How can we measure the improvements made to an account that arise from optimizing the account structure? Sometimes these changes obviously lead to an obvious increase in revenue, but sometimes they have more of a "behind the scenes" impact by reducing the time taken to manage the account or enabling further optimisations to take place. We can borrow an idea from economics to quantify the effect of structural changes on an account; although I will say now (and potentially save you the time for reading the rest of this post) that this is a pretty bad case of using derived metrics with little real/dollar meaning.
The Gini Coefficient for a country or society is a measure of how equally or unequally incomes are distributed. A value of 0 means that everyone has the same income and a value of 1 means that only one person has any income - maximum inequality.
You can see visually what the Gini Index measures using the following graph: The Gini Index is the ratio of area A to the total area of the triangle.
How is this useful for PPC?
In my last post about AdWords Ecommerce campaign structure I talked about 3 stages that accounts go through as PPC practicioners alter and optimize the structure.
One of the things people think about when making structural changes to their account is the idea of segmentation; splitting out different qualities or values of traffic into separate ad groups and campaigns in order to best use the different targeting and bid settings available.
At one end of the scale we have a poorly optimized account; here the Gini Index is high because a few keywords or ad groups generate a lot of the revenue. At the other we have an account where geo-targeting, day parting etc. are used to such an extent that every keyword has an equal share of revenue and the Gini Index is 1. (This is likely an impossible and undesirable situation but the same could be said of a society where everyone has the same income).
I cannot think of a situation where making a positive change in account structure would result in a lower Gini Coefficient. Thus, I put it to you that monitoring the Gini Index for your account is a useful way to monitor the health of your account structure.
Calculating a Gini Coefficiant
This is probably easier than you think. This example is for campaigns, but it will work for keywords, ad groups or adverts as well.
- For each campaign, calculate the percentage of revenue from that campaign (for example, the percentages might be 5%, 10%, 10%, 25%, 50%)
- Order the campaigns from smallest to largest revenue and calculate the cumulative percentage for each campaign (using the numbers from above this would be 5%, 15%, 25%, 50%, 100%)
- Also figure out what the cumulative percentage for each campaign would be if revenue was even (so if you have 5 campaigns the cumulative percentage goes 20%, 40%, 60%, 80%, 100%)
- For each campaign, find the difference between the cumulative percent revenue and the cumulative even split percentage (15%, 25%, 35%, 30%, 0%)
- Now divide each of these numbers by the number of campaigns (0.03, 0.05, 0.07, 0.06, 0)
- Add up these numbers (0.21)
- Double it (0.42)
What is Good or Bad?
Verticals which are very "head" orientated can have accounts with a very fine grained structure that will have a very unequal Gini Index. Accounts that focus on the long tail will be more equal.
This is just another number, in our vast toolkit of numbers, that can be used to guide our actions. As with all metrics it is useless without context. However, if one month my Gini Index was 0.42 and the next it was 0.83 I'd certainly start looking at my account structure.