-
-
Marketplace
-
Channel Resources
Articles from this Site
-
One brand, one Web site! DM Review is now the home of all the content you're used to at BIReview.com and much more. If you are registered at BIReview.com, you're already registered at DM Review. If not, take just a moment to sign up for all the free services we have for you at the new DMReview.com.
From Acorn to Mighty Oak
My introduction to business intelligence began early in my banking career. It was the early 1990s and I was working in San Francisco for the chief financial officer of Wells Fargo, and responsible for a "payback analysis" on new capital investments. A funding request for offsite ATM locations landed on my desk for analysis. The problem seemed straightforward until I began to plug in the potential revenue benefits and I hesitated - What is the value of an ATM location?
Commercial banking in the U.S. had only recently been deregulated and bank branches were very large and impressive building with marble pillars and a huge steel vault. ATM machines had been developed a decade earlier to deliver cash automatically and thereby save on the labor expense of tellers. They were largely deployed at existing branch locations although a few had made their way to shopping malls and office complexes. What is the value of an ATM location? The consensus view was that the value of an ATM machine was the usage fees collected from non-account holders and the handy convenience they provided to existing customers. They were a nice offering for customers when there were just a handful of offsite locations, but "nice" was not sufficiently rigorous to suit the CFO in my investment payback analysis!
Meanwhile, the traditional view of the value of a bank branch was equal to the deposit balances "domiciled" there when a customer opened their first account. Any additional business acquired from the customer accrued to the domicile branch of first account origination. Downtown main street branches had a huge base of domicile deposits since that's where folks opened their first account when they moved to town. The domicile deposit balances of suburban branch locations weren't nearly as large but, gee, they sure had a lot of teller activity on Friday afternoons!
In retail banking, customers value convenience and the ability to access funds from their accounts "all over town," but building marble monstrosities on every street-corner becomes prohibitively expensive after a while. The challenge is how to manage the trade-off - deliver more locations, which customers like, without blowing through the capital budget, which the CFO does not like.
The Value of Modeling Customers
The answer to both the question of "how to value an ATM?" and "how many branches do we need?" is to look at how customers currently behave rather than what they did in the past. Mapping deposit accounts to the branch location where a customer opened their first account used accounting systems as a preservative to fossilize long-ago behavior as current business intelligence. The customer behavior that we began to study was how customers make deposits and withdrawals at branch locations and, while we're at it, let's throw in ATM activity as well. The next challenge to overcome was that the detailed transaction information by location for teller and ATM were housed in separate systems managed by disparate parts of the bank. While I had the support of the CFO and the head of Retail Banking to undertake the customer behavior study, their support alone was not sufficient to cause the teller or ATM folks to deliver the data from their mainframe applications. I then embarked on an internal campaign to "sell" the benefits of the analysis to anyone who would listen, notably the managers of the separate ATM and teller transaction systems. I laid out the argument in many presentations and picked up the check for innumerable lunches. Eventually, I was able to convince the key stakeholders that they would each benefit from a global view of customer transaction activity, and a surreptitious exchange of files was made one evening.
Once we merged the ATM and teller transaction data, we found a great deal of interesting customer behavior. Most customers didn't perform the bulk of their transactions at the branch where they opened their first account - they had moved their transactions to a branch closer to their home or workplace. We cooked up a formula to rank customers' branch usage that we called "primary patronage." Beyond the breadth of branch usage, what we really wanted to know was the depth of usage - which branch was the linchpin in the customer's relationship with the bank?
We were proud of ourselves and proud of our Primary Patronage algorithm, but how did we know that we got it right? How could we be sure that it reliably measured actual customer behaviors and preferences? Fortunately for our research, pressure on the bank's earnings provided a unique opportunity to validate the customer behavioral segmentation. In order to trim expenses, a half-dozen underperforming branches were identified for closure. Quickly we moved to exhaustively track the transaction behavior of customers in the marketplaces that would be impacted by the branch closures.
After the branch closures took effect, we identified the customers who had attrited from the bank as well as the transaction activity pattern of the vast majority who were retained. Much to our gratification we found that the best single predictor of customer attrition in the impacted markets was when the customer's primary patronage branch was one of those that were closed. The patronage formula was robust - highly reliable and predictive of future customer behaviors.
Patterns of Customer Behavior
If we know how to identify through activity a branch that is so convenient for a customer that they will leave the bank if the branch is closed, then we can dimension the value customers place on the branches (and ATMs) that they use.
This insight led to an intense analysis of customers' transaction patterns and we soon discovered that customers used the branches and ATMs as a web - a network of banking services. A little digging in my economics textbooks surfaced the name of this phenomenon, known as a "positive network externality." Every additional node (branch or ATM) in the delivery network increased the value of the existing network for current and prospective customers. The key was to configure the locations of the branches and ATMs to maximize convenience in a marketplace. Quickly we developed management information to measure the value of a branch to its primary patrons and how they redistributed themselves when we closed or opened new locations. The patronage data was actionable as we could accurately predict the customers most likely to be impacted by changes in the branch configuration and contact them proactively.
Having the patronage information available in a database was a big step but in order to leverage the intelligence operationally we needed the line managers in the Retail Bank to believe in its validity.
I spent a lot of time visiting regional managers and brought with me branch level and individual customer patronage reports. The management team would review the reports with interest, albeit rather skeptically at first. But inevitably, a branch manager would exclaim "this is exactly what we knew was happening but didn't know how to measure!" Further dialogue with those stakeholders most impacted by the information (line managers) led to a few tweaks and twists to the algorithm to accommodate their suggestions and real-world experience.
Gradually, patronage moved from being a radical idea from headquarters to a core concept that was embedded in the retail banks' operations through branch scorecard and compensation planning. Today the concept is so prevalent in the bank that few realize that it was ever "invented."
This insight of measuring convenience via customer behavior gradually seeped into the strategy of Retail Banking. Now that we could understand which locations customers valued based on their activity, we could explore alternative strategies to locate branches. We had found that branches located in neighborhood shopping centers with major supermarket stores as anchors did extremely well at attracting new patrons. We quickly discovered that these were on the regular traffic path of customers who visited their local supermarket one or two times per week. What we if we located offsite ATMs within supermarkets in a massive deployment? We courted supermarket chains in the Western U.S. to explore their interest in the bank locating ATM locations in their stores. The supermarkets liked the idea of ATMs well enough but pushed for "mini-branches" located inside their stores. The desire to "lock up" major grocery chains led to the decision to rapidly roll out branches, of varying formats, in hundreds of grocery stores in several chains. In little over a year, hundreds of "in-store" branches were planned, constructed and opened throughout California.
A critical threshold in retail banking strategy was reached and crossed. The ability to measure and monitor customer patronage provided the intelligence to plan, deploy and adopt a radical innovation in retail banking service delivery. The benefits of rethinking retail bank services delivery continues to this day. Wells Fargo now has over 650 "in-store" branches in the US and more than 6,500 ATMs, many at offsite locations.
Conclusions
This methodology and the ability to acquire and interpret customer behavior in a rapid manner were subsequently applied in the telephone and Internet banking channels.
Key Lessons Learned:
- Customer Behavior can be uncovered through detailed data analysis
- Discovering a critical insight allows opportunity to re-think business strategy
- Leveraging the insight is the big challenge
- Exploiting the insight requires organizational discipline and a culture that challenges the status quo.
Other banks had similar ideas for leveraging customer transaction behavior to improve retail banking performance, but either lacked the ability to rapidly scale the data processing required, (remember this was the early 1990s!) or weren't willing to test new ideas through experimentation, feedback and adaptation. It should be noted, however, that not every experiment tried was a financial success - the drug store mini-branches never did roll out.
The critical information required for this innovation in retail banking strategy was there all along - buried in legacy accounting systems and literally printed out on customer statements every month. The little acorn of "how do customers actually behave?" became the mighty oak of network delivery redesign. Look around on your forest floor and you may find not only acorns, but seedlings already taken root - try not to step on them!
Gary W. Class is Senior Vice President, Internet Services, Wells Fargo.
Gary W. Class is Senior Vice President, Internet Services, Wells Fargo.
For more information on related topics, visit the following channels:
