Find The 4% High Churn Risk
NGO
(1-2 minute read)
Situation:
We’re here to talk about churn and why customer retention is sometimes more important than actual acquisition. So let’s dive straight into some of the whys.
Studies show that acquiring new customers can be up to 5x more expensive than retaining an existing one. Additionally, research shows that increasing customer retention by 5% can increase profits anywhere up to 95%.
Now, these are hard-to-believe numbers and they are obviously heavily dependent on the case itself. If retention is achieved through a new offer you will see an increase in profits.
In this case, however, we will focus on a case with a Non-profit organization (NGO), that wanted to know what long-term donors were leaving, so that they could act before it was too late.
Complication:
There is plenty of data available, and you might be tempted to start guessing by looking at some of the easy-to-spot tendencies. As an example, you can look at how people pay, how long they’ve been a member, etc.
The ice cream, plus sprinkles of business logic is a great starting point. But as more and more competing NGOs are using more and more data points, the ice cream starts to melt and fall apart.
Alright alright, that analogy was a bit far-fetched, but the point is.. we as humans cannot process the amount of data and use it to make an up-to-date and accurate calculation and result.
Opportunity:
If you think about it – There is so much data that can play a role in churning donors. We just need to look at the big picture.
What about age? The journey to becoming a monthly donor? The payment form? Value of donation? The lifetime value of a donor? Address? Gender?
As you can see, it’s a gift that keeps on giving (if you eat data points for breakfast, lunch, and dinner/dessert). It is possible to build logic that takes multiple points into account, but it will demand a lot of time to build and maintain while very hard to make automated.
A better option is to add machine learning (ML) to the mix. That is the single best way to achieve non-biased up to date information that is executable.
Resolution:
Looking at the available data from the NGO, there were some determining factors. Time as a monthly donor and the payment method was the most weighted data points.
That meant that we would make 8 models to predict accurately within these groups.
The first one was making a model for automatically transferring donors and those who do it manually. Accounting for 2 ML models. Then we split these up into different groups of, let’s call it: duration of donorship.
The split we ended up with, was looking at the duration of donorship divided into +2 years, 2 years, <1 year, and <3 months.
4×2 models equal 8 models in total. 8 models that will automatically retrain as time passes and donors mature to new levels.
With this, we were able to identify risk groups across the entire customer base:
4% of high risk
27% medium risk
69% low risk
Taking this back to the beginning – there is now a solid foundation available, from which it is possible to effectively increase customer retention rates (possibly even by more than 5%).
People at high & medium risk of churning, you can strategically decide what to do with i.e. should they get an “I love you call”, letter, or something else?
If you want to know more, we’re your Allyy in AI and Machine Learning.
