Meridian Awards 2012

Company Profile

Minneapolis-based U.S. Bancorp (USB), with $308 billion in assets, is the parent company of U.S. Bank, the fifth largest commercial bank in the United States as of Dec. 31, 2010. The company operates 3,069 banking offices and 5,310 ATMs in 25 states and provides a comprehensive line of banking, brokerage, insurance, investment, mortgage, trust and payment services products to consumers, businesses and institutions.

Introduction/Executive Summary

A top priority for U.S. Bank is producing organic household growth and to accomplish that, we wanted to design a prospecting campaign featuring our core products, i.e., checking accounts.  Given the inherent complexities of prospecting, we needed a provider who could offer a comprehensive suite of services including list generation, model development and campaign management services.  We found a solution with Pitney Bowes Software with its expertise in GIS, data, and now specialized modeling.  Having all these capabilities in one place has greatly enhanced our efficiencies for our prospecting efforts.   

Business Challenge

In the past, executing a prospecting campaign meant coordinating multiple vendors for distinct responsibilities such as strategy, list targeting, data appends, model building and analysis etc.

For example, we used geographic based targeting to determine the distribution of our mail.  For simplicity reasons, all participating branches had their mail distributed in similar trading area radii, regardless of population density, demographics, or singularities like proximity to rivers, airports, freeways etc.

We wanted to improve the geographic targeting of our mailings, but to preserve cost efficiencies, we selected addresses based on carrier routes residing within a pre-determined radius.  This tactic required that we overlay each radius with a USPS carrier route map.

After selecting the universe of carrier routes, we needed to attach all pertinent data corresponding to the selected carrier routes.  This process meant we had to work with multiple data vendors which took significant time, effort and money.

To further focus our targeting, U.S. Bank utilizes specialized predictive modeling.  Because we are most interested in how our mailings perform relative to a “hold-out” control group, we need to utilize unique modeling techniques that help predict lift over control.  Also, other modeling considerations had to be taken into account because of the relatively small sample sizes that the carrier-route approach produced.

For back-end results analysis, we typically focus on what the overall campaign produced and how the model performed.  Even though the mailing is primarily driven by geography-based selections via mailed carrier routes, we never looked at the results geographically.

Solution Description

Now that we have partnered with divisions inside of PB Software, many of the processes above have been streamlined resulting in better efficiencies and improvement of the overall campaign.

GIS-based targeting:

First, with Pitney Bowes Software’s MapInfo Pro software we’re able to better target our geography selections around our branches.  Our GIS team is able to dynamically specify radii around individual branches based on our existing customer penetration and other dimensions. 

These new individual radii can then be used to select the carrier routes that fall within them.  Because U.S. Bank licenses the carrier route component/data in MapInfo, we can do this selection internally or for efficiency reasons, Pitney Bowes Software is able to perform the same calculation in their shop.

Data Appends:

The selected carrier routes can now be appended with Pitney Bowes Software’s extensive suite of data elements.  Not only can we obtain the standard demographic fields from U.S. census data, we can also tap into the vast reservoir of Demand Insight variables for modeling.  Once again, because we have this data in-house, we could perform this task internally or have Pitney Bowes Software do the same on their end - which would be seamless to us.

Portrait Uplift Predictive Modeling:

Over the years it has been very difficult to develop stable models for this prospecting objective due to the limited data samples previously mentioned.  We have partnered with Portrait Software (now part of Pitney Bowes Software) to assist in this task.  Initially we attempted to do this modeling in-house, but with the expertise of the Portrait staff and the seamless connection with Pitney Bowes Software’s data sources, it made more sense to leverage Portrait’s resources.

GIS-based results:

Lastly, now that we have the geo-referenced carrier route files in-house, we will be able to show our campaign results by carrier route around each branch.  This, in turn, will lead to greater campaign insights while continuing to optimize future mailings.

Results Achieved

There are many benefits we have realized from our partnership with PB Software.

Efficiencies and cost savings with Pitney Bowes Software data append.

• A reduction in the number of vendors needed for data appends, from three to essentially one , shortening our production window and reducing expenses.
• Data transfer logistics have been greatly simplified, again positively improving our implementation ability and enhancing speed to market.

Modeling

The carrier route targeting approach, while exceptionally cost efficient, has made the modeling environment extremely challenging.  Nonetheless, we have made quantifiable progress in each successive generation.  With every mailing we test new theories and new data elements and are making strides towards a strong predictive model.  It’s a great advantage to have the experts in the data elements at the same company as the modelers using that data.  Partnering with the Portrait modelers have freed up our internal modeling analysts to focus on other business demands.

GIS Selection and Results mapping:

By using MapInfo and location analysis around our branches, we want to enhance the targeting of our next campaign by selecting radii more relevant to producing incremental results.

We’ve only recently begun to map the results geographically, but we are looking forward to analyzing the results in the hopes of data-mining useful attributes predictive of superior performance. We have already uncovered some interesting findings from analysis done at the state level and how the model is focusing the distribution of the mail.

Overall, we have achieved a number of successes by implementing many aspects of Pitney Bowes Software’s expertise and guidance.  We see further opportunities for enhancements by utilizing more of Pitney Bowes Software’s services for future campaigns and look forward to increasingly profitable campaign results.