Smart Customer Interaction Reduces the Demand for Data and Improves Response Time


Share on LinkedIn

In Silver Blaze, a murder story featuring that incomparable master of deductive reasoning, Sherlock Holmes, the key to solving the mystery lies in the “curious incident of the dog in the night time.” The clue? (Warning, spoiler alert if you haven’t read the book) The dog fails to bark when circumstances suggest it should be woofing its head off. The logic employed by Arthur Conan Doyle’s legendary detective in unmasking the perpetrator highlights a fundamental tenet of both successful sleuthing and customer-relationship building: The best interactions are often defined by what is not done or by questions that are not asked.

A traditional customer interaction might be described as “bottom up.” It requires first establishing a 360-degree customer view and then applying a set of rules to reach a conclusion. This tends to be both time- and resource-consuming and often necessitates that the customer answer a series of questions of dubious relevance to the specific purpose for the call or visit to the web site. In a “top down” approach, on the other hand, you minimize data requirements by first asking what needs to be known for the optimal completion of a transaction. This results in a faster, more efficient and generally more economical interaction with the customer.

Before her application can be approved, the company will almost certainly place an expensive call to one or more credit bureau databases.

Take the example of an existing female customer applying for a mortgage with a major international financial services company. Before her application can be approved, the company will almost certainly place an expensive call to one or more credit bureau databases. In a bottom-up scenario, the request for a credit report would take place at a very early stage in the interaction, as the company tries to define the creditworthiness of the borrower. A top-down approach recognizes the need for credit, as well as the significant cost involved in making the query. Rather than going straight to the credit bureau, in a top-down approach, you attempt to infer the borrower’s credit-worthiness based on what you know or can discover using lower cost sources of data. As it becomes clear that you will not reject the application for other reasons, you then move to request the credit report.

Rules definition

The existing rules for this particular financial services company have been defined with a goal of minimizing the costs of data acquisition for mortgage origination. This is accomplished through a series of steps. First, the company checks to see whether the applicant is flagged as a credit risk in its internal systems, a query that can be made quickly and at little cost. If yes, the transaction is ended. If not, the company retrieves elementary profile data from the back office (age and residence, for example). At the same time, decision logic dictates asking for the applicant’s income.

With this, the decision engine then determines whether or not the number is reasonable, given her age and place of residence. If not, the company will dismiss the application or, perhaps, suggest a smaller loan. In this case, our borrower continues to check out. Next, the financial services company retrieves transactional data (which is “expensive,” as it is against operational data sources) and product history from the back office (if a person is holding multiple products, prompt payments may get a higher priority). The company again applies the relevant predictive models and policy rules. When this analysis does not raise any red flags, the call is made to the credit bureau. The credit scores are retrieved and used as input for the final decision on the approval and the amount of the loan. This entire process can be completed in a matter of moments.

The contrast with a bottom-up approach is, of course, most significant in the case of those customers whose applications are not approved. The 360-degree review requires ordering credit up front, an investment that is essentially wasted for those customers whose applications for this mortgage product are rejected. Even assuming a nominal cost per report, this represents a major potential savings on this product alone using the top-down methodology.

A top-down approach seeks to collect the minimum amount of data needed to reliably conclude a transaction while meeting or exceeding the needs of the customer. As you can see in the example I described, the key to achieving this is having in place a very robust set of rules and predictive models—and understanding when and where in the process it becomes necessary to acquire new data (or, more opportunistically, to offer an up-sale based on the data acquired).

You can set different levels of certainty based on the nature of the product or service you offer. For a mortgage, for instance, you would mandate a high level of confidence. For a shorter-term, less expensive product, you might tolerate more variability if it allows you to save on upfront data-acquisition costs without significantly increasing risk.

On the front end, customers see a more intelligent, streamlined process, whether they are speaking with a sales representative or interacting with a web page, that results in a better customer experience. In the case of our mortgage borrower, she gets what she wants: a fast answer to her query and, in this case, approval for the requested loan amount.

The lender gets to book a loan and to deepen an existing client relationship. At the same time, it can employ next-best-action marketing techniques to suggest a further course of action to the customer: a bigger loan or insurance protection, for example. The system’s predictive analytics, enhanced by real-time data collected during the course of the conversation, significantly improve the chances for a further sale.

Rob Walker
Rob Walker, Ph.D., vice president, Marketing (EMEA) and Decisioning Solutions, is responsible for managing the strategic direction and development of Chordiant Software's predictive decisioning technologies. He was previously with KiQ Ltd. and Capgemini. Walker holds a master's degree and a doctorate in computer science from Free University in the Netherlands.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here