By Willow Moellering, Global Head of Customer Success, Lift
For many years I’ve worked by spearheading Customer Success at various companies. I’ve been responsible for ensuring clients realize value and want to grow their relationship. For many years I’ve used Healthscores and Segmentation combined with my growing experience as well as gut feeling to do this job. However this year, things are a bit different. I was super privileged to join Lift as Global Head of Customer Success and I realized there is a better way to predict and forecast churn rate and possible upsells: applying AI to predicting highly dynamic client relationships. Here are my 2 cents on how that works…
As 2020 is coming to a close, I’m sure you are like Lift and have started to create revenue forecasts to understand churn rate and possible upsells. In years past, I have used the traditional method of Healthscore & Segmentation to calculate targets, but this year I started to wonder if there is a way technology can assist in this process.
Can we use AI to calculate risk and health in a way that directly relates to revenue? But will this truly give us the pulse of customers in upcoming years, especially in such unprecedented times?
The B2C world has been using predictive analysis & AI to predict churns successfully for years: in a research article published by Financial Innovation in 2016, data scientists were able to to predict customer churn for electronic banking services by 92% by using 3 key components, first being a data set on customer satisfaction¹.
We know it works for B2C, but does it work for B2B. In a business case review from McKinsey on the AI Frontier, business cases related to “Churn reduction” found that text data sets correlate most directly to churn². This aligns directly with our approach to use evaluations and AI analysis to predict churn and assist in forecasting.
As an evaluation platform, we survey both Clients & Employees, and based on those survey results our AI tool uses the data points below to calculate likelihood to churn.
- Client scores
- Team scores
- Comments (semantic analysis + amount)
- Response rate
- VIP answers
- Question interdependency
- Client/Team score interdependency
- Response time and pattern
- Relationship tenure
- Participant proﬁle (scraped data)
While we use multiple data sets, we focus on Comment Analysis to create predictions with a 95% accuracy. Using this prediction, I can tie churn risk to specific clients and segments to understand 2 key elements in 2021 planning: revenue forecasting & customer mapping.
As I plan for 2021, I need to understand more than who is likely to churn, but also what effort I need to put into each customer to keep them from churning. Our tool creates a Lift Retention Score that calculates the likelihood to renew based on the AI, and then buckets customers into 3 buckets based on that score: Red, Yellow and Green. Overlaying those buckets with my current customer segmentation based on size, length of contract, expansion opportunity & contract history, I am able to see predicted churn, impact on business, and the customers within each segment for which we will need to initiate a save playbook.
Based on the data analysis, I am able to create a realistic forecast and budget based on customer needs and revenue risks rather than relying solely on CSM & Account Leaders “gut-feel”