Game of Indexes: A Song of Mx, Y, and Z
Note: an earlier version of this post was previously published on LinkedIn.
“The Scandinavian countries consistently rank on the top of Human Development Index. I should move there,” said Bill after the U.S. elections.
“The Dow Jones Index is decreasing. Is this yet another crash?” A nervous investor asked his fund manager.
“Our platform consistently ranks between number 5 and10 in the Verto Travel Index. We need to be in the top 5.” Future Unicorn Inc.’s head of growth and strategy told everyone in the company meeting.
“We are a better platform to find your target group than our competitors. Our composition Index for that target group is 200,” claimed Jenny from example.com.
“What’s up with these indexes?” I thought to myself.
All the above indexes are indicators of some entity. An index is generally used as a comparison and differentiation tool. A ranked list can indicate the position of an entity among its peers. The rankings, however, can be ordered on the basis of different measures. The construction of these specific measures provides the context upon which the index is to be interpreted. While measures can be atomic or composite, most indexes are composite measures.
For example, the Human Development Index (HDI), created by the United Nations Human Development Programme, is a composite measure derived from weighted combinations of measure of the levels of education, life expectancy, and other statistics of a given country. The Dow Jones Index (DJI) is a measure of the composite weighted share of prices of select commodities in the stock market. The Verto Index measures the number of monthly active users for digital services. At its core, the creation of an index gives us a quantified value that can be used for comparison and differentiation against similar entities. For instance, the HDI can be used to compare and differentiate countries; the DJI can be used as a reference to compare different stock portfolios and the Verto Index can be used to compare the popularity of apps or websites.
A measure, atomic or composite, has meaning only in contrast against a reference. That is why a simple ranked list is useful. This allows us to index a measure of an entity against the same measure of another similar entity. For example, we can see the number of monthly active users of digital apps on a ranked list and one can easily compare and contrast Facebook against Snapchat.
The Composition Index, Demystified
Advertisers often refer to a Composition Index, which indicates the representation of a target user group on a given platform compared to its representation in the total online population. It is scaled to 100, meaning that if 40-50 year old people have a composition index of 90 on example.com, then it implies that 40-50 year olds are underrepresented on example.com compared to the total online population. This is a bit different than the other indexes cited above, because the Composition Index has the reference also incorporated into the measure value. So, we already know by the value how the entity differs from the other similar entity (in this case, the total online population). Of course, this value can then be used to compare and contrast different platforms. If we create a ranking list of platforms sorted according to the value of the Composition Index, we would be creating an index of indexed measures. This allows us to determine where to best find the target group in the plethora of available online platforms.
So in a way, all these indexes are just measure values that can be sorted into a ranked list and/or used to compare and contrast similar entities. And some measure values are indexed (or “pre-indexed” if you may) measure values that inherently indicate the comparison against a reference such as the Composition Index. These are quite interesting and also very helpful in decision making. We need more of these indexed measures to answer complex business questions and for that, we use the following logic:
The Composition Index example above can fit into this formula as the percentage composition of target group in example.com indexed against the total online population.
How to Create Indexed Measures: Ask the Right Questions
At Verto Analytics, we create many measures to profile digital platforms and their audiences. In many cases, indexed measures are needed to generate better insights from the data. From the underlying engagement data of users on different platforms and the demographic data about the users, we are able to create insightful indexed measures by asking the right questions:
- Is the Facebook app more popular on Samsung Galaxy Phones? Calculate the reach of the Facebook app on Samsung Galaxy phones indexed against all phones, easy peasy.
- Do 24-55 year olds spend more time on Microsoft Word? Calculate the average monthly time spent per user of Microsoft Word among 24-35 year-olds indexed against all age groups.
- Do males have higher daily return rates in Tinder than females? Calculate the stickiness of Tinder among males indexed against females OR the entire online population (both provide the same information).
- Do Candy Crush players watch more TV? Calculate the average time spent per user watching TV on Candy Crush players (Candy Crush players would be Y in the above formula) indexed against general population.
The point is, you can create a wide range of indexed measures that indicate the comparative stance of a target entity compared against a referenced entity. And you can always resort to the ranking list of measures (whether indexed or not), to visually, manually, or mentally compare the measure value of the target entity against any other entity in the list, which, is also an index.