In order to use market share as a numeric adjustment variable, it is necessary to create numeric variables that estimate the market share for each respondent. It is commonplace for people to compute this incorrectly, so a bit more description is provided here.
Consider the example above, where we have three respondents who have rated their frequency of consuming two brands, Coke and Pepsi. With this data, the market shares of the two brands are 68.2% and 31.8% respectively (e.g., for Coke it is (10 + 5 + 0) / (10 + 5 + 0 + 0 + 5 + 2)).
A common mistake is to create adjustment variables by dividing one or more variables by the sum of the variables (at the respondent level). This is shown in the middle of the table. Why is it wrong? It fails to consider that respondent 3 consumes only one-fifth of the number of colas as each of the other two respondents. Consequently, the averages become 0.5 for each brand, and so Coke is under-counted. It should be borne in mind that this calculation can be useful – it is technically known as the average share of category requirements – but it is not market share.
The correct calculations are in the right table. The formula for each of the six cells is the value in the cell divided by the sum of the values in all six cells and multiplied by the sample size. For example, the 1.36 computed for Coke for respondent 1 is 10 / (10 + 5 + 0 + 0 + 5 + 2) * 3.
A practical challenge when using market share as a numeric adjustment variable is that the totals used in calculating the market share can change due to the weighting. For example, the total purchases before the weighting were 22, as in the example above, but after weighting this total is different. The way to address this is to recalculate the market share and redo the weighting a few times until it stabilizes.
Please sign in to leave a comment.