Predictive Analytics 101

Predictive predictive predictive! CoolaData uses predictive analytics to anticipate what users are going to do in the future, moving beyond the analysis of customer behavior and actions which took place in the past to improve current engagement and enable the prediction of user lifetime value or churn.

predictionGenerally speaking, people who currently behave in the same way as other people did in the past, will perform the same future actions as the original group performed in the past. Taking shopping cart abandonment as an example: say your average abandonment rate has been 60%, but in the past people who were associated with three specific variables only had a 40% abandonment rate. We can assume that other people who can today be associated with those three variables will probably show the same 40% abandonment rate.

These variables could be demographic, like gender and age, or behavioral, like purchasing specific items or clicking on certain links. At CoolaData these variables are called properties. We determine what a user’s future behavior will be based on the behavior of others who had similar behavioral properties.

The human brain can analyze a maximum of 5-6 variables. But it can’t analyze the value of 50 or 60 variables and it also can’t see the combination, or correlation between variables. Trends are a good example of variables humans use. If a stock’s value is going up, we generally assume it is a good investment. Or at a casino, games are designed to pay out small amounts at the beginning so that you stick around and lose large amounts in the long run.

CoolaData’s algorithms examine as many properties as possible and then add in the relations between them. We look at who abandoned in the past, focusing on the most influential properties. For example, the number of actions a user made usually impacts abandonment: the more actions performed, the less likely the user was to abandon. We then pick the most relevant properties and use them to predict future behavior.

The vast number of variables taken into account is what makes this type of analytics so valuable, because it increases the prediction’s quality and confidence level. Our advanced BI technology produces a predictive score for each customer or cohort. We then build our predictive models based on these scores to answer questions such as: what’s going to happen with a certain user who logged into a game and then left after the second level?

Below is a basic matrix of scenarios and actions which are useful in business terms. Typically you’ll want to focus your efforts on keeping customers who spend a lot of money and have a low probability of abandonment (the upper right corner).

The three main types of predictive models are:

  • Churn – who will stay or who will go?
  • LTV – what will my August revenue be? 
  • NBO (next best offer) – promoting merchandise a user is most likely to purchase

Each user’s predictive score informs actions to be taken with that user, so that predictive analytics optimizes marketing campaigns and online behavior to decrease churn and increase user response rate, conversion and clicks. 

Download our dedicated white paper for more on predictive analytics.

Share this post

Leave a Reply

Your email address will not be published. Required fields are marked *