Slaying customer churn with a better churn prediction model: A customer success case study
Success Stories | Technology

Slaying customer churn with a better churn prediction model: A customer success case study

Business Issue

Customer churn is the fire-breathing dragon of SaaS companies. The cost of new customer acquisition, combined with long conversion timelines, makes customer retention essential to sustained growth. It’s why Customer Success teams work tirelessly to slay the churn dragon by ensuring their customers’ success.

Our client, a leading SaaS company, needed to proactively respond to signs of stagnation and customer churn. However, their churn prediction model was primarily reactive in that it was not providing visibility into the root causes of customer churn. The customer success team needed to know when and why a customer was at risk of leaving in order to preempt it.

What We Did

Escalent conducted an extensive audit of the client’s existing churn prediction model. We realized it was:

  • Not accurately predicting churn for large accounts
  • Not good at predicting churn timing
  • Unable to show event root causes
  • Unable to account for market swings and seasonality
  • Not factoring customer issues/escalations, CSAT data, and account forecasts provided by customer executives into churn predictions
  • Time-consuming to maintain

Overall, not a great prediction model. We really needed to know what types of events, actions, or inactions are early indicators of churn. So we asked multiple account executives using a crowd-sourced insights approach. Next, we partnered with the client’s Data Science team to substantially expand the list of data variables to feed into a new churn ‘event’ prediction model. We also did an exploratory analysis to identify the most critical variables and helped source external data providers for any variables that didn’t exist on the client’s internal systems.

Result

Working with the client, we overhauled their prediction model, replacing it with a new, churn event prediction model that was:

  • More reliable
  • Accounted for multiple time horizons and customer account types
  • Delved deeper into underlying factors driving a specific event
  • Empowered customer success executives to take preemptive action
  • Up-leveled all event probabilities into a synthetic health score for each account
  • Provided a single, data-backed metric to proactively prevent customer churn

Need help slaying your churn dragon? Talk to us about building a prediction model that works.

Explore our thinking on Customer Acquisition & Retention

Show All

Explore our thinking on Customer Success

Show All

Keep me informed. I’d like to receive occasional newsletters, event notifications, and thought leadership materials.

Subscribe to Our Newsletter