Reducing Member Churn with Predictive Analytics with Fionta
In an ideal world, association members would remain highly engaged and renew every year.
Unfortunately, that’s not usually how it goes. How can we find those at-risk members to take action leading up to renewal? How can we understand the tendencies and motivations for those who renew and those who lapse?
During this webinar with special guest Katy Burns, CFO and analytics strategist at Fíonta, we discuss how data science and predictive analytics tools can inform your member retention and renewal efforts. We also share common methods to measure member engagement, talk about how predictive analytics tools can complement your membership retention efforts, and give you some pointers on how to get started.
-------------------
Transcript:
"...
6:03
why does this matter according to the asae center roughly seven percent of revenue is directly
attributable to members for trade associations and a bit lower for professional societies this includes
member dues conferences education webinars certifications etc
the danger in losing a member is the collateral revenue loss and
it's so much cheaper to keep an existing member happy and renewing than to bring on a brand new
one so how do we use predictive analytics to learn more about our
members and anticipate and avoid renewal risks i suspect some of you may be expecting
me to start with some code an algorithm to solve the problem but for those of you who might be
dreading that don't worry i'm not going to be writing code today this is more than just inputting
data into a black box and reading the output today i'm going to be walking you through the dating data mining
process and a type of modeling that we can use that's really tool agnostic
and then i'll hand it back to bill and he can walk you through how this process applies to acumen
let's start with defining the business problem
for the leaders in the audience it is our job to deconstruct these big problems into segments that can be
tackled by discrete analytical tasks and this process may be
possible with just your leadership team but i recommend that you consult with members of your organization who
really understand your data better yet have them at the table their perspective is super important because
they probably have a good sense of the strengths and weaknesses of the data and defining the business problem is
where a lot of the creativity and team collaboration can play a big role there's a few different ways to go about
this but i would like to leverage a framework we're going to use data to
generate predictions that can help us make better decisions to gain competitive advantage
and applying it to today's question let's leverage our member data to assess
who is at risk of churning in order to increase member retention
so this is what's called a supervised data problem and i want to tell you all a little bit about the difference
between supervised and unsupervised so we'll start with unsupervised
when you train a descriptive model that's called unsupervised learning
and descriptive models summarize data in new and interesting ways all features have an equal weight and so
no feature really is more important than the other and some examples of this type of model
are includes clustering and clustering divides data into groups or clusters that are based upon common
factors and similarities and your for your organization think about the attributes that you have
collected about your members like location gender age
job function perhaps credentials like cpa jd md industry
so we could do a pretty simplistic segmentation let's say by age and
bucket all of our members in decades but what about if we look at all of
these factors at the same time what other natural clusters might emerge and leveraging cluster analysis could
help inform perhaps what events you might recommend or even create new personas for your marketing campaigns
and another type of unsupervised model is known as association rules and a common
application is called market basket analysis so this is something that we actually all experience when we shop on amazon so first say for example i'm in the market for a new umbrella when i find the umbrella that i'm
interested in the other items that you might like section advertises for band-aids lip balm
and a nicholas cage coffee mug something in the amazon machine has highlighted that these purchases co-occur
at a greater frequency than you'd expect by chance
so for your organization you could use
association rules to recommend a series of courses that may might not be related
to you know overtly related but could perhaps lead to high engagement
okay so we'll jump to supervised learning so by contrast when a specific target or
variable is identified that's a supervised problem and predictive models are distinct in
that one variable is defined as the dependent variable and the other data points within the data set
known as the independent variables are used to predict the dependent variable
10:56
..."