How to Prevent Customer Churn

Shift from Customer Churn Prediction to Prevention

Highlights

  • The best way to manage customer attrition is to prevent it from happening in the first place.
  • By itself, the out of the box (OOB) CRM software customer attrition forecast is a static report that is too overly generic to be accurate or actionable.
  • However, extending the OOB CRM customer attrition report into an interactive predictive model that shows not just who, but why those customers are churning, with measurability, surfaces the root causes that can be resolved before they create customer attrition.
  • It's a shift in reporting from delivering bad news too late, to delivering an operating model to prevent bad news.
Johnny Grow Revenue Growth Consulting

Why Most Customer Churn Prediction Doesn't Work

CRM systems such as Salesforce Service Cloud and Microsoft Dynamics 365 use artificial intelligence (AI) and machine learning (ML) to create customer churn prediction reports. It's an outstanding idea as reducing customer attrition delivers an immediate, significant and sustained financial impact.

CRM Customer Churn Prediction Report

However, having implemented these reports for over a dozen clients we learned early that the CRM software out of the box (OOB) attrition report is a generic report built on only the most common data points used by the bulk of CRM software users. The default report is a good starting point, but insufficient for several reasons.

  • Because it's built on few, overly generic data elements it misses any impairments unique to your business. Company specific impediments tend to be the greatest causes of customer loss.
  • It doesn't apply essential variables such as Customer Lifetime Value (CLV), Customer Satisfaction (CSAT), Net Promoter Score (NPS), Account Engagement Score (AES), customer sentiment, account health score, RFM (Recency Frequency Monetary) score or most customer insights. These types of metrics are essential to predict which customers are most likely to defect.
  • It doesn't identify attrition by customer segment. It's critical to predict turnover by segment so the most valuable customers at risk can be prioritized.
  • It doesn't illustrate why customers leave.
  • It doesn't advise how to prevent or reduce customer attrition.
  • It doesn't show the financial upside impact of improving customer retention or any individual cause of attrition.

There's one additional problem.

The OOB CRM report attempts to show which customers are going to separate with the company. But the reasons for seperation have accumulated over extended periods, so what are you going to do? Are you going to call them and ask them to stay? You're unlikely to change frustrated customer feelings that have built and festered for quarters or years with a short communication, that is too little too late. Knowing which customers may defect and being able to retain them are two different things.

By itself, the OOB CRM attrition forecast is a static report that is too generic and imprecise to be accurate and without the insights to be actionable.

However, the report is extensible.

The CRM report predicts which customers are at risk. Johnny Grow takes that data, enriches it with the causes of customer churn, the methods to reduce turnover, and the KPIs that most impact revenue to create an interactive model that shows how to prevent customer churn.

Pro Forma Customer Churn Interactive Model

Knowing why customers turn over is more powerful than knowing which customers will exit as it allows companies to prevent customer attrition before it happens.

See the customer attrition model which uses CRM software to shift from customer churn prediction to customer churn prevention.

Click to Tweet

One More Challenge to Resolve

Better prediction is less about better AI algorithms and more about better data. Accurate predictions are dependent on the right data.

Customer attrition models don't fail because of the data you apply. They fail because of the data you didn't include. If you are not tracking relevant data, the AI or ML program is irrelevant, and back to using only the most obvious and generic data that misses the bulk of customers about to defect. A common adage in AI circles is that 10% better data will outperform 100% better algorithms.

To know the causes of attrition the Johnny Grow customer retention predictive model applies analysis of lost customers to identify the most influential factors for those losses. However, research shows that customers frequently defect based on things you are not doing. And it's difficult to measure what you're not doing. For example, you cannot measure how many customers left because of poor brand engagement if the brand has no engagement or doesn't track engagement response.

One way to resolve this underlying problem is to use industry benchmark data. We use our proprietary research to identify the top industry factors that cause customers to churn and to stay, to fill in missing client data.

For example, a common factor is poor onboarding. Sometimes customers find it difficult to setup or figure out how to use your product, get frustrated and eventually exit. If you are not doing onboarding or measuring onboarding, there would be no data to identify this cause of attrition.

So, we apply industry research. In certain industries, we know from the research that poor onboarding contributes to 13 percent of customer attrition overall, and 33 percent of that attrition occurs in the first 100 days.

So, we can measure the number of customer losses in the first 100 days and correlate that to product utilization and support issues. If we see low product utilization or initial support cases followed by a fall off, then followed by departure, we have data to suggest creating a test hypothesis for a new or improved onboarding process.

Another common industry attrition factor is poor customer support. But if you don't measure customer satisfaction (CSAT) and correlate to attrition, no AI model has the data to deliver accurate or meaningful predictions.

So, we again use our industry research to show the factors that most contribute to poor CSAT and customer churn and see if these factors exist at a company. For example, for certain industries, the research shows us that high AHT (average handling time) and transfers, and low SoA (speed of answer) and FCR (first contact resolution) directly drive low CSAT and high turnover. If we test and uncover these factors at a company, we can extrapolate the data to predict attrition.

The Johnny Grow research data tells us two things – the factors that most contribute to industry-specific customer turnover, and the multiple types of data that correlate to those factors. So, when a company doesn't track the primary types of data, we revert to other data that can be aggregated or substituted for what's missing.

The Point is This

Accurately predicting which customers will depart, and proactively mitigating causes of attrition are no easy tasks. But succeeding where competitors fail provides a strong competitive advantage. It also provides a significant and sustained source of increased revenues and profits. For many companies, increasing customer retention is the new business growth model.