XO Communications: Gaining insight into customer behavior to improve customer retention

Video Transcript

At XO, we believe that we have to constantly be innovating in how we go to market. Now that
can be applied whether how we formulate our sales strategy, how we formulate our customer
retention strategy but we always have to be looking at ways to apply today’s concepts,
methodologies and strategies to our installation and service of customers.

It is incredibly costly to acquire and install a new customer in business-to-business
telecommunications.

One of our major challenges at XO Communications not unlike any other telecommunications
provider is controlling customer churn.

Churn basically represents the concept of losing customers to competitors. We found a couple of
years ago that our churn rate was really hindering our ability to grow our monthly revenue.
There was a point in our history of trying to reduce churn where we had an epiphany and we
realized that looking at historic data and charts and graphs of churn reasons and things like that
just wasn’t cutting it. We really weren’t able to get to the true root causes of churn and address
those to help keep customers with us and better the customer experience.

There were some obstacles to overcome when we first started our churn reduction campaign. In
the beginning we started small with some proof-of-concept projects to see if Predictive Analytics
really could identify customers that were more likely to churn and we asked ourselves a question,
what if we could do this on an entire customer base.

When we went out to select our Predictive Analytics software, we wanted to choose a vendor that
we knew would be there for the long haul and that had a very robust solution. So we chose IBM’s
SPSS Modeler platform. It was robust, it had a great user interface and it was something that we
could drive ourselves when we brought the model in-house.

The predictive churn model that we deployed at XO Communications allowed us to talk to
customers before it was too late. So in other words, the model told us that they were very likely
to make a decision to go to another provider but it allowed us to call them before they signed a
contract and committed and made a decision which might result in some kind of a contract
penalty or something like that.

After we implemented the full-blown program we were pretty pleased with our results. A third
party came in and evaluated our program and found that we had 376% return-on-investment, a
payback in 5 months and a net savings of 3.8 million from the project. Said another way, our
churn rates almost halved since we launched the program.

We found that we enhanced our revenue retention rates by 60% and saved the company over 13
million in revenue each year of the program.

One of the things we found with our predictive churn model is that it had some side benefits to
just giving us a list of customers that were most likely to churn. It actually allowed us to identify
the root causes of that churn and eliminate them from the equation.

Modeling customer behavior can be pretty tricky. The thing to be careful of when building a model
is to let the data also tell you the story but always be checking the data back because it’s not
always going to be an intuitive situation, it’s not always going to be the 3, the 5, the 10 predictors
that you think they are going to be.

In today’s business environment, companies obviously need to be concerned about their own
bottom line but they need to be more concerned about their customers’ interests as well. So
Predictive Analytics is one way that companies can in very cost effective way get in touch with
their customer insights and drive them into actions that are win-win for both the company and the
customers.