Ken Mison is a vice-president of Market Facts of Canada, a Toronto-based market research firm. There, he leads The Commitments Management Group, which deals with strategic and tactical issues related to customer and employee satisfaction, attraction and retention.
If you knew a customer was vulnerable to a competitor, would you be able to think of something to say to them that would keep them buying from you?
If you were able to identify your committed customers, would you save any money by directing your marketing efforts only to those who are not committed?
If you knew two customers who were vulnerable to advances from the competition – one who was dissatisfied with you and the other who was drawn to a competitor – would you address them differently?
If you knew who among your non-customers was actually available to you and why, would you be able to tailor your marketing to speak directly to the individual rather than the mass market?
If you answered ‘no’ to any of these questions, then you need to put a little commitment into your database.
In my first article, (‘Commitment a more reliable measure’, Strategy DirectResponse, Nov. 8, 1999), I argued that ‘commitment’ and ‘loyalty’ are not the same thing. I defined commitment as ‘what a customer feels’ and loyalty as ‘what a customer does.’ I also introduced the Conversion Model, a proprietary technique for segmenting customers and non-customers on the basis of commitment rather than stated intentions (the basis of most loyalty measures) or observed behaviour (such as churn models).
I argued that there are only four questions that need to be asked to establish the commitment of a person to anything. They are:
To what extent does Is the How does that person rate other If they rate other In this article, I’d like to discuss the practical application of the Conversion Model. There are six steps, which I will discuss in turn. 1) Survey a sample of customers/non-customers to segment for level of commitment Many organizations attach Conversion Model questions to existing surveys, such as customer satisfaction surveys and advertising tracking programs. If you do not have an existing survey, then a one-time commitment survey can be created. Any survey methodology will do – telephone, mail, in person, e-mail or Internet. The sample size required is based on normal sampling techniques to account for variables such as customer segments and geographic distribution. Large samples are required, however, and one should expect to conduct more than 1,000 interviews in a typical mass market. For small markets and business-to-business applications, smaller samples can be used. 2) Fuse the survey data with relevant profiling information (revenues, products held, types of complaints, service history, etc.) The second step is to combine the commitment scores provided by the Conversion Model with information that you currently hold on your customers/non-customers. This will include geo-demographic data, product-use data, complaint history, internal classifications for profitability, revenue or value – in short, any information that might serve to explain a customer’s commitment. If you do not have a customer or non-customer database that contains these types of data, the survey can collect them and you can use this to identify what variables you need to start to track in the future. 3) Identify customer profiles relative to the commitment segments by using modelling software While there are numerous data mining software programs, we must select one that provides a predictive model. Our preference is to use CART (Classification And Regression Trees) but CHAID will also work. The idea is to use the information from your database to predict commitment levels. CART will then provide us with a set of rules for predicting levels of commitment (Entrenched, Average, Shallow and Convertible). 4) Segment the entire customer/non-customer database for commitment according to these profiles Using the classification rules developed through the CART analysis, we then score each individual customer/non-customer in your database. Each person is classified as Entrenched, Average, Shallow or Convertible. Non-customers would be classified as either Available, Weakly Available, Unavailable or Strongly Unavailable. 5) Validate the accuracy of the prediction through a validation survey Validation can actually be conducted several ways. We could just wait to see what happens – do the Convertible customers leave and do the Available non-customers become customers? We can also conduct a survey with a sample of the classified customers/non-customers in order to have them answer the Conversion Model questions and then compare their answers to the predicted scores from the model. Once again, this can be done through an existing survey or through a custom survey. We have typically achieved 60%-70% accuracy on the first prediction. This analysis usually suggests additional information that can be collected and included in future models. It is also typical that we will obtain an accuracy of 80% or greater after one to two years of refinement to the model. 6) Monitor the effects of new marketing and service initiatives. Once you have the data, numerous initiatives can be developed to respond to different types of customer/non-customer commitment. The impact of these initiatives should be monitored closely in the first few years. This can be done using any source of information, such as salesforce debriefings, customer service encounters and surveys. Finally, you should update your commitment scores on a regular basis. How often you do this will depend on how often you update your customer database variables. For example, if product usage data is updated every two months, you may wish to update your commitment scores at the same time. Some organizations re-run their model on a monthly cycle. This gives them a chance to flag changes in commitment and, if the customer is moving from secure to vulnerable, the organization can take a proactive intervention to find out what has caused the customer to become vulnerable. Other organizations update on a monthly basis in order to account for past marketing and service initiatives that were designed to build customer commitment or non-customer attraction. In my next and final article, I will provide a case study of how one organization – Lloyd’s TSB, one of the largest banks in the world – has used the Conversion Model and database mining to increase customer commitment, market share, revenue, new product uptake and profitability. I will also look at some guidelines for managing customer and non-customer relationships based on commitment. Ken Mison can be reached at (416) 964-6262 or by e-mail at kmison@marketfacts.com. Additional information about the Conversion Model can be found at www.theconversionmodel.com.
