Coefficient Quartile of Variation (CQV) to see how wide our pricing data is
- Do sales in Jakarta have any similarities between one sale and another?
- Will our customers prefer to buy a product at the same price over and over again?
- How different are the prices that we display on a product recommendation in one section?
- Etc.
In general, we usually use the standard deviation or variance to determine how wide or spread out our data is, where the closer the deviation is to 0 the data we have will be closer to our average value.
However, the standard deviation and variation will be affected by the outlier value.
- A high deviation is likely to have an upper outlier, on the contrary
- The low deviation may be influenced by the presence of lower outliers.
Therefore, in this article, I will introduce one of the values to determine a deviation provided that this value will not be affected by outliers, or in other words it will be more robust.
First of all make sure that Q3 and Q1 > 0, If the data says that a lot of data is 0, it can be said that we have a lot of freebies users, and why do we analyze user data whose price is 0 ?
For the pricing case, where Q3, Q1 > 0, and Q3 ≥ Q1, then 0 ≤ CQV < 1, with the following conditions:
- If CQV = 0 indicating that Q3 = Q1. For all business cases, CQV close to 0 indicates that our data is not wide or that the distribution is converging at a point.
Therefore, for pricing case, CQV is close to 0 indicating that the price is getting less wide.
Disclaimer
- Q3 != 0 and Q1 != 0. If the case used is that the price and we found that: (1) Q3 = 0 then it is clear that Q1 will also be 0. These users love freebies! and 0/0 is undefined anyway. (2) Q1 = 0 then CQV = 1 and we will find it difficult to determine the deviation because there is the possibility of infinite Q3.
- CQV is unlikely to be > 1 because if so then Q1 is negative. For price cases, this is not possible unless discount rates are considered :p
- CQV is unlikely to be < -1 because if so then Q3 is negative, which is not possible in pricing case.
So ? Wdyt ? Do you have any business case that suit with this calculation? Comment yaa :D
