Suppose that I have 3 soda makers and I want to know their share of market. I am just given the total sales in a period (ex. this year) for all the soda makers and the total sales per soda maker. However one of the data points register very low sales and thus if I perform a simple average I will get a significant decrease of the share of market for that given soda maker. So I was wondering in my case how should I perform a weighted average (or if I need to perform something else) in order to avoid that data point that is skewing my data. As an example imagine I have the following data:

soda maker   Sales_in_store_1  Sales_in_store_2   Sales_in_store_3
coke          $500k                 $550k            $4k
pepsi         $400k                 $450k            $4k
bigcola       $100k                 $0               $2k

in this case I have a total sales of $2,010,000 and using a simple average I would get a share of market of approx 52%,42% and 5% for coke, pepsi and bigcola respectively. However the low sales volume in store 3 is skewing my data. My question is, should I use a weighted average to calculate a more realistic share of market? and if so, How should I do it? (I am confusing myself with which should be my weights)

Any insight that you could provide me I will appreciate it


2 Answers 2


This was better conveyed as a comment but alas I can't.

How is this "skewing" your data? What is your definition of significant decrease?

If you exclude Store 3, you get 52.5%, 42.5% and 5.0% respectively, up to a single decimal point.

If you include it, you get 52.4%, 42.5% and 5.1% respectively, which I would hardly call significant (about 0.2% decrease for coke, about 2% increase for bigcola). This might be subjective though.

Anyway, it really doesn't matter where those sales were done. You might move half of Store 1 sales onto Store 3 and get exactly the same market share. Which makes sense: the (overall) market share is defined by the comparison of the total sales.

(If I worked in bigcola, though, I would seriously investigate what's happening at Store 2...)

  • $\begingroup$ Thanks @polettix, I explained myself incorrectly saying "skewing" my data. Now that I think about it is that it is not representative of the percentage of market share that I want to study. by significant decrease I meant that because I have a low amount of pepsi sales in store 3 this might not be as representative as sales from store 2, so I was wondering if maybe applying a weighted mean I could decrease the significance of store 3 with respect to sales. However I don't know which would be my weights in this case $\endgroup$
    – Oliver
    Sep 5, 2019 at 0:45
  • $\begingroup$ @Oliver to be honest I still don't get what you are after, the fact that Store 3 has low numbers is already by itself decreasing its importance with respect to the other places - in fact, its impact on the market share calculation is at most in the first decimal digit of the percentages. Isn't this sufficient? $\endgroup$
    – polettix
    Sep 5, 2019 at 5:48
  • $\begingroup$ what I am trying to accomplish is to infer the share of market of a population by just analyzing 3 stores (what I mean is imagine that in Texas there are 1000 convenience stores, however I only own 3 of them and I want an estimate of specific brand share of market by just analyzing the sells of that brand in my 3 stores). I know that just using 3 stores is not something accurate but that's all I got. If you know another method or could reccomend me to do something else I will appreciate it. $\endgroup$
    – Oliver
    Sep 5, 2019 at 14:14

The average is telling you exactly what you want, the mean share for each soda maker. You could use a weighted average to account for the total customer number for each store, since it is expected that smaller stores will sell less products. However, I think it would be more informative for you to use the total (sum) share for each soda maker.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.