Forum Moderators: DixonJones
So lets say I have widget 1 and widget 2.
Widget 1 has 6 reviews and a mean average of 7.8
Widget 2 has 1 review and an average of 9.
Therefore, widget 2 becomes higher in my sites ranking system.
I'm not too fond of this and would ideally like to iron it out.
Could someone please help me with ideas for balancing the system and/or with some more advanced math logic than mean averaging.
Any help would be great,
cheers!
[edited by: Tom_Cash at 9:07 am (utc) on Feb. 2, 2009]
Ranking from 1-10, average score is 5
Your most reviewed widget has 50 reviews, score = 7.8
Your 2nd widget has 10 reviews, score = 9.0
You could add 50-10 = 40 times the average score of 5
And end up with a weighted score of (40 * 5 + 10 * 9) / 50 = 5.8 for the 2nd widget
If that seems unfair, you could also compare the 10 newest reviews only.
Also, there are lots of other ways to describe "typical score" in the statistics world. The median, the geometric mean, etc. Try a search on these terms to get to somebody's statistics glossary.
And you can also ignore the idea of a middle point and go with something like "percent of reviews where the rank was higher than 3."
A good service for your readers would be a histogram, i.e. a bar graph, of the ranks. That way, people can decide for themselves based on the pattern, and they can also decide for themselves if there are enough reviews to pay attention to.
item_123_rank = [ 5, 5, 5, 5, 5, ]
item_123_avg = 5
Each time a "real" rating is added by your visitors, you push it to the front of the stack
item_123_rank = [ 8, 5, 5, 5, 5, ]
item_123_rank = [ 7, 8, 5, 5, 5, ]
item_123_avg = 6
I think I like that idea, caribguy! Me and my partner were sat down yesterday for about an hour trying to think of a system that would work only half that well!
Thanks a lot!
Her idea was:
Give the widgets a rank by score and a rank by review count and average those. I agreed it was a good idea, but couldn't see how to implement it without resorting the entire the database with each new review.