Calculating the incremental value of promotions to decide which customers to contact has been a dream of direct and database marketers for years. But despite promising research and testing, reliable results have been elusive -- until now.
The Search for Contact Elasticity
Communications often target customers with a high propensity to buy; for example, RFM and response models heavily reward recency of purchase. As a result, most marketing is conducted on the most active segments, leading to customer fatigue and discounted offers being delivered to customers that would have purchased on their own, and perhaps at regular prices. It also underestimates the potential value of those customers that buy less frequently. Incremental value was conceived as a way to target those customers that really need the stimulus of a promotional message to trigger a purchase. It’s an appealing concept because it is designed to assess a customer’s “contact elasticity,” to support marketing decisions.
But in addition to mixed results from various studies, many companies did not have good enough test and control data to identify the true incremental value potential across the customer base. Modeling practitioners realized the difficulty of estimating incremental values, especially to individual customers, who are either in the mail group or the control group of any campaign, so that an individual’s incremental value in this campaign is never directly observed. Moreover, methodologies were too sensitive to the particular promotion: changes in seasonality, format, channel, and offers often eroded the lift of incremental value. While there were successes, the great benefits promised by the approach remained elusive for many.
Spending -- or Not -- in Increments
One mistake was to seek to replace campaign-level tools, such as response models. Incremental value is not a one-time effect, but rather one that can only be observed over time. If a customer was mailed a promotion four times and responded only to the last mailing, we need to consider the effort we put into all of the mailings to judge if we have a gain.
An effective incremental value approach gives us a tool that estimates incrementality for more than just the current promotion it guides us on whether to promote a customer at any given time. We need to be able to evaluate whether spending money today will yield an incremental result, or if we have reached a point of diminishing returns with a customer.
General advertising has long used a similar approach. The “pulsing” strategy is really a decision built around incremental value. Television advertisements are introduced and their performance is monitored, and when returns diminish, the campaign is paused. It is resumed when it is likely to again produce an incremental result.
Going Beyond RFM
In order to get similar results for database marketing, we need to consider information from three areas: our customers, our marketing, and the business environment.
Our customer information should include transactional history, past promotions and responses, demographics, and many other items. We also need good information on our marketing to customers to understand the incremental effect of our planned promotion. Environmental factors, such as seasonality or interest rates may also play a role, e.g., the amount of consumer holiday buying vs. other times of year changes what is incremental.
We also need a “dynamic” approach, to incorporate the passage of time, changing customer behavior, new data and information collected. To again borrow an idea from media planning, the marginal effectiveness of advertising should be a function of the time path of advertising and the whole marketing mix. Similarly, historical promotions and a customer’s responses alter his incremental value dynamically over time.
This approach does not necessarily reduce marketing expenditures: the same budget will likely support communications to more customers instead of concentrating on a relatively small proportion. When we can derive incremental value, we should invest in marketing to a customer. We obtain balance by permitting a “pausing” strategy whenever the return begins to diminish.
Like many good ideas, incremental value has taken a few iterations to really get off the ground. Early results were mixed: some clear wins, but many cases where the potential was not quite realized. With a new perspective on how to use this approach and new analytical techniques such as latent class, mix models and Bayesian hierarchical models, a good idea has finally matured into an important and practical marketing tool.
About the author
David King (email@example.com) is chief executive officer of Fulcrum (www.fulcrm.com), a leader in advanced analytics, technology, and program solutions for marketing.
Please note that the Viewpoints listed in CRM magazine and appearing on destinationCRM.com represent the perspective of the authors, and not necessarily those of the magazine or its editors. If you would like to submit a Viewpoint for consideration on a topic related to customer relationship management, please email viewpoints@destinationCRM.com.