Many associations are still addressing daily inefficiencies and systematic problems with one of two extremes: short-term bandaids or lengthy "to-do" roadmaps. With these traditional approaches, it can take months or years to resolve pesky challenges that could be squashed in hours, days, or weeks with modern analytics techniques like propensity modeling.
You need a better approach.
If your association wants to do more with your member data, then it’s time to leverage a technique that corporations have used for years: propensity modeling. You may have also heard it called predictive modeling, predictive analytics modeling, forecast modeling, and even predictive personalization.
If you're not applying predictive analytics to the mountains of member data stored in your database, you're leaving value on the table. The magic lies in propensity modeling — the hallmark of better decision-making.
This blog post will walk through the steps to get started with propensity modeling. Here's how it works in six steps.
The first step is defining your organizational objectives, the challenges you want to address, and the insights you’re looking to gain.
Let's say your goal is event attendance. You could use your model to predict future attendance on an individual basis based on past attendance or whether someone new is likely to attend for the first time.
Member retention is another goal you could target through predictive/propensity modeling by honing in on the renewal likelihood at the individual level—and the reasons why.
You could even target by predicting who is likely to become a member in the future based on their types and level of engagement.
Other potential goals include: predicting and improving exhibitor and sponsor churn; recurring donations; engagement in your courses or credentialing programs; or the purchase recommendations displayed via content intelligence.
Mapping out the strategy informs the kind of data you'd want to include in your propensity model. It also helps to program the machine learning functions by telling it how to train and improve itself over time.
Say, again, that event attendance is our goal. Then, at the very least, we'd want these data points: member ID, registrant history, and attendance history, as well as financial transactions.
Now, if the goal is member retention, then we'd want to plug a slightly different mix of data assets to run the predictive/propensity model. Those data points may include member ID, renewal history, and purchase history. Chapter enrollment, committee engagement, and subscriptions would also be beneficial to include in propensity modeling software.
Before you feed the data into the model, it wouldn't hurt to review the consistency, accuracy, and completeness of the identified data assets. Many associations think they have dirty data or useless information, but oftentimes, this is not the case. Even in those rare instances, you can combine relevant third-party data for enhanced accuracy or as a starting point. Also, keep in mind, your organization's data quality will improve the propensity modeling journey.
This step entails determining the easiest and quickest way to get the data assets identified in Step 2 out of your database(s) and into the model. You may have this data scattered in different systems, so you need to figure out where the relevant information "lives."
How you then get this information into the model depends on the propensity modeling technology you're using, without interfering with data privacy rights. It could be as easy as matching up the fields in the interface. Or, you may have to dump your data onto an Excel or CSV download, and then upload it.
This step involves validating that the model will deliver more exact personalized insights each time it’s fed new behavioral data. During this phase, organize your target audience from least likely to most likely to perform a certain action, and begin to group cohorts of behavior.
The goal is to be able to accurately forecasts the propensity to purchase, convert, churn, engage, etc. on an individual-to-individual basis.
In other words, you want to make sure the model can train itself to give the 'most fit' likelihood score (aka propensity score or probabilistic estimation) of someone taking a certain action as well as the key indicators of that behavior.
Next, run the model on top of data from your next membership cycle, event, fundraising campaign, or whatever your target instance.
This will allow you to identify who’s likely to renew, attend the event, donate, get certified, accept a personalized offer, purchase a publication, subscribe to a newsletter, or whatever the target action at hand.
One of the main purposes of propensity modeling is to limit advertising and marketing waste on segments of individuals who are very unlikely to convert no matter what. We also want to limit spending on segments of individuals who always attend your events like clockwork or those who always renew, for example.
The reasoning here is that you don’t want to devote precious dollars to your entire audience base when you could instead direct those resources toward folks that have the best chance of converting into paid registrants, recurring members, etc.
The last part of the cyclical process is working together with the appropriate team members in your organization to develop targeted strategies for the identified goal(s) based on the actions, attitudes, and habits of individual members.
Propensity modeling is a cyclical process and continuous journey. That is, the model improves over time as more data is gathered, allowing you to hone in on which targeted outreach strategy works best for each member, attendee, donor, subscriber, or audience in question.
Regardless of your association’s size, you can run your data through these iterative models to determine member-specific events. For instance, the likelihood of purchasing a book, registering for an event, or even renewing membership. With this personalized insight on hand, you can focus your efforts and drive greater ROI.
After a few iterations, you should be able to build out individual profiles and pair those profiles with the best outreach approach, one-to-one.
As we look to the future, you want to keep progressing your organization along the propensity modeling journey. Ultimately, you're working towards data being embedded into every decision, interaction, and process.
Are you ready to do more with your data for less? Need help getting started with propensity modeling or automating the process? Get in touch with us to start making the most of your data.