Though I come from the data science background, I am compelled to avoid
using word “Predicting customer churn”. Because “Prediction” is a machine
learning term and it tends to bring along the associated overwhelm for the
business executives. This overwhelm has been the number 1 reason why most
SaaS enterprises focus on adding new customers and tend to ignore the fact
that their existing customers are leaving them every month.

Whether you are in the SaaS business or not, as long as you are in the
subscription business your customers always leave enough clues for you to see
whether they will continue doing business with you or not in the next cycle. So
lets start with some of the excuses that keep you away from using simplest of
the techniques to identify your customers who are likely to leave.

1. We don’t have Data: I have heard most business executives have been
coached by the expert data scientists that they don’t have data required
to identify the customers who are likely to churn. Generally when data
scientists look at a business problems their perspective is to get 90%+
accuracy in this identification project and therefore they are likely to
demand advanced or derived data sets which are definitely missing from
the young business operations. Good news is that, you don’t need to get
the model 90%+ right on the day one, your business challenge is to get
some sort of indication of which customers are likely to leave you and
the false alarm is not going to do much harm. So to start with in order to
start getting your list you only need basic transaction data for each
customer and their historical status as whether they continued or

2. It takes a lot of time: Data scientists are still in the awe of each
prediction problem and therefore they are right when they tell you that
building a model ground up for your customer churn analysis is going to
take at least a few months. So let’s take a step back and see the
situation again. Where is the time needed if data is assumed to be
available? The time taken by the data scientists is mainly for evaluating
which model performs better and for tweaking chosen models to your
domain specific requirements. You are not in the business of data
science, are you? So, as a business owner you should be asking a
question who can serve me with the result in less than 7 days and keep
getting better at the prediction.

3. It costs a lot: Cost is a relative term and every time you say that “it costs
a lot”, you have a comparing mark in your head. I often suggest to the
subscription business owners to compare the cost of making customers
churn list with aggregated life time value of lost customers. The second
dimension is that your marketing dollars are going to be wisely spent on
the customers need it more than the spend getting averaged out on all
the end of subscription customers. This second dimension is extremely
critical for fast growing SaaS or subscription customers because
associating large budget for customer retention in not scalable unless
you know which customers need it the most.

There are a few other objections such as, “We are not in a position to do
integration”, “We are planning change our CRM”, “Our product is undergoing
updates, we can’t engage into this right now”. Customer retention and
understanding the churning customers is a pure business requirement and the
mean to this end is easy to use machine learning that can give you results in
less than 5 days. So I strongly suggest all the subscription driven business
owners to dive into getting your customer churn reports right away, it has a
direct return on investment. You will be surprised to see your average MRR
becoming highly stable.

Self-Promotion: Datoin empowers subscription (SaaS) business owners to
institutionalize customer churn listing within a matter 1 week, doesn’t matter
what domain you belong to and whether you have any one in data science on
your team or not.