Predictive analytics are a powerful and proactive tool to prepare our associations for the future. The following animation walks you through using a predictive analytics model effectively. If you’d rather see it in writing, the transcript is below!
As markets are changing, companies are seeking business in new areas of the world. Associations are left to react, looking into the future and trying to determine what products and services to offer and where. Predictive analytics can serve as a tool to help our organizations keep up and look ahead into the future. They can draw us a picture of what our activities and revenue will look like in new markets.
We conduct predictive analysis based on internal and external sources that describe future member or customer behavior—to help us make better decisions about where to invest our time and money to support our constituents and the industry where they work. More specifically, they can help us retain or grow our membership, and predict product sales and conference attendance.
If we imagine predicting conference attendance through a predictive analytics model, we can look at historical data that is descriptive of a member (years in the profession, distance from the event, interests), behavioral (event attendance, purchases, community participation, web click streams), industry (prices, sales, company investments, hiring patterns) and macroeconomic data (GDP growth, unemployment rate, currency conversion rate).
When we look at the data’s relationship with the member’s behavior, we want to define what factors are most important. Do more years in the profession or proximity to the event correlate higher with attendance at a conference? Then we start developing our model. We begin with historical data and divide it into training and test sets. Training sets are the portion of your data that you use to create the model. Test sets test the model to see if it generates a better prediction than random guessing. Our training set could have something like 80% of the data, while the test set contains 20% of the observations.
We run the sample data through an algorithm—regressions, classifiers, clustering algorithms—and then we develop our model. Once we have a model, we can run our test data to find out how accurate the model is. Our accuracy is the number of correct predictions about who would attend the conference over the total number of observations in the test set.
When we know we have good model, we can input our new data and determine who will be at our upcoming conference and where we should hold it—and in the end, how much revenue the event will bring our association.