Intelligent Enterprise

Better Insight for Business Decisions

RSS
Webcasts
Whitepapers
Subscribe
Home

July 23, 2001

A Savvy Decision

Money and technology continue to shape the firm of the future

By Barry Grushkin

Investors are getting savvy. They are no longer enticed by measures of things past, like counts of eyes, hits, or even earnings. They want forward-looking customer plans - rigorously backed methods for improving profit stream for each of their customer segments. Promotion spending for unprofitable and low-allegiance customers will no longer do.

Investors have heard too many examples like this true story: An online brokerage goes public. Its stock soars because of exponential customer growth. But to rein in costs, it must stop its deep discounting. The customers scamper away, and the stock crashes.

To explore these issues, I interviewed a range of experts from the technologically advanced CRM firm, Xchange Inc., headquartered in Boston.

Steve Gallant of Xchange's Advanced Analytic Services argues that just as airlines quickly realized that without sophisticated, multitiered seat-pricing models they couldn't remain profitable, sophisticated customer modeling will become a basic business requirement for many other industries.

Management will have to make return on investment (ROI)-based plans appropriate to each meaningfully defined customer segment. This is breaking down traditional distinctions among marketing, sales, CRM, and analysis. Every customer interaction is seen as an opportunity to gain or offer information, improve the relationship, or increase profitability.

Knowing what actions to take - when and to whom - becomes part of a larger, integrated point-of-view.

Companies must use historical, realtime, multisource, and multichannel information to develop and implement strategies, learn more effective actions, and optimize customer value on an ongoing basis.

AN INTEGRATED POINT OF VIEW

Six key issues drive this integrated point-of-view called customer value management: customer profitability metrics, modeling and analytics, process, 360-degree view of the customer, learning and tracking what you learned, and a 360-degree response to the customer.

Customer profitability metrics. Two important metrics are value at risk (VAR) and value in play (VIP). VAR is the average probability-weighted loss that you can expect from customers as they begin to attrite, for example. Forecast models tell when attrition is about to happen and how various interventions can improve things. For example, a bank that sees an average, but regularly high, balance consistently dropping, may want to send a promotional gift.

VIP is the probability-weighted profit potential that companies can either snag or miss in a customer interaction. An example would be the highest expected return for the best cross-sell offer. A high VIP usually warrants taking action.

Modeling and analytics. Segmenting customers into useful groupings and scoring their potential interest in different products are two key demands placed on analytic methods. The better you can do these the more targeted your actions.

Segmentation, once seat-of-the-pants groupings, is now often performed by analysts using online analytic processing (OLAP) tools to slice and dice customers into sets that have promising characteristics. An example might be a lifestyle or regional grouping that seasonally buys high-margin items. Companies can then define appropriate actions. Additionally, automated methods for producing homogeneous clusters are currently in use, revealing wholly new, useful, and valuable ways of looking at customers.

As for scoring methods, Gregory Piatetsky-Shapiro, also in Advanced Analytic Services, is implementing an automated one he calls "an intelligent Ph.D. in a box." The goal is a robust technique integrated with the larger analytic infrastructure so that a new model can be generated as data changes. It is based on an algorithm developed by Charles Elkan at University of California-San Diego that is faster, more stable, more readable, and seems to work better on many customer data sets than many competing technologies.

Process. But wait, say those with street experience, "if you automate a bad process, you just get a bad process. A fool with a tool is still a fool." From this perspective, the questions include: Is this marketing campaign in tune with what sales is doing? Is it efficient? Are the right people engaged? Are they pulling in the same direction, undermining each other, or redundant? Do we know exactly what has gone out the door and when? Do we have a uniform voice or are we annoying customers with contradictions? Is the right information being used at the right time?

360-degree view of the customer. Integrating your back office (accounting) and front office (customer interactions records) is a start, but every customer interaction is an opportunity to collect information. You should ideally categorize and save interactions on all channels for future use. Furthermore, everything need not be in a single data mart; instead, you can poll information from sources as needed. This saves a lot of politics and money.

Learning and tracking what you learned. All activities, including marketing and sales, have a dual function - current needs and good experiments to learn from. As with experiments in a lab, a precise set of well-documented steps is necessary: what you did and why, the reasons, the method, the results, comparative cases, controls, and control cases. The goal is to maximize the use of the feedback you receive.