Data-Driven Prevention of Customer Loss

Long-term, profitable customer relationships are one of the pillars of a successful business. Given the high cost of acquiring new customers, securing existing customer relationships is an important challenge that is not easy to deal with. Given the levels of market saturation, the battle for new customers ready to change providers is stronger than ever. Special offers for switching providers, combined with special conditions and bonuses for new customers, test the loyalty even of customers not particularly looking to switch. 

Given that 20-30% of customers in Germany switch providers each year, it is extremely important for any business to do what it can to keep attractive customers for the long term. Companies that are able to detect a trend to switch providers at an early stage will have a strategic advantage in the market. Preventing customers from switching is therefore a necessary and complex challenge for any business.

S&N offers a solution to prevent this customer turnover based on a two-phase, process-driven model. The first phase is based on early detection. These findings are then used in the second phase, focused on preventing future switches and regaining former customers. Since the approach of the second phase is strongly dependent on the reasons behind customers' switching, we focus our efforts on building the necessary conditions for a successful early detection system.

 

This is done by obtaining more information about customers. For example, changing one's bank is not usually a spontaneous decision. Instead, it is preceded by a series of steps and events that provide clues that a customer might be flagged as a risk for switching.

The customer provides this information through various channels. In addition to the information from the company's CRM system, information gleaned from credit processing, accounting, dunning and all areas that directly or indirectly hold data about customers are relevant. In addition, there are a number of external factors that can influence customer behaviour.

All of this data must be merged to allow it to be analysed. If a data warehouse is available, the pure business and partner data are generally already present in a form that can be analysed for normal reporting purposes. CRM data, especially records from customer contacts, are usually in text form and must be further processed before analysis can take place. Analysing such data by keywords is one way to prepare this information.

External data are often not directly usable, but instead require transformation of some sort. All of this information is normalised and organised chronologically.

The search for patterns of customers at risk for switching is performed on the basis of data from customers already lost. Since we are looking here for new and unknown patterns, processes such as partitioning cluster analysis that recognise structures are of use here. Factor analysis can be used before the actual search to reduce the quantity of relevant attributes. The result of this search is usually a large number of clusters that share common traits before a customer switches providers. Other methods such as discriminant analysis can then verify the specificity of the traits. The results depend on analysing the gathered data according to a set of criteria and selecting a distance function that will be used to classify the data into clusters.

The cluster analysis identifies group-specific traits that can be representative for groups of different sizes and describe a number of possible scenarios. These clusters provide information about the reasons behind customers' switching. If, for example, a group shows an accumulation of external traits, this may indicate the presence of new products and services on the market that the customers find more appealing. If there is an accumulation of service requests, this may indicate a change in customer behaviour. The results are analysed so that appropriate strategies targeted to preventing customer loss can be started.

Those patterns deemed relevant are then used as search patterns in the second phase to locate customers that might be at potential risk to switch providers. This analysis thus results in a set of customers with an increased likelihood of migrating to a competitor. Customer retention measures targeted to these particular customers would then be initiated.

But successful prevention of customer migration is not a one-time thing. The results of a prevented migration means that the reasons for customer switching might no longer have the relevance that they once had. But, other reasons will arise, which will have to be identified and re-examined. A long-term customer loss prevention programme means searching, finding and countering the causes for migration on a regular basis.

The frequency depends on many factors. The complexity and range of products are as much factors as the amount of time it takes from early detection of migration risks to demonstrating that preventive measures were successful.

Contact: Jürgen Erdmann; Turn on Javascript!