BECU opens branches, cross sells by rolling the analytical dice.

BECU is using the power of software to "compete on analytics" as it makes decisions on such things as where to site branches and how much to let members take out of ATMs.

Competing on analytics is the brainchild of process management consultant Tom Davenport. The idea is that organizations, in this case credit unions and banks vying for business in the Seattle market, that offer pretty much the same commoditized products or services can "take advantage of the little conditional probabilities that lie buried in data," according to Cal Bierley, market research analyst at $8.6 billion BECU.

Bierley's case in point is BECU's use of Pitney Bowes Business Insight's AnySite and Data Flow solutions. The former is used to study potential and existing member density, competitor branch density, workforce population, consumer loans and deposits and other factors to decide where to open, close and maintain branches. The second is used to target members for select marketing messages such as preapproved loan offers and identify those who are perhaps one product short of qualifying for a bundled product.

Together they help Bierley and his colleagues "become like analytical gamblers. You don't win every time but by taking advantage of the little conditional probabilities that lie buried in data, and doing it consistently, you can win consistently and come away with more," he said.

"There's absolutely no question we're making better decisions with this, and we're able to extend that value to our members in terms of better rates, fees and services," he said.

For instance, the AnySite solution allows the 614,000-member BECU to graphically depict where existing and potential members and assets are in a given area and analyze how well a branch might compete there, using census, drive time and other data.

The big credit union uses AnySite data and its own analysis and modeling to both predict branch performance and identify those that are perhaps underperforming given their market characteristics and to look for locations for new branches-neighborhood financial centers in BECU parlance-that indicate they might yield above-average performance.

"Up until a couple years ago, we were doing this by ZIP code and census data. It was very time consuming and not particularly accurate. In fact, it was probably only a little bit better than doing it blind," Bierley said.

"Now we're probably adding four or five branches a year, sometimes closing two or three or four or five, consolidating a couple. The sands are always shifting," he said.

BECU currently has more than 170 ATMs and 45 neighborhood financial centers, according to its Web site. A notably tech-savvy organization, it also is moving heavily toward cashless branches and is using its new tech tools to help automate such decisions as whether to allow a member to withdraw more than the legally required minimum from ATMs or waive deposit-hold requirements.

"We can automatically provide members with greater access to their demand deposit accounts based on a predictive model of a member's risk of a deposit charge-off," Bierley said.

Such tasks are normally reserved for people at desks "and since BECU's operating model has many locations without tellers, there was a need to automate the process that grants this additional access," the BECU market research analyst said.

The data warehouse and analysis solutions also are replacing many of the functions associated with traditional MCIF technology, Bierley said, especially in the areas of communicating with members.

The AnySite and Data Flow solutions are used, for example, to geographically identify members to receive notices of openings, closures and consolidations.

"We locate and notify members that either live in the trade area, opened their membership at that location or make a substantial number of transactions there," Bierley said.

Other tactics include preapproved loan offers the software indicates would be most likely to respond based on predictive regressive modeling of their current product usage, balances and demographics.

The system also is used to produce mail lists to members who have had an auto loan with BECU in the past five years or are near the end of an existing note, he said.

"This modeling also works well to de-select the 30% to 40% that are very unlikely to need a given product," Bierley said.

He noted that the software is used to generate lists of recipients not generate the messages themselves.

Bierley said BECU now is looking at an add-on to AnySite called WinSITE, a market-simulation and statistical modeling tool that adds predictive analytics aimed at branch planning and ATM networks.

[e-mail]