Why does this business have an average of 4.8 and not 5?

[irapasternack]

I was doing some research today and came across a business that has 5 reviews, all 5 stars. For some reason, their average is showing as 4.8, instead of a perfect 5.

Looking at some other businesses, I see one with 4 5-stars, 1 4-star, and 1 1-star. 25/6 = 4.1666666.  That would round to 4.2, but Google lists them as 4.3.

Is Google just bad at math, or are they using some other factor in calculating these averages?

[MiriamEllis]

Hi Ira!

Mike Blumenthal wrote about this same question at LocalU in 2014. See here. The discussion in the comments there is quite good.

[EBSAR]

Hello!

Google, on its help page says: We calculate an overall rating based on user ratings and a variety of other signals to ensure that the overall score best reflects the quality of the establishment.

What is going on?

When you use reviews for ranking, like Google and many other companies do, you have a problem. Is a business with eight 5-star reviews actually better than a business with 48 5-star reviews and two 3-star reviews? Should the one with eight actually rank higher?

Probably not. Sooner or later the company with only eight reviews is likely going to get some lower reviews.  Or so Google assumes and they don't want to give them more credit than is their due. What is a company like Google to do?

What they always do. Write an algorithm.

Google clearly is weighing 5-star reviews differently if there are less than a certain number and they are probably using an algorithm that is something like a Bayesian average.

What is a Bayesian average you ask? From Wikipedia:

A Bayesian average is a method of estimating the mean of a population consistent with Bayesian interpretation, where instead of estimating the mean strictly from the available data set, other existing information related to that data set may also be incorporated into the calculation in order to minimize the impact of large deviations, or to assert a default value when the data set is small.

In other words, when you have a very small sample (i.e. few reviews), Google is using its large dataset of review stars to model and estimate what your average would be IF you had more reviews. All very Googly, as they say.

Hope this helps!

Thanks.