Research Tip: How to easily spot "Real effect" of variable changes vs "Mix changing effect"

In the following article, we will be reviewing some techniques to analyze and present data. These are very suitable for explaining compound metrics when the mix of single elements is taken into account. Especially, when those are defined by a weighted average using the Share of Market to ponderate.

How to compound metrics based on Share of Market?

In order to compound metrics that reflects better the effect of the single elements, we could use the weighted average instead of the simple average. For example, if we want to calculate Purchase Intention for The Coca Cola Company (as manufacturer), initially we have two possibilities:
In the simple average calculation, we don't take into account the weight of single brands and we bias the entire metric. If we take into account some weight, the metric will be healthier in terms of reflecting the reality. It is not mandatory to use the Share of Market, but we will use it just as an example since the idea is to show how to separete this effect when measuring changes of variables within different periods.
With "Fi" being the value of the Purchase Intention for the "i" brand itself, and "Si" the correspondent amount of Sales from the period.
For more information about the Weighted Average, please refer to the following link

How changes in Mix can impact on variables variations?

As observed in the following example, the average can change due to variations in the mix of Sales (weighting variable)
The increase of sales of Coca Cola boost the metric since this is the brand with the best Purchase Intention. This is a bit tricky since there is not any change of each of the brand from the company. As conclusion, if we are not careful this can lead us to think that Coca Cola Company is increasing it Purchase Intention,

How to separate the change of the variable from the effect from mix?

After the previous introduction, we now are ready to dive in the main paragraph of the article. The idea is to spot the real effect. By real, I mean the pure effect (or so) of the variable, separating and measuring the mix effect. In order to do that, we need to order the spreadsheet in the following way:

To fill the grey shaded cells we use the following formulas:

To complete the example, here are some possible insights and conclusions about the methodology (not on the data, remember it is fake data).

  • In order to check the variation, both must sum up the same
  • The same analysis of brands/manufacturer can be done with manufacturers/market
  • The 0% of mix effect does means that this brand has evolved in the same rate than the market

As always, we really hope you enjoy our article. Thanks for taking the time to read!

1 comentario:

  1. Thanks for sharing the formulas and tips to find the weighted average of compound metrics when the mix of a single element is in the count. This helps to calculate the market trend and plan strategies accordingly. Being a marketer in the section of phd dissertation writing services, We have a lot of data on users from the different regions provided by google. We need to analyse them and target potential users for business growth. The weighted average formula and its implementation are helpful in classifying the data.