The Art of the Predictable

Customer relationship management (CRM) is on every organization’s radar screen these days, but perhaps nowhere is it a bigger target than in the banking industry.

Traditionally, banks have compartmentalized their customers, seeing individuals simply as savings-account customers or mortgage customers and so forth. In many cases banks have had multiple records of different dealings with the same person and have not put them together to form a complete picture. But that blinkered approach to the customer has changed dramatically over the past few years, largely through the use of CRM techniques.

“Banks are without a doubt among the leading appliers of CRM technology worldwide and the reason is that they are in a service business where the only thing that sustains them in the long term is customer loyalty and long-term relationships with their customers,” says Barton Goldenberg, founder of CRM consulting firm ISM, Inc., Bethesda, Md., and publisher of The Guide to CRM Automation.

Goldenberg maintains that institutions that do not learn to see the entire customer will not survive. Banks will need to follow their customers through their lives, anticipating their needs and forecasting their behaviour. Thus, predictive modeling and careful targeting of marketing contacts with customers are trends in banks worldwide.

Scotiabank’s Life Matters Program

A Canadian institution that is taking CRM to heart is The Bank of Nova Scotia, popularly known as Scotiabank. It is using predictive modeling and targeted marketing techniques to develop marketing campaigns 25 to 40 per cent faster than before and nearly triple its contacts with customers, resulting in return on investment averaging better than 100 per cent, bank officials say.

With more than 1,000 branches and about 2,100 automated banking machines across Canada, as well as investment and corporate banking services and extensive international operations, Scotiabank is among Canada’s largest chartered banks.

Promoting a wide range of service offerings to hundreds of thousands of customers across Canada is no small challenge. Scotiabank’s old way of doing this was to put together separate marketing campaigns for dozens of different service offerings. Each contact with a customer dealt with one specific offering, and the bank might contact the same customer twice in a short period to promote different things. To justify promoting a particular offer, the bank had to be able to foresee enough results from the promotion to cover the full cost of a mailing campaign.

The new way of doing things is a two-year-old program called Life Matters. Every two or three months, the bank chooses a dozen or so service offerings and uses a predictive model to decide which ones are best suited to each of its customers. Based on patterns of customer behaviour and predictive models that are continually being refined, the analysis determines one to four offers to be included in a direct-mail package sent to each customer. The necessary data is sent to a mailing house, which puts together the customized mailings, and information is distributed to the branches about what is being sent to each of their customers. The cost of promoting each individual offering is lower because the expense is spread over multiple services, and customers get one consolidated package rather than a barrage of uncoordinated mail.

The Origin Of Bundling

“The creative format allows customers to quickly sift through what they want, and because it’s targeted, there is a good chance that we’re hitting things that they don’t already have or on which we can offer a better deal than our competitors,” says Jonathan Huth, vice-president of relationship database marketing at Scotiabank. “There’s a brand impact as well. Because products that wouldn’t normally be out in the mail on their own get into the bundle, I think we’re seeing stronger brand recognition among our customers.”

Scotiabank’s evolution toward Life Matters began when the bank built a data warehouse of customer information more than five years ago. The bank began mining that data warehouse to identify which customers were the best prospects for certain service offerings, and contacting those customers based on the results.

This worked so well, Huth says, that more and more parts of the bank wanted to promote their services this way, but that put a strain on existing staff. “My phone was ringing so much that we said if we don’t bundle, we’re going to have to grow this group.”

The idea of bundling different offers together came partly from Huth’s experience in a previous job with Loyalty Management Group Canada Inc., the company that runs the Air Miles program. “Instead of bundling together different financial services offers, we were bundling together different partners in the Air Miles program. That idea was something I brought into Scotiabank.”

Huth was faced with the task selling the bundling approach to the different lines of business. Traditionalists in the banking industry often resist this kind of program at first, and Huth admits there were some reservations about it at Scotiabank. However, he says, “any banker likes the idea of getting more returns at a lower cost, so that was a selling factor. We could show each of the product lines that we work with that they were contacting more customers for the same marketing dollars.”

Thanks to the bundling approach, Huth has been able to promote more of the bank’s services through targeted customer mailings without expanding his staff. “We wouldn’t have been able to handle the volume and complexity of all the different campaigns had we tried to do a single-campaign approach,” he says.

Utilizing The Data Warehouse

The Life Matters program draws customer data from Scotiabank’s data warehouse in order to choose what to send to which customer. The choices are based on demographic data and customer activity. For instance, a young family that does not yet own its own home might receive information about mortgages, while a middle-aged customer whose home is paid for might be told about investment opportunities. The computation involved in matching customers with service offerings is staggering.

“It’s very slick,” says Kyle McNamara, vice-president of data warehouse and decision support. “Depending on the iteration of the campaign, we’ll have a number of different offers that we can send to customers. We don’t want to send them twelve offers simultaneously, so we restrict ourselves to sending them between one and four offers. It becomes a really interesting mathematical problem to select one to four offers from a possible twelve.”

McNamara says the predictive models take into account the demographic characteristics of the bank’s customers, their current holdings and what other cus-tomers with similar profiles have been buying.

How The Models Work

Using predictive modeling, Scotiabank can begin to understand the customer’s likely behaviour and how it can group customers into logical segments. For instance, the analysis to determine whether a customer is a good candidate for a mortgage would start with whether he or she already has one. Assuming the answer is no, one of the next things to look at would be age – between the early 20s and the mid-40s is the most promising age range. Then the system would look at the customer’s last 25 months of banking activity for clues to his or her needs. Paying down credit cards or building up savings could well indicate that the customer is thinking about buying a house.

Meanwhile other predictive models would determine how likely the same customer is to be interested in other Scotiabank services, such as a credit card or mutual funds. Having ranked the customer’s likelihood of being interested in the various offers available, the system would then choose the four – or sometimes fewer – most promising offers for that customer. To choose between two offers equally likely to interest a customer, the analytical models look at which offer is potentially more profitable to the bank.

The number of offers available to choose from each time varies somewhat, McNamara adds; it might be 12 one time, 14 the next. The number of unique combinations is well into the thousands.

The predictive models were developed in-house. McNamara says his group has been refining them since the beginning and expects to keep doing so more or less indefinitely.

“We’ve got a solid test-and-learn program in place where we’re constantly testing different aspects of the campaign,” he says. “With each iteration of the campaign, we’ll improve or replace those models that we use to predict the best offers. We’ve got a systematic way that we can simply drop one predictive model out and plug another one into the process. I don’t think we’ll ever be done; we’ll always be improving it with each iteration.”

McNamara says his staff can create a prototype of a new predictive model, implement it and put it to work within a matter of days.

Integrity Of The Data

The integrity and consistency of data often become issues when companies start mining their data warehouses for CRM initiatives. It is not unusual to find incorrect customer information or inconsistencies in how different parts of the organization interpret data.

Scotiabank never had much trouble with incorrect information – probably not surprising since accurate information about customers has always been more critical to banks than to many other businesses. Nonetheless, the bank went through long and careful procedures to make sure the data was accurate.

McNamara says a bigger issue than the accuracy of data is ensuring that everyone in the bank has the same understanding of what each piece of information really means. To do that, he says, the bank has built a solid metadata repository – that is, a collection of data about the data, to ensure that when people access the data they know exactly what it represents.

The IT Underpinnings

Close co-operation between IT staff and the rest of the business is always an important factor in the success of CRM projects, and Life Matters was no exception. “We’ve been involved right from the beginning,” McNamara says. “We’ve become more involved over the last two years as the data mining algorithms have become more sophisticated and there’s been greater dependency on them. With each iteration of the campaign, we go through a process, and we’re involved right from the start.”

Earlier this year, Scotiabank upgraded its IBM server hardware, so that a modeling process that previously took between 30 and 50 hours to run now takes only three to five hours. That means the bank can now afford to run the process, look at the results, make adjustments to the predictive models and run the process again – sometimes several times – before proceeding with a campaign, McNamara says.

The data warehouse from which customer data is drawn is built mainly on IBM technology, including IBM hardware and the DB2 database management system. Most of the development of the data warehouse was done internally.

Scotiabank uses business intelligence software from Cary, N.C.-based SAS Institute Inc. to extract data from its data warehouse for the campaigns, and Affinium, a campaign-management tool from Unica Corp. of Lincoln, Mass., to manage the campaigns themselves. The resulting data goes to a mailing house, which uses selective insertion systems to put together a custom mailing for each customer.

The same data also is sent out through Scotiabank’s branch network to internally developed contact management software called SalesBuilder, so that branch staff know what offers have been sent to their customers. Depending on the nature of the offers sent out, branches may choose to follow up with their customers by telephone.

The Customer Can Just Say No

Though mailings are more or less continuous, the program runs on a three-month cycle. Roughly every quarter, the design of the mailing pieces is updated to tie in with current advertising campaigns and branding.

An individual customer normally won’t receive a mailing more than once every three months. That gives customers time to respond to the offers they receive, and the branches a chance to update their customer information so that decisions about the next mailing to a customer are made based on up-to-date information. The customer’s response – or lack of response – to offers in one mailing are taken into account in planning the next.

Of course, some customers prefer not to receive this kind of package or to have their bank call them up to promote products and services.

“If a customer ever tells us that they don’t want to receive any offers through the mail or be contacted with any offers, we’ll make sure that we never contact the customer again,” McNamara says.

Paying attention to those concerns is important. Not only customers but bank management are very sensitive about the privacy of customers’ financial information, says ISM’s Goldenberg.

Many banks have even encountered resistance to sharing data among departments. “It’s a real barrier to entry,” he observes.

Looking For Triggers

Scotiabank has recently completed pilot testing of another CRM initiative aimed at responding quickly to specific customer activity.

“We’re looking at triggering contacts that are more timely after some event is detected in your transacting behaviour,” Huth says. “In other words, we’re using ups and downs and changes in patterns to determine whether or not that customer should be called, and called quickly. Some examples of triggers are a significant deposit or a significant withdrawal.”

If a customer makes an unusually large deposit, for example, it may be a good time to offer that customer advice on investment strategies.

The Canadian Approach

Since their business depends so much on personal relationships and requires them to collect a good deal of information about their customers anyway, banks are naturals for CRM and they have been in the forefront of adopting it.

Because competition in the Canadian banking industry is not as intense as in the U.S. and many other countries, Canadian banks have been a little slower than some to adopt CRM, says Perry Marshall, vice-president of emerging technologies at Montreal-based consulting firm CGI Group Inc. He notes that this conservative approach has one advantage: It has allowed Canadian banks to learn from the experiences of banks elsewhere.

There are obstacles to doing a good job of CRM, including the need to integrate disparate databases and the resistance of traditional bankers to the new way of doing things. Nonetheless, says consultant Barton Goldenberg, “if you get to the success stories – and there are many – then the rewards far outweigh these question marks.”

Huth and McNamara certainly believe that’s true in Scotiabank’s case.

Grant Buckler is a freelance writer and editor specializing in information technology and IT management. He is based in Kingston, Ontario.

Would you recommend this article?

Share

Thanks for taking the time to let us know what you think of this article!
We'd love to hear your opinion about this or any other story you read in our publication.


Jim Love, Chief Content Officer, IT World Canada

Featured Download

Featured Articles

Cybersecurity in 2024: Priorities and challenges for Canadian organizations 

By Derek Manky As predictions for 2024 point to the continued expansion...

Survey shows generative AI is a top priority for Canadian corporate leaders.

Leaders are devoting significant budget to generative AI for 2024 Canadian corporate...

Related Tech News

Tech Jobs

Our experienced team of journalists and bloggers bring you engaging in-depth interviews, videos and content targeted to IT professionals and line-of-business executives.

Tech Companies Hiring Right Now