, 20%, 15% weighting on the previous four generations and then the remaining 15% for any previous generations would be better. Though to be honest I think the whole buyers guide could do with an overhaul as it really isn't very good, new "generations" are sometimes the addition of a new model or a colour change, which isn't ideal.

Are we going to need an actuary?:p

Seroiusly though, the more accurate we try to make it, the more complicated it will become. At some point, the reader has to do some analysis on their own.

alphapat
Sep 30, 2007, 03:17 PM
I'm an actuary. Or at least I will be.

I'm happy to do this calculation for a small (large) fee.:rolleyes:

On a more serios note, I also think the more recent intervals should be weighted more heavily.

flamejob
Jan 29, 2008, 09:09 AM
I totally agree; it seems very automated at the moment and would make a Lay-person mistrust it.

BlizzardBomb
Feb 11, 2008, 11:31 AM
On the other hand, if we factor in large waits too much (which happen regularly around redesign time), it could push the average too high so the "Don't buy" status will come on even later, which would be worse than the current average method. It's still a very good point though.

arn
Feb 11, 2008, 11:32 PM
The buyers guide is due for rehaul. I'm open to suggestions... but this may take some time to get done.

arn

jecowa
Feb 14, 2008, 11:00 AM
I think giving users the ability to share their insights could be useful.

Maybe users could have the option to predict the release dates of Apple products. The average could user guess could be displayed with the average release date. Maybe the top closest guessers could be recognized.
Because the development new rumors will allow for more accurate guesses to be made, perhaps the individual guesses could be removed from the average guess after an amount of time, and the users would be allowed to make another guess. Older guesses would still remain in the system to be considered for a list of best guesses, this would give users incentive for guessing early on.

Additionally, perhaps users could make comments on items in the buyers guide. For example a user could write something like, "The Mac mini won't be upgraded until the MacBook is. If the mini was upgraded now, it would be too powerful compared to the MacBook." Comments could be rated by other users as "useful" or "not useful."

JNB
Feb 14, 2008, 11:32 AM
A simple minimum-mean-maximum should suffice. Trying to be too clever with weighting is not going to be more accurate (as we're trying to predict future behavior), and may cause even more consternation as members feel a more "accurate" method is leading them astray.

Also, then we need to get into definitions and weight them as well, such as the differences between updates, refreshes, major model shifts, etc. We can be as clever as we want with it, but the fact remains that we will not be providing any more useful intelligence than we do now, and may actually make the situation worse.

Folks have to be their own judges (and take a little responsibility for their own purchase decisions), I think that providing the raw data essentially as we do now is fair, honest, and transparent.

Eraserhead
Feb 16, 2008, 04:22 PM
^^ All my previous suggestion is doing is taking the view that what Apple has done more recently is likely to continue ;).