Membership churn is a major challenge for associations, yet often organizations simply accept their attrition rates rather than proactively addressing the real reasons members drop. It can feel impossible to figure out why members become ex-members, but predictive analytics offer a data-driven approach to finding at-risk members before they leave.
In this post, we’ll explore how associations can use engagement scoring, machine learning, and predictive modeling to predict and mitigate renewal risks. While we’ll be talking about Salesforce, much of what follows here is tool-agnostic, allowing you to use technology your organization already has in place.
Member Churn Is Often An Unsolved Mystery
All kinds of association revenue are directly related to membership. In addition to dues, members also attend events and conferences, pursue certifications, pay tuition for continuing education opportunities, and more. According to ASAE, 70% of association revenue comes from membership-related sources. And we all know it’s more cost-effective to keep current members rather than going out to find new ones.
Every association knows that member retention is essential to organization and mission success. However, associations still rely on historical attrition rates for planning purposes rather than actively working to prevent churn. More organizations try to treat the symptom – the retention rate – rather than the root cause of the problem.
Luckily, technology can offer valuable insights to help your organization improve retention rates and overall member engagement. We’ll start with engagement scoring – a basic but limited approach.
Using Engagement Scoring to Identify Members At Risk
Perhaps you’ve already used engagement scoring, which is a straightforward way to use data already at hand to predict member churn. Engagement scoring works by assigning values to specific member behaviors – think membership, tenure, event attendance, volunteerism, and community participation, for instance – and then calculating a composite score to rank engagement levels.
Engagement scoring is simple to implement and understand, and it helps you identify members at both ends of the engagement curve. However, this type of analysis has significant limitations as well. Middle-range scores are inconclusive, making it difficult to determine the risk of churn; only individual behavior is considered, not trends across the entire membership, and the tool won’t adapt over time or learn from patterns in member behavior. Certainly, engagement scoring is a good place to start, but there’s a better tool.
Moving Beyond Engagement Scores with Predictive Analytics
Predictive analytics uses machine learning and historical data to find patterns in data. In this case, we are concerned about churn risk, but you can use predictive analytics combined with your existing data to analyze all kinds of issues. Unlike engagement scoring, predictive models can weigh multiple factors and uncover hidden risks. It’s a much more sophisticated type of analysis than engagement scoring.
Machine learning offers multiple advantages in membership management. The tool identifies early warning signs before members disengage, learns from patterns across thousands of data points (not just a single score), and continuously improves as you add data. For example, risk profiles may differ between your members who are active online and in real life. Predictive modeling examines both behaviors rather than treating them equally to create a more nuanced risk profile and give you a better shot at engaging members where they are.
How Predictive Models Identify At-Risk Members
Of course, predictive models aren’t created in a vacuum. It’s essential to spend time identifying and cleaning the data you intend to use to build the model, as well as deciding the target question that is most useful for your association to answer.
This can get very complex, of course, but here’s the basic data that you could use to build a model. All of this data already lives in most associations’ databases or AMSs, but it needs to be clean and correct for the best result:
- Membership tenure
- Event attendance history
- Volunteer and committee participation
- Online community engagement (logins, posts)
- Payment and renewal history
Using this dataset, an organization can put a model in place that learns from past churn behaviors to predict future risks. In addition to the informed predictions that this model can make based on thousands of data points, the outcome is that you can personalize your retention strategies based on identified risk factors for an individual member (or group of members).
Further, once the model is built, it continuously adapts to changing behavior based on the new data it consumes. Salesforce Einstein Prediction Builder is one example of a platform that allows associations to build no-code predictive models, but there are others as well.
Using Predictive Insights to Reduce Churn
Of course, the point of building a predictive model isn’t just the joy of building it. You want the data to inform your strategies for retention at a granular level. The model finds your at-risk members, and then you can build a comprehensive strategy to increase retention. Tactics you might employ include:
- Personalized Outreach: Automated emails or personal calls to high-risk members.
- Exclusive Engagement Opportunities: Targeted invitations to webinars, networking events, or mentorship programs of specific interest to identified members.
- Content Recommendations: Using data-driven insights to serve relevant articles, courses, or event recommendations.
- Financial Incentives: Early renewal discounts or flexible payment plans for at-risk members.
One common scenario would be to set a range of churn risk by percentage, say 75-100%. You can then prioritize outreach to the members in that group, using your time and resources budget to focus retention efforts where they matter most.
Lessons from 2020: Adjusting for Unpredictable Member Behavior
The COVID-19 pandemic drastically changed how members engage with associations and vice versa. Even five years after the start of the pandemic, it’s important to consider these lessons. Is your organization much more likely to host virtual events now? Are your online learning and certification offerings still more popular than your in-person classes? Has your geographic base widened or narrowed?
The lessons learned from the pandemic can also inform your targets for the organizational challenges that inevitably arise. When building or updating your predictive models, you can segment data into pre-challenge and post-challenge datasets, adjust your models to reflect new engagement patterns, and keep feeding new data into your models to continuously update evolving risk factors, no matter the challenge you face.
Next Steps: Implementing Predictive Analytics in Your Organization
Let’s get started!
The very first step is critical to organizational success at all times. Name the member data you have already collected and indicate where gaps in member knowledge exist so that you can start collecting that data as well. Standardize and clean your data so that your model has accurate information to work with, because we all know the old data maxim, “garbage in, garbage out.” Deduplicate member records, standardize state abbreviations within member addresses, ensure critical data isn’t kept in text fields, etc. If necessary, combine disparate data sources depending on the tool that you’re using or correcting.
Next, experiment. Use Salesforce Einstein Prediction Builder (or the AI tools you have at your disposal) to create initial models, and run small-scale pilots to validate predictions. In this case, you are looking to prove what you already know is true, so that you see the model is working as you want it to. You are also evaluating your dataset for balance to ensure that you are stripping as much bias from the results as possible.
Once your pilots are successful, you are ready to put machine learning to work for your organization. Identify the critical target questions you need to answer with regard to churn, and run your predictive models on an org-wide scale. You can then develop dashboards that allow your leadership to easily visualize and track churn trends, and empower membership and marketing teams to act on high-risk predictions.
Predict and Prevent Churn Before It Happens
Machine learning provides a deeper, more accurate view of renewal risk compared to traditional engagement scoring. Predictive analytics gives associations the advantage of informed initiative—instead of reacting to lost members, you can anticipate and address risks before members churn. By finding at-risk members early, your association can focus its resources on personalized retention strategies that keep members engaged.
The key takeaway? Stop guessing and start predicting.