Using Predictive Analysis

Data Insights is currently in Beta. If you are interested in opting in, please submit this form.

Table of Contents

Introduction

The Predictive Analysis for Data Insights uses data modeling algorithms to predict metrics such as retention and monthly revenue. It will also provide segmentation analysis and grouping members by engagement data, to show patterns, trends, and predictions that are difficult to see simply by looking at charts and graphs of static data. It includes three tabs along the top:

  • Membership Retention Analysis
  • Membership Segmentation
  • Revenue Forecasting

Back to Table of Contents

Membership Retention Analysis

At the bottom of this tab is a list of all member profiles. It includes the “Predicted Outcome” which is either “Predicted Drop” or “Predicted Join.” Next to that, you will see the probability of that outcome as a percentage.

This uses a binary classification model to run through dozens of data points, comparing them to the simple determining factor of whether each profile dropped, or retained as a member after joining.

Some of the key factors used to determine this predicted outcome:

  • Location/Demographic information
  • Event signups
  • Referral data
  • Member login data
  • Contact records created/Emails read
  • Length of time as a member
  • Dues and non-dues Revenue
  • Open/outstanding invoice status

You can download the detailed data used to determine these results by clicking Download Source Data Used in Analysis.

At the top of this screen, you’ll see a list of the five data points in the analysis that appear to have the highest impact on whether or not a member is predicted to drop, and how “sure” the data model is of that prediction:

At the bottom of the panel are buttons to move between pages and see more data points.

These data points can be configured in the Data Insights settings.

Back to Table of Contents

Member Segmentation

Similar to the Retention Analysis, this screen analyzes many factors on each member including demographic data, revenue and payment information, and engagement information.

This uses a K-means Clustering algorithm to create groups of profiles based on similarities and patterns among the data, which helps you clearly identify the “types” of members you have. These groups and patterns help you strategize your efforts differently for different types of members.

For example, if you want to increase attendance at your events, reviewing the data patterns of the cluster with the highest rate of event signups will give insight into how you can target profiles most likely to attend your events.

The data is normalized into “points” on a scatterplot to help visualize the spread in values within each group, which are all different sizes depending on the impact of each data point.

These data points can be configured in the Data Insights settings.

If you click on one the cluster names in the legend at the top, that cluster will be removed/re-added. This can be helpful if you are trying to get a cleaner view of a particular cluster.

Move the mouse over a point to see the name and ID of the profile being represented. In the example below, the mouse was moved over an area with several points, showing a list of profiles.

On the right, you can see a list of the five data points in the analysis that appear to have the highest impact on determining the cluster of each profile.

At the bottom of this panel are buttons to move between pages and see more data points.

At the bottom of the screen is a list of clusters which shows the average values for each cluster:

Two things to note:

  • There are 6 clusters in this example, but you may have more or less.
  • This screenshot shows 9 columns, but there is a horizontal scroll bar to see more.

If you click on a cluster in this list, the profiles within that cluster, along with predicted outcomes, and other information, will appear at the very bottom:

There is also an option to download this data.

Back to Table of Contents

Revenue Forecasting

This screen shows the next 12 months of predicted revenue, broken down into three segments: Dues, Non-Dues, and Events, with a sum of all three for each month.

There is also an option to download this data.

The line graphs below the chart display all revenue, as well as forecasted revenue from Dues, Non-Dues, and Events for each month.

Back to Table of Contents

Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request