When you want to calculate the risk of a customer leaving the company, you use algorithms - what is called Advanced Analytics or Machine Learning. These are concepts that are characterized by mystery and are almost considered the alchemy of today. Many people think that it is "very complicated" and requires "very large amounts of data". With the modern analysis tools, it does not have to be complicated and the data that you have to find in order to be able to make that kind of prediction is actually found in most companies in ordinary business systems.
Data you need to identify Customer Churn
When you select relevant data for a calculation of risk for Customer Churn, you look for data that can explain why customers have historically left the company. You look for patterns that you can recognize, for example: "Previous customers who have churned, had x, y and z characteristics before they left us".
So, if you e.g. can find three factors that occur immediately before a customer chooses to end their agreement with your company, you can start to react proactively and enter into a dialogue with the customer before they leave you.
When working with churn predictions, keep in mind that you are calculating a probability - it may be that you can identify a pattern that applies in 76% of cases, and thus you can work on retaining this group.
In other words, you look for explanatory parameters in the various data that describe the customer and the relationship during the period in which the customer has purchased from the company.
For companies that sell their products as subscriptions, we know from experience that there are explanations to be found in the following data:
-
Product
-
Product details
-
Product categories
-
The transaction pattern - the overview of all transactions, payment terms, telephone calls, e-mails and any complaints/incorrect deliveries
-
Turnover and quantity
-
Discount and promotional information
In addition, there is a number of specific data that can both be explanatory and help to segment the results of the risk calculation, so it becomes easier to find customer-relevant ways to prevent Customer Churn. Examples of this can be:
-
Demographic information such as age and gender
-
Geographical information such as regions or city/country divisions
-
Marketing data such as response to newsletters, web behavior, social media and the like
Information about which customers have left the company is typically found in customer-, CRM- and marketing systems and in some cases in financial systems. If you have data for more than two years available, the result will be the best possible, as you can take into account any seasonal fluctuations or the like.
GDPR is no obstacle to optimizing Customer Retention
The algorithms do not need names, addresses or specific CPR or customer numbers, which also helps in the GDPR context, where this personal data is classified as sensitive. Anonymous customer IDs are a great way to secure a link between the projected risk and the specific customer who is at risk of leaving the company. It gives you the opportunity to initiate an "action" in relation to the individual customer - and it is "action" that provides the value.
Hear more about our Customer Churn & Retention Framework.