CASE STUDY - Top Retail Bank More Than Doubles Campaign Profitability Using Latest Uplift Modeling Techniques
- By Mark Smith
- October 18, 2007
Commentary by Mark Smith, Executive Vice President, Portrait Software
A large successful American bank that had recently launched a new high-margin product was aiming to increase new product bookings. Should the bank mail all its customers to increase sales of the new product and incur a high mailing cost while still providing a positive return on mailings? Or could the bank increase incremental sales while reducing mailing costs by focusing its mailing on those customers best influenced by the marketing?
The value of many banking products is high, so even an increase in product sales of a fraction of a percentage can provide a positive return on investment for direct mailings. Leveraging the latest techniques in uplift modeling, this bank discovered that it could increase total sales by about 10 percent while reducing mailing volumes by 60 percent. These results more than doubled campaign profitability compared with the bank’s previous state-of-the-art approach to modeling.
Relative Campaign Profitability by Targeting volume
Figure 1. Targeting 20 percent or 30 percent of the population achieves an uplift nearly 80 percent higher than doestargeting the whole population.
The graph in Figure 1 shows the relative profitabilty of targeting different volumes with the two models, using the profit achieved by targeting the whole population as a base. As can be seen, targeting 20 percent or 30 percent of the population achieves an uplift nearly 80 percent higher than does targeting the whole population. More significantly, the profit achieved using uplift at cutoffs in the typical and desirable 20-30 percent range from 60 percent to 125 percent larger using the uplift model than with the bank’s champion model.
Maximizing Returns on Direct Marketing
What is uplift modeling and how does it differ from the traditional approach of response modeling? The core benefit of uplift modeling is the ability to optimize targeting to maximize the returns from direct marketing. Uplift modeling achieves this by predicting, at an individual level, the change in behavior likely to result from a particular marketing intervention. Uplift modeling enables us to target customers whose behavior we can change, rather than targeting those who would have behaved in a way even without the direct marketing intervention.
As in the case with the leading retail bank referenced above, uplift models were able to generate higher returns on marketing spend through lower targeting volumes. This constitutes a multiple win, because it means an increase in sales while simultaneously reducing targeting or mailing costs. Additionally, where incentives are used, the cost of these is usually significantly reduced because they are not “wasted” on customers who would purchase anyway. Customer satisfaction also tends to improve because customers who dislike being contacted are less likely to be targeted.
The Multiple Benefits of Uplift Modeling
Uplift modeling allows businesses to optimize their targeting of customers in such a way as to maximize the return on investment of marketing spend. The benefits from this in the areas of demand generation and customer retention include:
- Reducing the number of customers required to achieve a given level of business stimulation and thus reducing costs
- Increasing the level of business generation achieved for any given level of spend
- Lowering customer dissatisfaction by decreasing the level of negatively received material
- Enhancing understanding within the business of the effectiveness of various kinds of marketing spend
- Eliminating many or all of the negative effects associated with mistargeted campaigns
- Increasing customer retention
Uplift Optimizer: The Solution
Portrait Software’s uplift modeling solution has been successfully applied at five of the world’s top 10 financial services firms. Portrait’s Uplift Optimizer is the culmination of more than eight years of focused R&D effort, and is the first automated solution that predicts the incremental impact (uplift) of marketing campaigns. Singled out for its innovation, Uplift Optimizer differs from traditional modeling by identifying the two most important segments for targeting:
- The Persuadables—people who buy (or renew) who would not have done so if the campaign had not been run
- The Do Not Disturbs—people for whom the campaign triggers a negative response
Portrait’s Uplift modeling solution can be used as a stand-alone package or as an integrated component of an SAS package or other modeling applications.
This article originally appeared in the issue of .