Share the article
twitter-iconcopy-link-iconprint-icon
share-icon
Digital journeysOctober 28 2009

Can decision science improve customer relationship management?

As rising unemployment promotes higher rates of customer delinquency, banks are having to get to know their customers all over again. Decision science can help. Writer Nicholas Pratt
Share the article
twitter-iconcopy-link-iconprint-icon
share-icon
Can decision science improve customer relationship management?

Customer relationship management (CRM) are three of the most provocative words in the retail banking lexicon. Since the term was coined in the late 1990s, there have been countless claims made as to how the proper use of client and customer data can transform the uneasy relationship between a bank and its retail customers and act as a potent fertilizer for cross-selling.

These claims have been difficult to prove and the ambitious plans for promoting banking brands and encouraging greater customer activity have often been based on the suspect supposition that customers actually like their bank and see it as a retail outlet rather than the place that looks after their money.

In the past 12 months there has been an important change in the priorities for banks, and the industry's focus has moved from customer acquisition to customer retention. "The game of musical chairs has stopped and some banks have been left with customers they don't want and are now trying to get rid of them," says Olann Kerrison, head of analysis and published content at UK-based research firm Lafferty Group. "The problem was they didn't know what made a good or bad customer."

The classic credit scoring approach, which has traditionally formed the basis of CRM, is now being applied to default and delinquency management. The teams looking at reams of data for screening loan applications and targeting potential customers are finding that the techniques employed - projecting customer incomes and propensity for default, behaviour scoring to model customer indebtedness - are as equally suited to the collections department, particularly in an economic downturn.

cp/64/Russell Anderson.jpg

Russell Anderson, director of Transaction Analytics

Delinquency forecasting

But these are unpredictable times for the retail banking industry, which presents a number of challenges for the decision scientists. "The UK retail banks all have models for delinquency forecasting that are based on the last downturn rather than the current one, so they need to be adjusted," says John Hitchins, UK banking leader at PricewaterhouseCoopers. For example, the propensity for default in traditional models is impacted because they do not have present-day interest rate drivers factored in.

The ultimate objective for banks is to have early resolution of the issues rather than resort to legal proceedings or repossession. This necessitates an understanding of the customers' ability to pay, the willingness to pay and also using value models to try to examine which of the alternative strategies available presents the most likely successful outcome and gives the most value to the bank and the customer.

But the lack of relevant historical data makes this a difficult task. "As the current issues play themselves out from week to week and month to month, it becomes challenging to overlay a statistical model with constantly changing behaviour," says Steve Davies, partner in the performance improvement consulting team at PwC. "Instead you are relying on the most recent set of data you have without any historical data to establish statistical relevance."

Furthermore, a lot of the traditional measures used in delinquency management are less applicable in the current climate. The link between unemployment rates and indebtedness is not as clear cut as in the early 1990s, and there are more middle-class customers unable to pay their mortgage. "It is a big change for banks," says Mr Kerrison. "What were once their best customers are now their biggest liability. But they have to decide if these customers will be a liability for ever or if they can be helped. There is the thinking that if you help a customer in a time of need, that customer will be more loyal to the bank over time."

A place for decision science

This is where the practice known as decision science comes in - not just in terms of identifying who are the best customers, but also in deciding the best way to interact with them where collections are concerned, says Janice Horan, senior director of solutions management in the EMEA region at US-based producer of credit scores FICO (formerly Fair Isaac).

"It can help banks identify higher risk customers, get good contact information and prevent them being offered any additional credit products and building up further losses. It can also be used to identify the clients with more moderate debt, allow the bank to reach out to them and let them know that it will work with them and, using information from their payment patterns, find ways to take mitigating action." Ms Horan cites the example of payment portals - online debt collection facilities for higher-end customers - where visitors can structure their repayment plans and avoid the humiliation or embarrassment of talking to an operative.

Decision science is also being used to model various economic triggers and the effect that they have on retail customers' behaviour, says Horan. "We have been producing forward-looking scores for banking clients that are using them for customer behaviour rather than investment portfolios. Retail customers have always been subject to market risk but this has not always been modelled."

However, all of these developments are meaningless if a bank does not have the right systems and platforms, says Kurt Thearling, head of decision science at business process outsourcing vendor Vertex, and formerly the head of the advanced technology group at Capital One.

"In debt management, banks need to focus on looking at the data to find the patterns that predict client activity and behaviour, but there also needs to be the agility in the systems that can react to these indications - the timing of call, the tone of a letter, and the workflow into different cues for collections. A clunky system may take so long to react to changes that they are no longer relevant. These changes can be very subtle but they can have a big impact in how the bank engages with its clients." Russell Anderson is the director of Transaction Analytics, a UK-based company which provides decision science consultancy to the financial services and medical informatics industries. Mr Anderson was also head of retail decision science at UK-based retail bank HBOS between 2005 and 2008 and led the transaction analytics function at Chase Card Services before that. In Mr Anderson's view, banks have been slow to realise the full potential of decision science. "Many people in the banking industry equate decision science with analysts, scorecards and Basel models. I view it differently. Decision science is essentially an in-house engineering consultancy and our job is to use mathematical, statistical and engineering methods to automate and optimise decisions, make predictions and monetise data assets."

cp/64/Kurt Thearling.jpg

Kurt Thearling, head of decision science at business process outsourcing vendor Vertex

Making data work harder

To some extent all of the service industries are beginning to understand the possibilities of thorough data analysis in terms of customer management, but very few banks realise the full potential, says Mr Anderson. "The banks have more data than anyone and know more about the economy than the government but none of them has any notion on what to do with their data other than managing risk, improving collections and selling pre-payment insurance."

The risk function, which is a heavy user of mathematical modelling, was unsurprisingly the first sector within banking to recognise the value of decision science, followed by the fraud detection units. But the marketing function is still mostly limited to demographic profiling and response modelling, says Mr Anderson.

"The opportunities with decision science in the UK retail banking industry are huge," says Mr Anderson. "There are benefits in terms of staff utilisation, call scheduling, prioritising customer contacts, deciding which debt to forward to collections agencies, and finding correct contact addresses. But these benefits too often go unnoticed by banks."

Collections teams rarely contain more than a modest number of analysts rather than a team of statistical modellers and mathematical engineers. Consequently the biggest problem in collections is the lack of statistically valid testing. "This is unfortunate because experiments in collections could be designed like medical placebo trials. Once you have confirmation of success, you stop the experiment and go into full production."

Internal politics

Alongside the reluctance to fully embrace the principles of scientific experimentation, banks have also hindered the development of in-house decision science teams with the usual departmental politics and institutional inertia, says Mr Anderson. He talks of lengthy battles with IT departments over the provision of appropriate technology tools for his team and of being accused of wasting company time in writing patents for the bank - "I assured the bank that I did all of my innovation on the weekends."

Vertex's Mr Thearling agrees that more experimentation makes sense but he stresses that the experiments have to be done correctly. "When I was at Capital One we ran thousands of experiments a year that tweaked with the parameters of an offering - from the colour of an envelope in a direct marketing campaign to a 1% change in interest rates. Over a period of a year we would learn a lot about our clients and what they were looking for.

"But you have to have someone that knows what they are doing to design the experiments, particularly for those experiments that grow very quickly. If you make a mistake in an experiment, you no longer have a valid experiment. If you are not careful in the design of the experiment, you could end up with data that is much less useful than if you had done it correctly.

"The real challenge is when you start to put the experiments into production and are presented with compromises: you have changed the experiment. The realities of the production process often get in the way of the theory of the experiments. And that is the problem - finding a way to maintain the integrity of the scientific process in the face of the reality of production constraints."

Solving this problem becomes much easier, says Mr Thearling, if there is a good relationship between the statisticians and the banking business and if the correct systems and platforms are in place. "The challenge is not the mathematical side of things but causing that maths to have an impact on the business. Most banks' systems are not built to take in all this sophisticated analysis and put it into practice.

"Fortunately, the industry has learned to build software and systems so that the decision sciences process can work more fluidly. The data can more easily be taken from a debt collection system and fed to the people that are doing the data analysis."

Commercial deployment

Brendan Clancy, head of customer management and insight at Co-Operative Financial Services, agrees that the commercial deployment of decision science is the critical challenge. "In the past, the analysts have been stuck behind their computers building models but the key has been to deliver the change, end to end, throughout the bank. You can have the best technology and models in the world but if you don't get analysts to think commercially or other areas of the business to understand what you are trying to achieve, it doesn't change a thing."

The organisation has just completed a three-year investment programme into its CRM programme. "It started with getting the data right and getting some good analytical views across the whole customer base," says Mr Clancy. "We then had to buy the right tools and software that were fit for our needs. Then it was about getting the training to the front-line colleagues so that they were buying into the strategy and using the various sales and service messages to help with their decision-making and providing great customer service.

"The single analytical view of the customer was a major part of the programme. We keep a 24-month history of all customers' holdings and transactions across all of our product lines and channel activity. This gives us billions of records to analyse. We use a modelling tool from KXEN to do the building and it's not really about the statistical methods used but about how the models are interpreted by the business and getting the models more commercially focused."

Mr Clancy's efforts will be further helped, he says, by the recent merger between the Co-operative Financial Services and UK building society Britannia, creating a customer base of more than 9 million people. He says: "We are currently in the process of replacing all of our core banking systems. This will give us a truly greenfield site on which we can develop all of the things that our old legacy systems prevented us from doing. From a CRM perspective, it is absolutely critical as it will allow us to develop the products and services our customers need and use the background technology to present the right offer at the right time through whichever way the customer prefers to do business with us."

Was this article helpful?

Thank you for your feedback!

Read more about:  Digital journeys , Fintech