Situation:
As you probably know as a subscriber to various services, we are prone to changing services whether it is phone subscription, streaming services, insurance, etc. and we do it for a lot of different reasons. This might be due to getting a better offer, customer service, and so on.
If you run a subscription business, you know that churn is a focal point that should always be minimized if possible. Churn and loyalty go hand in hand, and basically, the churn prevention/loyalty building should already start from the point of landing a new customer.
Why? Because the cost of retention is typically lower than the cost of acquisition. Therefore in these lines of business, it is important to focus on raising the customer lifetime value and thereby defend the market shares that have already been conquered.
But! At the same time, you don’t want to wake up a dormant customer.
The risk of waking a dormant customer (a customer who wouldn’t churn) is just as costly as not contacting a churning customer. It might even have the opposite effect of pushing the customer to churn.
Therefore the precision of identifying customers who are most likely to churn is critical.
Complication:
So where do we start? The data and system landscape are labyrinthine (yes, apparently that is a word – i.e. complex). The data scientists at the client had a hard time gathering data before analysis. Because what data points of the trillion ones available would have the right indicators?
Another predicament is, that systems and data sources change, and as well do data scientists – they start and move on to other tasks which meant in this case, Churn prediction was in limbo or in a state of bewilderment if you will.
Opportunity:
So what would you do from here?
The existing people-intensive method proved unreliable in the long term. Characterized by lengthy manual processes, dependency on people, and a changing landscape of data sources.
Do you hire more people? Well, that would only solve part of the problem.
Instead, they found some greener grass – a SaaS Machine Learning alternative.
This removes a lot of the manual tasks, automates processes, and quickly gets hold of the feature engineering and new data sources.
Resolution:
Using existing data such as usage, purchased services, activity, and responses to previous marketing efforts, Allyy.io predicts each individual customer’s probability to churn.
In less than 2 days after uploading customers’ data, the Client was able to predict accurately 76% of the customers’ churn behavior.
This helped the Client to make informed decisions on the next actions towards each Client while limiting the loss of waking up dormant customers.
This enables the retention team to focus their resources on the customers most at risk and offer them incentives to remain loyal.
Guess you could say that we made the churn situation, a little bit more hyggelig (cozy) and turned it into loyalty building.
