Rating statistics change

[ozobpozz]

Until recently, we have had high average, average,
and low average numbers on all aspects of a
vendor's rating. Even with the limitations of
such figures, they at least gave us something
approaching a statistically meaningful way to
interpret the raw numbers. Now these overall
population figures appear only on the overall
ratings.

This is unfortunate because the high average,
average, and low average numbers do not apply
to the subcategories, which may have very
different interpretations. A value of 7.12,
for example, may mean somewhat below average
in one subcategory but well above average in
another.

As a tradeoff, we are given ratings values over
the past six months, but these figures are
almost always based upon a small number of samples
and vary widely from month to month whether or
not the quantity being measured doesn't change
much. In other words, most of the variation is
statistical noise.

I would like to see the old figures brought back,
or even better yet, a similar rating based on
something with even more statistical meaning,
such as quartile boundaries. The first quartile
would replace the low average, the median
replaces the average, and the third quartile
replaces the high average.

ozobpozz

[Warthog]

I see.

[DocHopper]

Maybe you can translate wart... I have no clue what he's talking about....

[daveleau]

He wants a statistically based rating system in RR.com and an option to see each store's performance by quarter.

I understand. I have no decision making power in this area, but I understand. Scott, Mark and the others involved will see it and give it though.

Thanks for your input.

Dave

[DocHopper]

Thanks for the translation.
why didn't he say that in the first place?

;-)

[Warthog]

lol

[ozobpozz]

Well, not exactly. I didn't say anything about
wanting statistics by quarter-year. I just
wanted some way to tell what was supposed to be
good, bad, and average. What you had for a
while were high average, average, and low average
scores for each rating that could be used for
that purpose. Now only the overall rating has
this feature. Unless you happen to remember about
where these different averages fell for the
other performance measures, it is difficult or
impossible to judge the performance ratings of a
vendor based on the numbers given.