User Metrics

User Metrics are tools for your marketing, BI and community staff to answer these questions:

What have the users done? What are they likely to do? Which users will do what?

and, when used with Segmentation:

What was the result of our efforts on their past and future actions?

The most common categories here are Churn (Quitting) and Conversion (to go from not spending to spending), though Katana will eventually add Virality, Time Spent and Advertising support to User Metrics.

User Metrics are broken down into three sections: Global, Historical and Projections

The Global page for any given category shows you what all of your users, or groups of them, have done:

Or, you can see what they are likely to do:

This projections graph has upper and lower bounds known as "confidence intervals." These mean that 95% of the time, the value will be between the upper and the lower line, and the mid projection line is the best guess. Consider this graph analogous to a projected path of a hurricane from weather forecasters: it's probably going down the middle, but be prepared for anything in the cone.

The Historical page is straightforward. It's simply what every user has done, both on a case-by-case basis, as well as with summary statistics at the top. If you segment, this top summary table can be very insightful. The bottom tables give you snapshots of your top and bottom users, but you can also use the search field to find a particular case. Or, if you use the exporting option, you can select a range of values from all of the users.

The Projections page is very similar, but tells you future behaviors rather than past ones. Just like the Historical page, it gives you a summary table as well as your top and bottom users, and exporting and sorting, etc.

Projections are different, of course, and there is never 100% certainty. In fact, you can see the confidence level of the data on the page at the top right.

This number is an F-Score, and it's telling you how much you can trust the individual projections. This number gets larger when you do two things: #1, give the system enough time and data to accomplish the "learning" part of machine learning, and #2, implement all of the calls you can. The basic set of calls we recommend will usually get you decent confidence, but the more optional calls you implement, the higher this number will typically rise.

If you use Segmentation on this page, it will combine with the predictive analytics to tell you which groups of users are likely to take actions in the future, and how much. It may be that those who have experienced different content, CRM, etc., will churn or convert in the future at different rates.