There is nothing like the heady sniff of a windfall pay-out to make a building society investor suddenly learn to spell demutualisation. Last month, one of the last stalwarts against conversion, Bradford & Bingley, took the plunge on the stock market and floated as a public company.
But exactly how many of the society's three million plus investors and mortgage holders have found themselves shareholders? Identifying who is eligible for share allocation has been a major project for the last 18 months.
The problem has been that, like all banks and building societies, B&B's IT systems were originally developed along accounting lines. These were held on a clutch of core, in-house-developed, mainframe-based applications which looked after things like mortgage and investment management. This made it easy to tot up how many mortgage or various types of investment accounts there were, but less easy to see how many customers the society had in total, since many people had several accounts. Establishing a single, comprehensive customer database, rather than a set of accounts databases, meant sorting through all accounts data to find out what accounts were accreditable to which customers.
When it came to demutualisation, big money was at stake. Rules of flotation meant every customer who qualified for a share allocation in the new company had to be contacted and invited to apply for shares. If B&B missed out on anyone who then claimed allocation after flotation, the new company would then have to find extra shares. This would have to come from the small stock of shares set aside by the building society for this purpose. If they proved insufficient, the society would need to buy the requisite amount of shares on the open market at the prevailing price - a potentially expensive exercise. Naturally, the society wanted to keep this liability to a minimum, by ensuring members eligible for shares were identified before flotation.
This meant customer records had to be exceptionally accurate. "We have 3.1 million customers," says Steve Broadbent, project manager at B&B, "Our assessment was that we would be making payments to 2.7 million of them."
If the customer database was out by a few per cent, it meant several tens of thousands of qualifying customers could be left in the cold. A 5% inaccuracy meant 135,000 customers not getting their windfalls on flotation and whose share entitlements would have to be sorted out afterwards. "It would hit at the wrong time for the City," says Broadbent.
Together with the society secretary, Broadbent decided what level of database accuracy should be achieved, given the financial risk to the new company against what it would cost and the tight timescale before flotation.
For Broadbent, tasked with ensuring the society's computerised records would be up to the demands of the demutualisation process, there was good and bad news.
The good news was that B&B already had a customer database. This was set up as a stand-alone DB2 database in l995, extracting and collating customer information from the disparate account systems into a single set of records.
The bad news was the database was not accurate enough. "It was great for marketing, but not good enough for demutalisation," says Broadbent.
The society had used software from customer information quality specialists, Innovative Systems, to check the accuracy of records when the database was first created. When the demutalisation programme was launched, another audit started. "By the time we decided to float, it was 93% accurate," says Broadbent.
The society knew 93% was not good enough, but the questions were: what was good enough? How much would achieving it cost? And could it be done in time? "I suggested from a purely technical perspective that we could get 99.5% accuracy," says Broadbent. "Any more would be in the realm of fantasy."
Because the implications of being only a few percentage points inaccurate in a customer list of over three million are huge, the society decided anything less than 99% accuracy was unacceptable. "That was a fairly ambitious target," says Broadbent.
"We do not think anyone else in financial services has that accuracy. We know other societies demutualising haven't achieved it - when they issued shares, they still had five years of sorting. The Halifax has said it still has 150,000 share allocations unattached."
Reckoning about 2.7 million customers would be eligible, even "if we are 99% accurate with our membership database, then 27,000 records could still be wrong," says Broadbent. "That is 27,000 customers who have been paid twice or not at all."
Flotation has made customer shareholding worth about £600. An error of 1% could mean the newly floated company would face a maximum liability of £1.5m. That is not a hit a new company would want to announce to unforgiving City analysts and institutions. But against that risk had to be set the feasibility of forcing accuracy of the membership database any higher.
With the overall demutalisation programme costing around £67m, the IT proportion of the cost was £5m, says Broadbent. But setting that cost against the financial risk from a higher level of membership inaccuracy, the sum was reasonable.
But money can, if necessary, always be found - the same is not true of time. The issue of demutualisation was first raised in April 1999, which gave the society the impetus to start working on what needed to be done for demutualisation. A flotation date of 4 December 2000 emerged early last year to be recommended to members, and was formally agreed in July 2000. The flotation went ahead successfully.
Hardly catching breath from Y2K and euro compliance programmes, Broadbent was appointed IT business head of the flotation programme and its IT domain architect. "I am project manager for IT effort. My team make the technical decisions needed," he says.
He was faced with a tight timetable. The new, highly accurate database needed to be ready early in 2000. "We needed it in February 2000 for everything to take place for a December launch of the public company," he says.
"Between April and September l999, we decided what systems and processes were needed," says Broadbent. Because the society had already used Innovative Systems software and was not only familiar with it, but already held a perpetual licence, the decision was taken to use the same software to clean up the existing customer database to the level of accuracy needed for demutalisation. The society also enjoyed "an extremely good relationship" with Innovative Systems, and was one of the supplier's first UK customers. A member of the society's staff even sat on the software company's board and represented UK customers in the US.
"The choice of tool was a no-brainer," says Broadbent. But the challenge was not in software, but in size. "My challenge was scaling it up," says Broadbent. In l995, we used seven people to create the original DB2 customer database." For the demutalisation project, the scale was larger. "Between September l999 and February 2000, the [data cleansing] team peaked at 200 people," he says.
Before the data cleansing process could kick off in September l999, Broadbent ensured the rules were clear and agreed. "We revisited all rules for duplication we had set up in l995 [for the original customer database]," says Broadbent. "Setting up parameters was complex".
The society then mailed members asking them to check their details, to set up a baseline. One problem was of family members who shared the same name, who were asked to confirm they were separate account holders. "We got an 85% response rate," recalls Broadbent.
He attributes this high figure to society members realising money in the form of share options was at stake, which gave incentive to return forms. The exercise netted over two million changes, which needed to be made to database records.
With the duplication parameters agreed, Innovative Systems' tool set, Match, could start the cleansing process. The front end collects data from the legacy accounting systems and identifies duplicate name and address information despite misspellings, character transpositions and missing information. It identifies copies by analysing word meanings and patterns within records. "If there are two similar pieces of data, like the name R Jones and RB Jones, it will categorise each element into up to four thousand different combination of attributes," says Broadbent.
The software then applies rules drawn up by B&B as to what constitutes a single customer, despite variations in details such as additional initials, by collating other fields such as date of birth and the role a member plays on an account. The software "can then say which records are an automatic match [definite duplication], which are not, and which are fuzzy matches".
A record definitely identified as the same customer but another record is automatically merged on the DB2 database. A record definitively identified as a separate customer is not. The fuzzy, unsure matches are sent to a human operator to follow through and ascertain whether they should be merged.
Because of the financial implications of an inaccurate customer database, the rules on whether two records were or were not about the same member were set not only by Broadbent but by the society's company secretary, just as the overall level of accuracy attainable had been set jointly.
"We needed duplication rules set very high up in the organisation, so if anyone disagrees, tough - it is what the company secretary says," says Broadbent.
Setting Match to trawl through the DB2 customer database revealed 300,000 fuzzy matches. "That was more than expected - we'd anticipated 250,000," says Broadbent.
By repeated checking, the project closed in on the target level of accuracy. Using Match, it took the team 26.5 hours to flow data past the duplication rules and recalculate the database. "We did it every fortnight instead of the usual once a quarter in normal working practice," points out Broadbent. "We did not need extra processors because the new version of the Innovative Systems software had been improved, so we could check data faster," he says. It was a gruelling process for all. "For seven months we wrote, checked, chased. Every fortnight we revised it," he says. "In the early part of the project, between September and November l999, we saw a 6% improvement."
That pushed accuracy up to 98%. Then, with the quick wins over, "it slowed down", says Broadbent. By February 2000, the database was ready for auditing. "KPMG said it was 98.2% accurate," says Broadbent. "In early February, I did final technical tweaking and got an extra 0.9% accuracy, bringing us to 99.91% accurate." One tweak was identifying customers who were split up and bringing them together again.
"About 150,000 customers had been split, even though they were the same person. We brought them into the fold," says Broadbent.
One respite was granted to the data cleansing team. Since April l999, to prevent profiteering from the flotation, no new members were admitted. Only existing members were allowed to open a new account if they wished.
With the audited customer database ready in February, the society used it as the basis of its March mailing to invite members to the April AGM, and again in May for a special general meeting, using the clean data. "We had no significant complaints. We felt we had done a good job," says Broadbent. Using automated data cleansing "saved us months of manual data sifting and checking". The society estimates that without automation, it would have taken 200 people six months to check records manually, instead of a few weeks.
Eligible members were mailed again prior to flotation asking them to specify whether they wanted to keep or sell their shares. The clean database has now been used to issue payments to those who have sold shares and register those who kept them as shareholders.
Broadbent is already thinking up things to do with his new database once the dust from demutualisation has settled. "Customer relationship management is the perfect opportunity," he says. "We have started to understand our customer base in great detail, so we can go forward. That is our strategic policy," he says.
He is also keen, having put so much effort into achieving high levels of customer data accuracy, on keeping it there. "We are setting up a department of around seven people to keep data quality high," he says. "At the moment, the highest accuracy is 99.8%. If we keep it at that, we will be awfully pleased."
The organisation: the Bradford & Bingley Group is the UK's second largest building society with over 200 branches and nearly 180 local agents. The largest high street provider of independent financial advice, it has 630 independent advisors and is also the UK's fourth largest estate agency with nearly 380 branches.
Bradford & Bingley needed to assess its 3.1 million-member database accurately to meet the legal requirements for share allocation during demutualisation, saving an estimated £27m cost per 1% error rate.
Use of an automated data cleansing process enabled the IT team to achieve the 99.5% accuracy target, make 2.3 million changes to customer records, and save 100 employee years 'of effort.
Bradford & Bingley chose Innovative Dictionary, Innovative Match and Innovative Household from Innovative Systems.
Innovative Dictionary enables customer profiling and analysis by identifying customers and their relationships to each other, standardising significant information and flagging data inaccuracies within customer databases. Innovative Match identifies duplicate name and address information by identifying similar records despite misspellings, character transpositions and missing information. Innovative Household groups customer records based on similarities in name, address, postal code and other user-defined fields.
What the BuyIT experts say
Chairman, BuyIT, and president, CSSA
Customer intelligence is a powerful competitive weapon. Whereas this data was once regarded as customer "history", it is now used to help companies identify and develop profitable customer relationships. Keeping any database up-to-date is a chore, but when the database comprises millions of records accumulated over long periods, accuracy is a real issue.
Customer profiling is a technique for leveraging customer intelligence. It provides a complete view of a company's relationship with a customer. It provides a comprehensive insight into a customer's spending patterns and needs, and can be used to support highly targeted sales and marketing campaigns.
We thought this example not only explored use of customer profiling and matching tools but also told the real story - the cost benefit calculations, hard work and careful planning needed to achieve results. Like all tools, it is what you do that counts. Achieving 99.5% accuracy required crafty manoeuvring through data. Every business benefits from increasing value and accuracy of its customer intelligence. Bradford & Bingley Building Society had no choice. But it can now use it to offer a professional and personal service to customers. That will repay investment over and over again.
Marketing communications director, PSDI, and member of BuyIT
The message is clear: understand customers. In today's competitive business environment, knowledge is power. Detailed information about customers can provide you with the competitive edge. Bradford and Bingley's example highlights the importance of establishing, developing and maintaining detailed customer profiles through well-planned and implemented technology. The building society turned necessity into a resource that will allow the company to personalise relationships with its customer base and bring long-term benefits.
The key to the case study is understanding the mutual benefits of a strong relationship between an organisation and customers. Detailed data enables businesses to target sales and marketing campaigns accurately, providing potential customers with products and services that are of interest to them, rather than bombarding them with "junk-mail", wasting time and the organisation's valuable resources. Businesses learn a great deal about customers needs, tailoring products or services, even creating new ones, allowing them to take advantage of this knowledge.
Profitability depends upon building trust, goodwill and loyalty with your clients. This is the essence of effective CRM. Customer retention must be integral to any strategy if profitability is the ultimate goal.
Managing consultant, CRM at Logica, and chairman of the CSSA's CRM group
This case study is an excellent example of a project with clear objectives using cost-benefit analysis to set a target of bringing its customer database to 99.5% accuracy.
Although Bradford and Bingley had a very specific reason to establish a clean database, the problem faced is common to many organisations.
Companies either have historical product-based systems which record customer details separately or have merged with others and end up with several databases. Either way, the hardest part of setting up a database to provide the single customer view central to customer relationship management is the cleaning of data. Even with the use of sophisticated rule-based matching software, this was a major project for Bradford and Bingley. Without tools, it might never have been done.
The project was cost-justified in terms of the demutualisation programme. Provided processes and systems are put in place to maintain accuracy, Bradford and Bingley now has a major asset in its database, which it can also use to increase customer service levels and loyalty, develop relationships and grow its base - all of which become more important in its new role as a public company.
Could you do better?
Do you think your organisation has shown best practice in its implementation of an IT project? If you want to tell us about it e-mail us at [email protected]