Retailer Rating

The scoring system for the retailers seems to be a little unfair. From what I see, the overall score is calculated by the number of ratings multiplied by the average overall rating. This means that the overall rating is not an acurate score of the quality of the retailer, just the number of customers. For example, a retailer with 30 ratings all of 5 of 5 would score the same as a retailer with 50 ratings all of 3 of 5, or worse, a retailer with 150 ratings all of 1 of 5. While I understand there has to be some weighting given to the number of ratings, or else a retailer with 1 rating could jump to the top of the table, the way it is at the moment the overall score isn't representative of the best rated retailers.

On the other hand, I'm loving the idea of it!

Come up with a better system then. The top retailers are shown by average rating and by overall score on the left in the stats section. Overall Score is useful as it is also an indicator of how many people use a retailer as well as how well the retailer performs. It is therefore both an accurate score of quality and frequency. It shouldn't be taken on its own as an indicator, it should be taken in conjunction with the other figures such as average rating.

if you can think of a more accurate way to indicate retailer scoring then please let me know.

OK then! :D

The best way I can see to work out would be to use logarithm of a power of the average score, multiplied by the number of ratings. Using my examples above, and a cube, for 30 ratings of 5, we get a score of: 8.22, 50 of 3 scores: 7.20, and one rating of 5 scores: 3.27. Of course, the higher the power you use, the more important the rating becomes, and the less important the number of ratings. My choice would be a power of 5 or 6. Of course, incorporating a formula like this into the automatic system is far beyond my already over-stretched brain.

Believe me I don't mean to be over-critical of what is an amazing system, something we've been needing for a long time!

The problem with appplying something like that is that at the moment all the calculations are being performed in a SQL query so it is relatively quick and has little impact on the server. Using your approach means returning multiple rows in the query and then looping through them to calculate a result in code which takes a lot longer. In addition, the end users inevitably won't have a clue how to correlate the overall score into a meaningful value i.e. if they have a current score of 8 and someone rates them it won't be easy for them to see how that rating affects their score or what effect future ratings will have

Your input is valued but the HEXUS view is that the number of ratings is just as important as the rating itself hence the easy calculation. It all comes down to individual preference though.

Apologies for the abruptness of my earlier post, I'd just been round Asda and was ready to kill :)

Thanks for using the system and feel free to continue to give us feedback, it is much appreciated