AI saves customers

SAP wants to become the leading provider of machine learning in the corporate customer business, according to company CEO McDermott. The company is therefore investing heavily in AI and has created a foundation for machine learning applications and artificial intelligence software with its Leonardo innovation system.

Both on the supply side and on the demand side, companies are increasingly investing in AI and its subfield, machine learning. 73 percent of companies expect the use of AI technologies to increase customer satisfaction and 65 percent to reduce customer churn, according to consulting firm Capgemini.

ML solutions are predestined in this environment to determine churn forecasts, also called churn prediction or customer retention.

Customer retention can be used to identify customers at an early stage who are about to leave for the competition, for example when their contract with a service provider expires. Customer retention is particularly relevant for companies with a large number of customers, because it is difficult to assess the loyalty of each individual customer.

Here, too, SAP has an ML-based solution up its sleeve. According to the solution description, SAP Customer Retention can be used to derive and predict customer behavior based on transaction data and digital interaction points.

However, in addition to possible churn parameters, the algorithm of a customer retention system requires, above all, as large a data set as possible about the respective customer history. This includes all customer information, i.e., master data and transaction data.

These include address data, purchasing behavior, purchase history, preferences, and the traces the customer leaves behind on the Internet and social media. But all of this data must first be made available to the ML system, processed and validated.

And this is precisely the challenge for companies. After all, customer master data and transaction data are distributed across several systems in companies - be it systems such as SAP CRM, SAP Service Ticketing Intelligence, SAP ERP solution or call center applications.

This makes it virtually impossible to merge the customer data managed in the different systems and make it available to the ML system. Therefore, companies need a solution and process methodology that brings together all customer data - address data, purchase behavior, purchase history, preferences, and the traces the customer leaves behind on the Internet and social media - from all their individual company systems, without siloing and redundancies.

Garbage in, Garbage out

In addition, companies must keep in mind that the use of ML only brings real benefits if the database that companies provide to the system is also of high quality.

This is because the basis of every machine learning system is data sets, which are used to train ML systems. To prevent the system from learning incorrectly and making erroneous predictions, it is therefore critical that the underlying database is absolutely error-free.

In the case of customer retention applications, customer data must therefore be up-to-date, correct and comprehensive - and errors must be weeded out in advance. Potential sources of error are incorrect spellings, duplicates, outdated data, and semantic problems.

ML only achieves the best success rate if the database it is provided with for learning is also of high quality. The more correct a database is, the better an algorithm will draw its conclusions from it.

For this reason, data must be maintained, protected and monitored throughout its lifecycle. Only in this way can AI and ML systems develop their full potential.