This dimension of the analysis tells the chronological story of the development of the Multinational System for X. The sections refer to specific aspects of the development using all references in the docments irrespective of whether they are historical, contemporary or forward-looking. Within each section, evidence items are ordered so that comments about similar topics are together.
S.1 Organisational context
S.2 Strategy study (Nov 91 - May 92)
S.3 Multinational Requirements study (May 92 - Aug 92)
S.4 Phase I (Sep 92 - Jun 93)
S.5 Use of phase I (Jul 93 - Mar 94)
S.6 Phase II (Sep 93 - Mar 94)
S.7 Phase II goes live (Apr 94 - Aug 94)
S.8 After phase II (Aug 94 - Feb 95)
S.9 Discussion
S.1.1 History
X-Group is the parent company of a large international insurance group that writes almost all classes of insurance business. In 1991 a subsidiary, X, was created to bring together those parts of the group underwriting world-wide insurance risks for multinational companies. Initially, X specialised in providing a package of insurance services for multinational businesses. By 1993 the focus had been diluted as Marine business had been included to give the company critical mass.
S.1.2 Key factors for X
The market for large international insurance was well-defined with few major players. It was a difficult time for the market when X was formed with companies competing on the basis of service. Service needed to be provided locally for handling premiums and claims but centrally for account management. X was created to provide a common identity and to give the business the opportunity to 'stand by its own merits' and so the focus was very much business-led.
S.1.3 Multinational Insurance
In Multinational Operations, underwriters worked with brokers acting for large multinational companies that wanted a programme of insurance covering a variety of risks (eg fire, employer's liability, public liability) for operations in many countries. The business was written by the underwriter arranging for policies to be issued in each country of operation. The premiums were collected locally and transferred back to the controlling office for the whole programme. Claims were mostly handled locally but the local office was reimbursed by the controlling office. Apart from currency differences, there were also local legislative requirements that needed to be met. Re-insurance allowed X-Group to spread the risk but required them to be able to aggregate exposures at a location.
The business was long term with renewal annually and a limited number of opportunities to quote. At the start of the development there were estimated to be 6000 multinational policies world-wide. Multinational companies were becoming more sophisticated in the way they managed their risks and thus more demanding as clients. The current systems in X could not handle these demands.
The nature of the business meant that data was often incomplete and distributed over a number of sites with no efficient methods of collation or aggregation. Existing systems and operational procedures were preventing X from achieving its potential or exercising an adequate level of control.
S.1.4 Physical organisation
The company headquarters of X were in the City of London with the biggest multinational underwriting team. There were also teams underwriting in the States (New York, Los Angeles and Chicago), Holland and about 80 offices in other countries that were more involved in servicing the business. The Managing Director, Tim, was an aggressive Australian. Multinational Operations reported through the General Insurance Manager to Tim and was initially managed by David who had two Account Managers each managing a team of 10-15 underwriters.
Multinational programmes were serviced by a back office Services team in Essex, claims handled by a separate group in Essex and accounts in the City of London. Consolidation of X results was done at the Group head office in the North of England. All X-Group servicing was moved out of London to Essex in 1977 but Multinational Underwriting Services was only formed in 1992. Prior to this, servicing had been done by the underwriters. Lucy became section head for the Services team shortly after it was formed and at that time had four staff working for her. Although she reported for supervisory matters in Essex there were no other groups in Essex with similar working procedures because the XUK business was quite different.
S.1.5 IT culture in X
IT was perceived to be the way for X to compete by delivering accurate information, but in 1991 the competitors' systems were regarded as being old but better. When X was set up it decided to take services, including IT from local existing X-Group companies. In the UK, X got virtually all its IT support from XUK IT Department, including a computer and voice network, at a projected cost of £2.2 million in 1992. However the XUK mainframe systems were inappropriate for large case based multinational insurance which was quite different from XUK's business and X had little influence to change things. For these reasons, and because of the timescales necessary to get work done on the mainframe, a large number of PC systems had been developed by X. Some of these systems were critical to the business but had not been developed by IT professionals to conventional system development disciplines.
When X was created the main problem was a lack of complete accurate information to help with underwriting control, customer reporting and management control. Although it was recognised that X would have to continue to depend on local operating companies for IT for some time, the difficulty of ensuring IT was aligned with business priorities was acknowledged. The problems of data availability and quality and the need to clarify work procedures were also identified early in 1992.
S.1.6 Existing Multinational computer systems
In Multinational Operations local mainframes were used for handling financial transactions and claims. These systems were generally not designed for multinational type business and did not hold the necessary level of detail. In the UK PC systems were used for Underwriting, Services, Overseas claims and Client reporting. In terms of computerisation Multinational Operations was far behind the rest of the X-Group operation mainly because of the one-off nature of their business but also because there was not the emphasis at management level on computerisation.
The first computerisation in Multinational was initiated by an underwriter using Lotus 1-2-3. Around 1990 a previous attempt at computerisation (by XUK and involving David) was stopped by Tim who argued that if X was not part of XUK they would not want to use the mainframe for Multinational. This was one of the reasons for initiating the IT Strategy study. When the XUK initiative was halted, one of the underwriters, Keith, who had a background of technical management in a joint venture in Jakarta but no experience of developing IT systems, suggested to David that he should make use of the work already done with XUK and develop a PC-based system for Multinational. Q&A was used because it was the standard tool in use on X-Group PCs at the time. Many of the underwriters used the Q&A system but it was seen as an underwriter tool and there was no pressure from underwriting management on the accuracy and completeness of the data. When Services was created a Q&A system was established in Essex with data transferred between the two networks at night. Later a Claims system was developed as well. The Q&A system was a single user database with separate files for each major class of business. The system produced basic operational documents and allowed the users to generate their own reports.
S.1.7 IT within X-Group
Although there were some communications betweeen the IT departments within X-Group, each company largely operated independently. In the UK the two major operating companies were XLife and XUK and they both had big mainframe oriented IT departments processing huge volumes of data; XLife used PL1/IMS and XUK used Cobol/IMS. At the start of the Multinational system development separate initiatives for electronic communication were underway in the UK and North America and although the group had standardised on PC hardware (IBM PS/2s with at least a 386 chip), there was no standardisation on PC based tools.
S.1.8 Group Information Services (GIS)
In the UK, X drew IT services from XUK, XI and Information Services Department (ISD) of X-Group but in April 1992 the IT group in XI merged with ISD to form Group Information Services (GIS). GIS was formed to give priority to IT work for smaller operating companies. With these companies being smaller, GIS tended to specialise in PC local area networks (LANs) and package solutions but had inherited its system development methods from larger companies in the group. While there was some support for rapid development techniques, projects were expected to produce their own procedures for new tools or techniques. GIS had an internal Quality Assurance function.
S.2.1 Conduct of the study
The IT Strategy examined X's overall business objectives for the next five years, assessed the current position of IT in X and set out what IT systems needed to be put in place to meet the business objectives. There was a view within X that the strategy was a necessary formality to justify the expenditure because with Tim as the new MD the systems were already seen as critical to getting the business going.
S.2.2 IT strategy findings
The IT Strategy proposed systems for Financial Consolidation and Reporting (an MIS), Multinational Operations and Non-marine Re-insurance which together would let X 'gain significant competitive advantage' and 'avoid significant mistakes'. It recommended a PC LAN based architecture linked to local mainframe systems with separate LANs exchanging data by diskette or file transfer. The use of a graphical user interface and iterative development were seen as key aspects of this architecture with packages being used wherever possible. It was recognised that there would be a need for local variations and that not all cases could be computerised but that development would be controlled at the centre. Moreover, as X was continuing to evolve, the IT systems would need to be flexible.
The emphasis in the IT Strategy, arising from a presentation that Colin did in October 1991, was of the new systems driving the local mainframe systems and progressively taking over the main business functions. It was felt that this would give X the necessary control over its IT but that it would require 'significant management effort and commitment' to achieve.
The Strategy highlighted the problem of incomplete data and the problems encountered in other parts of the group where senior professionals regarded data entry on a PC as a clerical task. It also identified that procedures would need to be clarified and common classifications (eg for territory and peril ) would need to be used throughout X.
The Strategy proposed a plan of work for the next two years with a review at that point. After three years there were no plans to formally review the strategy although new work plans had been developed.
S.3.1 Requirements study findings
During the Requirements study interviews were held with representatives of X in the UK, USA and Holland that resulted in three brainstorming sessions in the UK focussing on Underwriting, Claims and Finance. During this process, many of the separate requirements for Re-insurance and MIS identified in the IT Strategy study were incorporated into the Requirements study for the Multinational system. The IT Strategy study suggested that the existing Q&A system could continue to be used, at least on an interim basis, for data entry but there seems to have been no consideration of this option in the Requirements study. As part of the work an entity model was produced.
Development was divided into two phases with phase I covering Underwriting and Claims for UK, USA and Holland and phase II covering Accounting functions. Phase I included a core database, automatic download and import of data from local mainframes, reporting and analysis of data and efficient exchange of data between X offices. Completion for phase I was scheduled for the beginning of May 1993 at a total cost of £785,000. This was in line financially with the IT Strategy study but the delivery date had already slipped by at least two months.
S.3.2 User involvement
A key feature of the project was the joint management of the project by David from Multinational Operations and Gordan from GIS. This partnership was seen as important by both parties and was made to work as an equal 'man and wife type' partnership with the business dictating what had to be done not the technology. The fulltime commitment of both a senior manager, David, and an underwriter, Keith, was regarded as a significant factor in the success of the project. Senior people were also seconded part-time to the project from Holland and the States but they did not work as closely with the team.
S.3.3 IT architecture
In parallel with the Requirements study an architecture study was undertaken to determine the platform, database, development tool and communications. It was intended that this would provide an architecture for IT within X. A decision was taken to use Unix servers for the larger sites but that development would proceed using a PC server whilst evaluation continued. In the event the power of PCs increased so fast that they were able to go live with PC servers in all locations. This saved a lot of money on hardware and so the project came in under budget. The LANs in London, Essex (2), North of England, Holland, New York, Los Angeles and Chicago were all linked together using ISDN.
S.3.4 Software
There was not an applicable off the shelf package available so the team evaluated a number of client server options. None of the relational databases had adequate Windows development tools and after an evaluation of the alternatives they chose Sybase for the database and PowerBuilder for the development tool.
S.3.5 Methodology
The team developed their own prototyping methodology with users and the development team working together to establish the system structure, validate the entity model and design the database, and, determine the functional requirements of Property underwriting, Casualty underwriting and Claims.
S.4.1 Proposed functionality
Although the IT Strategy study had recommended that the Multinational system should drive the local mainframe systems it soon became apparent that this would not be feasible because it would not be acceptable to the IT people in the different companies. Instead, it was decided that the programme and policy information would be keyed into both the mainframe and the Multinational system. This did not involve much duplicated data because the mainframes held such limited information about Multinational policies and there were only a limited number of programmes involved (just over 500). On the other hand it was decided that claims would be downloaded from the mainframe systems into the Multinational system, presumably because the volume of data made it worthwhile.
Data would be transferred around the Multinational wide area network (WAN) so that all information relating to a programme written in London would be in the London database and all information relating to policies issued by an office would be in that office's database. For offices without the database the information would be faxed to the appropriate office and entered into the system. So, for example, for a programme written in London with policies in New York and France, the policy details would be transmitted electronically between London and New York but by fax between London and France. Claims would follow the same routes with the policy details and claims from France being entered into the database in Essex.
S.4.2 Delivered functionality
Phase I concentrated on the data input and transfer aspects of Multinational with only a basic set of reports. The system consisted of modules for Quotations, Programmes, Policies and Claims. Acceptance of a quotation automatically generated a skeleton programme and confirmation of a programme generated policy records for each country. The system also included a basic ability to identify re-insurance parameters. The data input aspects of the system were much larger (737% was given by one source) than expected.
Phase I also included the installation of the LANs and WAN. PCs were installed on underwriter desks with standard Windows software and IBM terminal emulator software. Email systems were installed to conform to local standards. The users were impressed with the Windows technology.
The system was delivered in June (two months later than planned the previous September) and 3% under budget. This was regarded as a success because the system was so much larger than planned. However the system was very slow on the client PCs even though by the time they installed, the standard machine was a 486 33 MHz PC. Both Services and Underwriting were disappointed with the delivered functionality. In the case of Services there were things that the existing Q&A system did that the new system did not do and for the underwriters things were designed but not implemented. The developers recognised that phase I was overly ambitious and that it might have been better to have left Claims out although there were benefits in including Claims in phase I. They also underestimated the work involved in installing it in the States but felt this was necessary for it to be seen as a global project.
S.4.3 Prototyping
A classic prototyping approach was used where requirements were defined in a workshop, converted into a prototype, presented and refined at subsequent workshops. Prototyping continued for some time because of the complexity of the business and because the team was not experienced enough in rapid development techniques to call a halt.
The prototypes were driven by the business model working first on Quotations, then Programmes and then Policies and Claims in parallel. Property and Casualty insurance was prototyped together but it was felt that it would have been better to develop the Property screens first and then extend them for Casualty. Although developing Claims and Policies in parallel was a bit messy it had the advantage of ensuring that the Claims system was considered from both a Claims and an Underwriting perspective.
Initially the workshops were attended by management with underwriters only being brought in at a later stage to check usability. The approach was very much based on starting from a blank sheet of paper, they did not analyse the way that the existing Q&A system was being used. Moreover they concentrated on the underwriters and did not really consider the Services area although Lucy was involved in most of the workshops.
The full day workshops started in September and continued every week or fortnight until February. Briefing papers were produced before each workshop setting out the aims of the workshop. Early workshops brainstormed with paper models but these were quickly followed up with PowerBuilder screens. Although there were only three or four major prototypes, each screen typically went through three or four versions. Technical debates were held outside the workshops.
Each workshop was attended by David, who generally chaired them, Keith, representatives from Multinational Operations, Services and sometimes Claims. A conference room in the City was wired so that the system could be demonstrated. The workshops were completely 'open floor' but sometimes did go round in circles because of changes in attendees. More was designed at the workshops than could be implemented in phase I.
The prototypes were coded directly in PowerBuilder, which the team found easy to use although it did crash on occasions. Once the prototype had been built it was installed in London and the underwriters were encouraged to put real case data in. The prototypes were also installed in New York and Holland to get their feedback.
Managers found it challenging being given a blank sheet and asked to define their systems but found it very helpful having a working system to comment upon. However it was necessary to manage expectations because development seemed so quick. In general the team found prototyping harder to manage because it was easy to get sidetracked by business issues and hard to say 'no' to changes particularly as they started with a very open attitude to the development and did not have clear boundary lines.
S.4.4 User involvement
The commitment of David, Multinational Business manager, to the project was important because he had the authority to make things happen and the background to see the impact of decisions. The management focus of attendees in the early workshops meant that sometimes the impact of decisions on underwriters and Services was lost. However, it is not clear that involving underwriters who were not particularly IT literate would have been very helpful. In a similar way Services staff sometimes found it hard to contribute to the workshops because they had a mainframe mindset and could not think about alternatives.
There was a view that the workshops gave the business ownership of the system but this does not seem to have been universally acknowledged, although the attitude to GIS was generally more positive than the attitude to XUK IT.
S.4.5 Design
The system was deliberately designed from a client perspective rather than the traditional policy approach of insurance systems. It was designed to support the underwriting process and to be easy to use with table selection wherever possible to improve data quality and speed input.
Mostly the system was developed directly from the prototyping. Re-insurance, which was highly complex and had never before been systemised, had to be rewritten. There were problems with the database design, which just grew out of the original entity model. Although some rework was done in phase I, the database was scheduled for revision in phase II. There were also some inconsistencies in validation and messages. Although they wrote standards specific to the development approach, these were not particularly good and were not always adhered to.
S.4.6 Testing
Originally GIS planned to do unit testing and system testing. However the unit test phase was trimmed because there was no real consolidation period between the end of prototyping and live running. They also found it difficult to unit test in a traditional manner because the system was Windows based but system testing was carried out. Underwriters and Services were supposed to test the prototypes but this testing was limited because they were very busy. However, the system was tested extensively by David and Keith.
S.4.7 Installation
The networking in each location was a major element of installation for which insufficient time had been allowed. The installation of phase I was phased with the Windows technology introduced first along with the Client and Quote parts of the system. The rest of the system followed a month later but this meant that Services could not use the system initially and the rather slow, somewhat problematic start allowed some underwriters to ignore the system.
A manual was not produced for the system because Tim thought the system was intuitive and there was online help. Unfortunately there was a bug in the online help system and America really wanted a manual.
S.4.8 Project management
There was a big push, with GIS working long hours, to get something in by the original deadline of May - although this was supposed to be the beginning of May not the end. The remainder of the system went in only one month later and under budget. The initial plans were formed on a sound basis and allowed for the inexperience of the team with the tools. Project controls such as weekly and steering group meetings were in place. However the plan was not redrawn when the system started to grow and the project team did not have the experience with prototyping to call a halt. The power of the programming language partly cushioned the team and they also worked very hard and cut corners with configuration management.
S.4.9 Team
In phase I the GIS team was led by Gordan who had two people working to him: Scott, his number two, was a technical team leader responsible for the networking aspects of the project and Stuart was the business analyst. Working under them were four or five analyst programmers including Elizabeth, Jim, Matt, Nigel and Ned. Elizabeth was one of the key analysts supporting Stuart but she left in March before the system went live. Gordan ran a flat, individually empowered team, which he felt was very strong with a good mix of skills. Stuart had a good understanding of the business, Matt and Nigel combined good programming skills with an understanding of the business, but Jim and Ned were both more technical and less good at seeing problems from a business perspective.
S.4.10 Quality assurance review
In November a post implementation review of phase I was undertaken by the Quality Assurance person with GIS. He recognised that the project had been successful in meeting budget, time and quality requirements. However he identified that the development approach could be improved in change control, database design, unit testing, technical documentation. He recommended that the approach to prototyping be changed so that it was only used for modelling the user interface. He also included a recommendation to appoint a team leader to manage programming in phase II but this was not included in the management summary. The team felt that his report was biased because he was 'steeped in traditional development methods'.
S.5.1 Organisational preparation
At the time of the Requirements study the resource implications of the new Multinational system had not been examined in depth but it was assumed that additional clerical staff might be required in Essex and that underwriter input could be absorbed within existing resources. It is not obvious that this was ever looked at in more depth or that system co-ordinators were appointed in the USA and Holland as recommended. There were some comments that the system was introduced at a particularly bad time for the underwriters when they were short staffed but that it was politically expedient to introduce the system on time because David was about to leave Multinational Operations. Once phase I had been installed they did have an operational review between October and December 1993 which took a broad look at information flow within the Multinational business in the UK and made some minor recommendations for changes in the Multinational system.
The job of getting the system used by underwriters fell to Jenny. She joined Multinational Operations as an underwriter manager (one level below David) on 1 June 1993 and suddenly found that she was taking over the IT liaison role from David on a part-time basis. Although she was concerned about the need to prove herself on the business side, Jenny generally liked the role of IT liaison.
In Services the need for additional staff was eventually successfully argued but took time because the local head of department had little involvement in the unit as it was different from the other units in Essex. Jenny's background in servicing cases and the XUK discipline helped the Services unit to put forward their point of view.
S.5.2 Phase I conversion
In the UK and Holland data in the existing Q&A system was converted and downloaded into the new Multinational system. The idea was that underwriters and Services people would go into each programme and complete the fields that were not in the Q&A system. The conversion process took all of June 1993; for the first two weeks Services were updating the Q&A system and preparing the data for download, this was followed by two weeks when the data was converted and the system installed. The outcome of this was that Services started using the system with a backlog of work as they were without machines for at least two weeks.
About 85% of the data was downloaded but the quality of the data was doubtful because of problems with conversion and the poor quality of some of the data in the Q&A system. Conversion problems only affected some of the programmes so it was not until they were updated that the problems were found. One of the frustrations was that validation rules on the new system prevented some of the errors from being corrected by the users. The problems with data impacted on the image of the system and provided underwriters with an excuse for not using the system.
The UK decided to check the programmes on a rolling basis when they came up for renewal. Although this would take a year to complete they did not have the time to update the data in one hit at the start. Even though systems were clearly 'flavour of the month', Jenny still had to persuade the underwriters to do the updating and in this she was not always supported by the other managers whose focus was on getting the business.
S.5.3 Use of phase I by underwriters
The official view of phase I was that it was a success as a system, a development strategy and as a new PC based architecture. However from an underwriter perspective it offered little, was not particularly convenient to use and it did not work well enough. In part this was because the system encouraged underwriters to be more systematic and culturally they were not interested in doing this.
In the UK it was decided that the underwriters would be responsible for checking the accuracy of the data at programme level with Services responsible for policy level data. The aim of having underwriters involved in data entry was to ensure they had ownership of their cases in the system and because they were best placed to enter programme details. Although new programmes were quite time-consuming to enter (one to three days was quoted), most work was renewals which was much simpler. Although this was a change from the existing situation where Services input all details to Q&A, it was not long since underwriters had input all details because Services was only formed the previous year.
The underwriters varied in their enthusiasm for using the computers; some believed in keeping control of their cases while others did not see the systems side as part of their job. This attitude did not seem to be related to computer expertise; in fact it was often the younger underwriters who were most resistant and their attitude was described by their manager as a 'cultural thing'. In addition, the underwriters only had one machine between two. One of the consequences of the rolling programme of conversion was that it could be several months between training and underwriters using the system in anger. There was no manual but they did have a refresher workshop in January highlighting system shortcuts.
Much of the process of case handling provided by the system supported Services and Accounts rather than the underwriters. In phase I the system was seen as a burden by many underwriters but Jenny likened it to a baby - it gives very little but when you see the first smile you know it was all worthwhile. Once the data was up to date, the system would provide management information so that underwriters would no longer be asked to provide exposure information. However, an ongoing problem was knowing what level of information to store about location data in order to provide the necessary detail on exposures.
Three months after the system had gone live David thought that Holland were very enthusiastic about it and had checked that all their policies had been correctly converted. He thought that the US were making good progress on using the system but had further to go. In practice, three months later, when Gordan visited the US for a management review, he reported that the Claims download was not operational in the US because of missing policies. Amongst the developers there was a view that 'the further west you are the more IT illiterate you are' so that while the Dutch embraced the IT ideals and there was some enthusiasm amongst underwriters in London, the US did not even know how to use a mouse.
Gordan was disappointed that there was resistance to the system given the use of prototyping and put this down to an innate dislike of change. However nine months after delivery there were signs of increasing enthusiasm.
S.5.4 Use of phase I by Services
When phase I was initially installed it was of no use to Services because it only allowed clients and quotes to be input. Throughout phase I, Services continued to use the Q&A system to service programmes which had not yet come up for renewal. One of the features of this kind of insurance is that there can be long delays in getting all the information necessary to process a case. This use of two systems meant that they had a very heavy workload and because they were full-time users of the system they suffered more from the crashes. Services also felt that phase I was inferior to the Q&A system which it was replacing because it was slow and long-winded. Multinational Claims did try to use phase I but was not successful, in part because of missing policy data.
S.5.5 Phase I system problems
Phase I suffered from data inaccuracies, system crashes, errors in calculations and reports for overseas territories, and, was very slow. One particularly visible error was that help screens did not display properly. There were database crashes caused by a problem with Sybase that tended to affect Services more than London because they were accessing the database through a router. As this was a more complex system than Q&A it was more difficult to diagnose problems. Throughout the operation of phase I it was necessary for Keith to smooth things over to improve the image of the system.
S.5.6 Phase I data quality
Throughout phase I the quality of data in the system was poor. This arose because of a combination of problems with data conversion from phase I and a lack of commitment from underwriters to get the data right. In the UK this was partly because of the decision to wait until the programme came up for renewal before checking the conversion, so that it would be at least a year after installation before all the data was accurate or before management reporting was useful. So, for example, early in 1994 when Gordan was in the US to try and encourage greater data accuracy, there was an earthquake in Los Angeles but the system could not report X's exposure. In addition, not all underwriters entered their new programmes in the system.
The development team realised at this stage (six months into phase I) that they should have put more effort into ensuring data quality and actually built some time into their plans for phase II, but they still saw it as the users' problem. There was a view that there was a loss of continuity when David left that contributed to the data quality problem. Actually Jenny was temperamentally better suited, with her Services background, to sorting out data, but she did not have David's authority. Although there was on the spot help from Keith on how to use the system and what data to enter, there was no detailed monitoring of data quality throughout phase I and a certain lack of management commitment (excluding Jenny). This was a long standing problem that was inherited from the Q&A system and was one of the reasons why conversion was not useful. In retrospect it would probably have been better not to have converted the data from Q&A, then it would have been more obvious what was on the system, and there would have been fewer problems with data quality.
There was some discussion about using data entry clerks to enter data but the consensus was that this would cause more problems because it was not just a matter of entering the data, it needed to be interpreted correctly. One of the problems in America was that some fields were interpreted differently.
S.6.1 Transition to phase II
There was a period of indecision following the installation of phase I while the business was deciding the priorities for phase II. This allowed the development team to focus on support which was useful but the project lost some momentum as a consequence. One of the reasons for the indecision was that the scope of X had broadened out; at this point the Marine project was added to Gordan's responsibilities. Work on phase II started in September but some urgent development work continued to be done on phase I right through until the end of the year.
S.6.2 Rationale for phase II
One of the main elements of the first stage of phase II was a re-engineering of the database and the PC client application to improve speed. This included upgrading the version of Powerbuilder. Phase II also included an extra policy class, security features and additional reports. One of the key limitations on the speed of phase I was the client PCs and the time taken to load screens. This was addressed in phase II by loading screens on system start up, so trading start up time for screen switching time. Phase I suffered from a lack of design arising from the prototyping process and phase II aimed to redress this by improving the design of the database and the application. The database was re-engineered to include referential integrity and the application was re-written in a modular fashion. The overall aim was to have an application that would be easier to maintain and extend.
S.6.3 Functionality in phase II
In practice there was very little change in functionality between phase I and phase II. The proposed enhancements to minor classes were dropped and a limited number of new control reports were included. The development of a suite of MIS reports and the installation of an adhoc reporting tool, Q+E were postponed. The view seems to have been that without the data, the MIS reports were not essential. The main change was that the whole system was effectively rewritten although with two exceptions it looked the same: the method of initiating renewal was deliberately changed and the order of policies was inadvertently changed. During this period work also continued on the import of claims data.
S.6.4 Development approach
As a consequence of the indecision from the business and the departure of Gordan's number two in September, detailed plans were not produced for phase II but the re-engineering was scheduled to go live at the end of February 1994. This was later revised to mid March to incorporate some specific reporting requirements. There was not quite the same pressure to deliver as with phase I but of course phase II was replacing an operational system so had to work properly from day one.
The development approach in phase II did not follow the same prototyping approach as in phase I and a number of technical design documents were produced during development. However, the development team underestimated the extent of re-engineering as phase II was effectively a rewrite of phase I. This led them to view phase II as more straightforward than it actually was.
S.6.5 Design
The new database structure was redesigned to incorporate the comments made by Sybase in their review of phase I. The job was done by a strong member of the team, Matt, who was later recruited by another company as a database designer.
The Powerbuilder code in phase I was favourably reviewed by consultants from the suppliers but was rewritten in phase II to make it more modular and thus more maintainable. An object oriented structure for the client was designed that limited the communication between windows and database access. This improved error handling and made the system more robust. This architecture was designed and documented by the two senior developers Matt and Nigel. There was nothing in Powerbuilder to support the approach (and some workarounds were necessary because of limitations in Powerbuilder) so it relied on the programmers to stick to the rules. It might have been better to write the system in an object-oriented language such as C++ but it would have taken longer and they did not have programmers with C++ experience. The reporting side was separately designed by two more junior members of the team
S.6.6 User documentation
No training or user manual was provided for phase II because there were no real changes.
S.6.7 Hardware problems
During the development of phase II there was a major problem with the use of Sybase in phase I. PCs accessing the database remotely over a router (those in Essex, Los Angeles and Chicago) began frequently to lose their connection. This impacted on the phase II work as the team spent two weeks putting a trap in to handle the error gracefully. Sybase initially washed their hands of the problem but, after Gordan threatened to convert the system to Oracle, located the problem in the Novell communications protocol. Converting the system to a Unix communications protocol at a cost of £10,000 seemed to solve the problem. This reflected the fragility of the PC/Novell version of Sybase that eventually led to the replacement of the larger servers with Unix machines late in 1994.
S.6.8 Phase II testing
Phase II was subjected to unit and system testing but there was no acceptance testing because the change in functionality was so small. System testing also included some further unit testing and only revealed minor errors.
S.6.9 Installation plans for phase II
Installation of phase II involved converting the database so the changeover had to happen over a weekend. Stuart drew up a plan for the installation period and they planned to have someone in America and Holland when it went live. There was no fall-back plan for what to do if installation failed. Installing new versions of the software was much simpler as they had a utility that checked when a user logged on that they were running the most up to date version and upgraded them if they were not.
S.6.10 Team
At the start of phase II a Finance Director was appointed in X, George, who took over overall responsibility for IT from Tim. George was a qualified accountant who had been a consultant involved in projects with a high IT content. At the beginning of 1994 (two months before phase II was due to go live), George appointed another accountant, Kevin, as IT manager for X. Kevin had little experience of IT and although it gave Gordan a focal point for communication with X, it also absorbed a lot of his time and made it difficult to resolve complex technical decisions. While Gordan was reporting progress to Kevin he was still reporting to GIS on resourcing matters.
Under George's new structure, Gordan was also given a more consultative role distancing him from a day to day involvement with the Multinational project. This only really formalised what had already happened because Gordan had been too involved in the Marine project in the latter part of 1993 to have time for the day to day management of Multinational. At this point the Marine project was taking off and Gordan diverted Scott's replacement, Manny to run the Marine project and broadened Stuart's role to lead the Multinational project including the phase II installation. He was concerned at the time that Stuart did not necessarily have the managerial or technical skills but aimed to continue having a close involvement himself. At the time Gordan was fully aware of the importance of project management in IT project success.
Stuart saw the team as a happy bunch of intelligent people and his new job to advise them and provide a buffer between the development team and the business. At this point the strong technical people on the team were Matt and Nigel. In fact Matt was regarded as so good, he was promoted two grades. Jim was still regarded as not good at seeing the business perspective and Gordan still did not feel he could completely trust Ned although in working with a new graduate, Alan, he was improving. Alan joined phase II of the project straight from completing a PhD in mathematics and after a week's Powerbuilder course started writing the reports with Ned. Each member of the team was given a lot of responsibility and contact with senior business people particularly when compared with the working conditions in other parts of X-Group.
Throughout phase II, business involvement was led by Jenny on a part-time basis but Keith continued to support the system in London on a full-time basis.
S.6.11 Project management
Phase II was originally supposed to have been led by Manny but his arrival was delayed by GIS. Gordan did raise the under-resourcing within GIS but it was not effectively raised with X who were not necessarily aware that Gordan was spending a lot of his time working on the Marine project. The complexity of phase II was under-estimated by the team so that the under-resourcing was not viewed as that serious.
An overall plan for phase II was produced which had already been extended by two weeks before Stuart took over the coordination of phase II installation. When he took over he produced a detailed estimate of what was left to be completed and the deadline was extended by a further two weeks. There was limited monitoring and control throughout the development of phase II. Team meetings were held every one to two weeks but generally Gordan led by giving people responsibility and then letting them get on with it. There were regular meetings with the business people that involved some of the team members. When Stuart took over he recognised that more structure was required and aimed to have more detailed plans with short scale milestones. Unfortunately in drawing up the plans, Stuart had to rely on the judgement of the team members because he had not been involved in the development of phase II and did not have the technical knowledge to question their estimates. His revised plan did not allow enough time for testing.
S.6.12 Financial Accounts investigation
The other element to phase II was a Financial Accounting extension to the Multinational system. The analysis for this was undertaken during the Autumn of 1993 in parallel with the re-engineering of the database. At this stage the plan was to leave the Accounting in the UK and US with the mainframe systems but develop a Financial Accounting system based on Sun Accounts to run alongside the Multinational system in European countries. Sun Accounts had versions to run in each European country and was written in Sybase. At this stage the Financial Accounting sub-system was regarded as a relatively straightforward integration job. The plan was to develop it using a small team with a mixture of accountancy and technical expertise to run initially in Holland. The plans for this were presented at the end of January 1994.
S.6.13 Marine investigation
The Marine system was a separate development for X only interfacing with the Multinational system when reporting information about X as a whole. The analysis for this work was also undertaken by the team in the autumn of 1993 with the first presentation to the business in December 1993. A package solution was recommended for Marine and the choice of package was presented to the business on the same day as the Financial Accounting presentation at the end of January 1994. Marine was a big project (£2 million) involving a complex conversion and the management of an external supplier who was doing the tailoring work on the package. Contracts were due to be signed with the external supplier shortly after phase II went live.
S.6.14 Future plans
Around the end of February 1994, X led by the new Finance Director, George, started to think about longer term IT needs for the whole of X. While this did not directly affect the immediate developments on Multinational, the system was seen as the core of many of the future developments. Despite the negative comments, the business was proud of the system and saw its expansion as important to their future competitiveness.
S.7.1 Installation of phase II
The week before phase II was due to go live, Keith was asked to do some acceptance testing and he found about 27 immediate problems with phase II. As a consequence the business users wanted to delay installation because they were aware of the importance to PR of a good launch to phase II. As most of these errors were relatively straightforward and many had already been found, Stuart wanted to continue with the release. He discussed the issue with Gordan who was on a short visit to America and they decided to go ahead in part because Gordan would then be in America when the system went live. The users raised the issue with Kevin but he was guided by Stuart and Gordan.
The day phase II went live the users were not necessarily aware of it being installed or of the changes. Phase II required the database to be converted and this led to some further data problems because of inadequacies in the phase I database design such as null entries in key fields.
S.7.2 Phase II installation bugs
People were unanimous in declaring the installation of phase II a disaster: there were 120 errors and queries in the first month, far more than they had with phase I. There were a large number of relatively minor errors, some serious data conversion errors, some unexpected changes in functionality and a Sybase problem that caused the system to hang. By the middle of June most of the bugs had been fixed but data problems continued to emerge. In particular records transferred to the States under phase I could not be accessed in phase II. The data problems caused a loss of faith in the system. There were also one or two cases where people used the conversion problems as an excuse for work that had not been done. The problem with Sybase was serious in that some servers just hung sporadically for no particular reason. The Sybase problem was not necessarily related to phase II but further sapped confidence in the system. The team did not have the low level access to the Sybase application to do anything about it and the Sybase suppliers could not fix it.
Although there were a large number of errors, the number reported included many enhancements and repeated reports of the same error. The team felt that the underwriters tended to over-react to the errors while Services were far more sanguine although they were far more dependent on the system. This was perceived to be a cultural thing influenced by Jenny's attitude, which was that of a customer rather than co-developer. The over-reaction of the City was illustrated by a problem escalated to management level that turned out to be a LAN cable that had come unplugged. In contrast the underwriters in Holland were far more laid-back about the system problems. System problems were also used as a scapecoat so that management reactions tended to be far more extreme than the reactions from those using the system. In general Gordan protected the team from the worst fallout from the problems of phase II, although there were some comments that the developers were unnecessarily defensive in their reactions to the problems.
Shortly after phase II went live, Alan was given the job of managing the process of error reporting and correction. Errors were supposed to be reported through Keith using a standard form. Services found it difficult describing problems, which sometimes GIS could not recreate. They recognised the usefulness of a central point for recording errors but going through Keith just made communication more difficult. Services found it useful when they had a programmer on the spot but generally felt that the errors had been handled well. Some of the underwriters reported errors to Jenny who passed them through Kevin thus bypassing the normal reporting procedures. Jenny felt that the business users were not always given accurate information about error fixing although there were regular management meetings and lower level meetings to ensure the users were kept informed about the progress of bug fixes. In July 1994, GIS started a Multinational newsletter to aid communication with users. However during this period the frequency of the IT Steering Group meetings appears to have slackened off.
Alan's aim was to solve as many of the small errors as quickly as possible to minimise the number outstanding. He established a release process to ensure that fixing one problem did not introduce another problem. He also tried to make sure that things were tested properly before release although with the adhoc correction of data errors it was not easy to have a set procedure. One of the things that he discovered was that it was possible to be too helpful so that the users expected the team to solve data problems that it was more appropriate they should solve themselves using the system.
As far as the users were concerned the problems with phase II were extreme and continued to emerge for the next six months. They felt that if only they had been allowed to do thorough acceptance testing, the problems would have been avoided. They also felt that some of the errors were so basic that they just showed a lack of care. However there was also a sense that part of the problems were caused by a lack of management commitment to properly resource the development team. There was some resentment at Gordan's involvment in the Marine project. Despite the problems, people I spoke to seemed to be committed to the system and kept telling me what a good system it would be.
The development team felt that they almost pulled it off but under-estimated the scale of the rewrite and hence the importance of user testing. The Sybase problem also clouded the issue. From their point of view it was disappointing that the technical improvements in phase II were ignored because of a large number of relatively minor bugs.
S.7.3 Phase II capabilities
Even without the errors the actual capabilities of phase II were not well regarded by the users. The speed of phase II in most areas was significantly improved but was still regarded by the users as too slow with some reports being particularly slow (15-20 minutes). The electronic transfer between countries was still not working fully. The number of locations used meant that re-insurance reports were inadequate and in general the reporting facilities were poor. The adhoc reporting capability of the Q&A system was due to be replaced by Q+E but this was not available in phase II and reporting was hampered by the lack of confidence in the quality of the data.
S.7.4 Views of phase II
Within the Multinational system in phase II there were 822 programmes (8925 policies) with 514 controlled in the UK, 149 in the US and 159 in Holland. Attitudes to phase II varied considerably between the different groups using the system although there seems to have been a general view that the system had potential but was not yet delivering that potential because of problems with data quality and a lack of functionality in the system.
Management in the City of London believed the system was good and would form the basis of ongoing developments that would improve their competitive position. Use of the system was just held up by the data quality problems.
There was a view that phase II was a public relations disaster with the City underwriters who were supposedly very critical of the system. It seemed to me that it was more that they were not particularly interested in making the system work and so were less likely to overlook system failures. Theoretically the system should have helped the underwriters to provide a better service to their clients but in practice the underwriters continued to work with their manual paper files in part because they needed the history. There was no standard form for these files and much of the information was not precise so that loading details of cases into the system was not straightforward. The problems with phase II allowed the data loading impetus to drift.
The system did make processing of cases easier particularly at renewal but my impression was that this was a relatively small part of an underwriter's job given that each underwriter was only responsible for between 9 and 20 cases. The other facilities on the PC, particularly email, seemed to be more useful to the underwriter than the Multinational system. So while there was a cultural antipathy towards using computers ('I was a highly paid typist', 'we don't like using PCs') this did not extend to stopping them using the system when it brought them real benefits ('the mail system is a brilliant tool ... it's replacing faxes'). In addition both underwriters I spoke to could see the potential of the system. The development team could not understand why, with so few cases each, the underwriters did not just get them in and right.
Despite the relative newness of the Multinational Services team there seemed to be a 'them and us' type relationship between the underwriters and the servicing team. This perhaps reinforced the underwriters' view that entering the data was not really their job although Lucy, Services' section head, reckoned that use of the system made the underwriters more sympathetic to what the Services team had had to put up with. As far as Services were concerned, phase II still failed to replace all the functionality that they had had in the Q&A system. Despite this, and the problems of phase II, Lucy at least was still very positive about the system's long term use.
Despite some positive comments from the development team it was apparent that neither aspect of the Claims system were effectively used in phase II. The download of claims information from the various mainframe systems suffered from some problems with the data extracted and the data quality problems of the policy data in the Multinational system. The loading of claims data coming in manually did not appear to have happened. The main problem seems to have been a lack of interest in the system in the Claims department but I did not manage to talk to anyone from Claims.
The problem of getting the underwriters seriously to use the system in America continued into phase II. There were problems of differences in terminology between the UK and USA but perhaps more important there seems to have been a lack of senior management commitment to the system despite their involvement during prototyping.
Holland was reported to be generally happy with the system although they were affected by the same bugs as everybody else. When the system was installed on a standalone basis in Germany, training was provided by the Services Manager in Holland.
S.7.5 Post implementation review
In July and August 1994 a post implementation review of phase II was undertaken for X. This was a more thorough review than that undertaken following phase I because it was perceived that phase II had suffered far more problems on installation that phase I. It was conducted by Colin who had had no involvement with the Multinational development since the strategy study and so was seen as independent. He also interviewed me as part of the review to increase the independence of his report. The objective was to learn from the problems with phase II particularly with respect to GIS procedures and to seek views from the business on the benefits of the system. The review was not expected to look at phase I but as it found that many of the problems stemmed from phase I, it was extended to consider these issues.
The report found that the system was 'widely regarded as a good system, well conceived and potentially offering significant benefits to X' with the potential to be important in X's future business success. Although the bugs were under control, there was a continuing lack of project leadership, the data load was incomplete and the system was not fully used. Despite the problems of phase II, GIS were regarded positively within X.
Project leadership of phase II was weaker than in phase I, although not non-existent (which is how Jenny interpreted the report). However, Gordan's reduced involvement did mean that the scale of the rewrite was not fully appreciated by all concerned. The report questioned the technical design of phase II arising from the lack of leadership but this criticism was not accepted by the development team. The report suggested that the development team were under considerable pressure from X senior management to deliver and that this may have led them to take short cuts with standards and procedures.
The report did highlight the need for the data to be complete for the system to be useful but this was not itemised and there was some confusion even six months later about the distinction between increasing the use of the system and improving data quality.
The report's main recommendations focused on the leadership issue by recommending the appointment of full-time business and IT project managers and increasing the role of the Steering Committee. The report recommended that a key task for these two project managers was to get the system into full use and get the data right. It suggested that X and GIS senior management should address the resourcing problems caused by lengthy reshuffles. It also recommended a number of technical checks.
The reaction of the directors to the report seems to have been fairly positive in that they saw it as a way forward. The development team felt that the report was unnecessarily negative. The good aspects of the system were not emphasized and some problems were reported out of proportion. Although most of the inaccuracies in the first draft had been addressed they still felt the report had an unfortunate slant. The report did recognise the role of the senior management in the problems of the system but its reception in GIS was influenced by the organisation's fear of failure. Despite their criticism of the report, GIS felt that the report 'focused the venom' and that subsequently there was a more positive attitude to the system.
Jenny was much happier with the report and felt it was fair. This was probably because, by Colin's own admission, the report had to satisfy Jenny and Neville. As a consequence of the report, she attributed most of the system problems to senior management in both X and GIS.
S.8.1 Unix migration
In June 1994 Gordan recommended migrating the large servers to Hewlett Packard hardware running Unix and a more advanced version of the Sybase database software at a total cost of £200,000. The main aim was to cure the Sybase crashes and improve reliability but this version of Sybase also offered Replication server. There was a view within the team (perhaps with the benefit of hindsight) that they should have gone for Unix initially as the PC/Novell combination was technically not appropriate for a database server.
The original plan was to migrate the UK server in mid August but the decision to go ahead was delayed because of a breakdown in communications between George and Gordan. In the end migration was scheduled for October 1994. The biggest unknown in the migration was the acquisition and set up of the hardware but they also allowed for five days Sybase consultancy in the budget because of the development team's inexperience with Unix and Replication server.
Matt, who designed the database for phase II planned the Unix migration but left the team at the beginning of September to go to a better paying job in New York. Alan, who had only joined the team a year earlier with no previous computing experience, took over and managed the Unix migration. He was assisted by another member of the team and some new team members who were used for testing.
Migration to Unix was only planned for the large sites (UK, US and Holland) because the database problems only affected those servers supporting large numbers of users. In practice Holland with 20 users had not had problems but was migrated to Unix because of the Financial Accounting sub-system; Sun Accounts had had problems with the Novell version of Sybase. A Replication server running on the development machine would provide a warm standby machine. They also planned to use the Replication server for adhoc queries to protect performance on the live machine.
In theory migration was a simple task of running the scripts to build the database on the new machine with an identical structure and use a bulk copy program to copy the data from one database to the other. In practice there were some minor changes needed to the application and the original scripts did not match the live database but overall the migration was straightforward. Generally speaking Alan seems to have done well on the migration particularly given his inexperience but Gordan did report a communication problem with the support people which jeopardised the live date because a comms card was not ordered.
The Unix migration was very intensively tested with unit testing, system testing running the two systems in parallel and stress testing with the users running two systems in parallel. Gordan admitted that they had probably over compensated for the lack of testing on phase II and the testing did delay the release date.
Migration was generally trouble free although there were some security issues on day one. However, everyone had hoped that there would be performance improvements because the servers were so much more powerful, and in practice the migration did not make any real difference to performance. In consequence, performance remained a real issue and when the new project manager arrived, Jane, she blamed the database design despite all the work that was done on it in phase II.
S.8.2 Data audit
By August 1994 there were still quite a number of underwriters who were not up to date with loading their cases into the system. There were still problems emerging from the original download from Q&A which meant that those cases had not been looked at on the system for over a year. The teams were reorganising and the underwriters had agreed that cases could not be handed over until they were up to date in the system. With the proposed Financial Accounting sub-system it was essential that the data was completely accurate. A data audit was initiated from within X to assess the quality of the data in the Multinational system. Originally it was going to cover London, Holland and the US but America got a stay of execution because they did not have the data loaded.
The audit was undertaken largely by one of the underwriters I spoke to, assisted by Keith and Jim from the GIS team. The audit was managed by X's underwriting development manager reporting to the MD. The aim of the audit was to validate the paper records with the data in the system. It also aimed to look at how underwriters were using certain fields and identify training needs.
The first cases to be audited were Noah's, the underwriter doing the audit, and despite his computer literacy, this still revealed some errors in his data. The audit considered a 10% sample of each underwriter's cases and produced a report itemising the discrepancies. Most of the audit was undertaken by comparing paper records with print outs of the system but they did speak to each underwriter about their cases. It was a time-consuming exercise and could take a day to audit a single case because of the numbers of paper files involved. Jim investigated the reasons for discrepancies looking into the history of policies and identifying common mistakes. He also built a little system for recording the results of the audit in order to do trend analysis.
Each underwriter was given a report of their cases. A separate report was produced for management in UK and Holland and an overall report was produced. The reports did not compare the UK and Holland but in some aspects the UK was better and some, Holland was better. The report was sent to the directors but was not generally released and it was October before Neville got a copy of the report.
Overall the report found that the data completed by both Underwriters and Services showed a lack of accuracy and attention to detail. The report identified those screens and fields where people did not appear to understand what to insert, the number of transcription errors and those cases where it appeared just anything had been filled in. One of the underwriters claimed that most errors arose from the initial data loading period. As a consequence of the report a number of changes to the system were recommended including changing the layout of screens, tightening up on validation and some data quality reports.
S.8.3 Organisational changes
By mid 1994 X had established itself as a serious business and saw IT as a key to its development. By the autumn they were predicting significant increases in income from Multinational insurance without increases in staff. The company was restructured around the concept of profit groups with each group containing an IT person.
At the same time, Services were expanding with a view to them eventually taking over responsibility for entering all the policy details for renewals and new programmes. However, the expansion meant that they took on new people with no experience of Multinational insurance and it would be several months before they would have the necessary expertise.
In June 1994, Gordan's role was further refined although he was still working for GIS. He had the title of Systems Development Manager working to X's IT manager, Kevin. This meant that Gordan had a far wider role than just Multinational. However, Kevin resigned and although X initially sought to replace him they had difficulty recruiting. In the end they split Kevin's job into development and operations. In October, Gordan got the job of Development Projects and Standards Manager working directly to the Finance Director, George. He was now employed by X and spent about half his time in London. The advantage of this promotion was that he controlled the whole IT development budget for X including the overseas offices. He also no longer had to worry about the networks because that was part of the Operations Manager's job. When he took on the job, he did a presentation to the Directors of X where they pledged their support and said they wanted to see all of X systems renewed or replaced. This included a plan to stop using the XUK system in a few years' time. Gordan was also responsible for hardware and development standards.
As part of the new IT management they decided that each project needed an IT manager, a business manager and an executive sponsor. Neville was given the business responsibility for Multinational. He was supposed to be working on it for 50% of his time and his job was on the line if he did not get the business issues sorted out. Although this had been decided in October, it was not announced until early January when his appointment was backed by the MD giving him authority to divert resources into getting the system running properly. Neville was the same level as David but Jenny also continued to be involved in the project.
In June 1994, in response to the problems with phase II, the directors of X decided that the Multinational project needed a clear project manager. Gordan agreed that Stuart was not up to the scale of this job as Multinational was rolled out over Europe. The need for a project manager was also echoed in the recommendations of the Post Implementation Review. In part this was because the scope of the Multinational project had grown and was seen as the core of a range of developments for X.
Normally recruitment of project leaders in GIS would be from within the organisation but in this case Gordan did not feel there was anybody internal who was up to the job. He recruited a contractor by approaching leading management consultancy organisations. He found that all the candidates he interviewed were strong in some aspect: good business knowledge, experience of the client-server architecture, experience of working overseas, but all were expensive. In the end he employed Jane from Touche Ross who was very strong on the methods side but had no previous experience of insurance. She started in late October so the recruitment process took about four months and the budget for her was 10% of the spend on the overall project. One of the reasons for employing an external person was that they would find it easier to be tough on X senior management and that seemed to be true.
Colin was also brought into the team to manage the Financial Accounting sub-system and other accounting elements of X systems. Although he was nominally working for Gordan he was if anything senior to Gordan in GIS terms.
When I stopped observing the project the new management structure seemed to be working in that the project was re-energised.
The development team continued to be part of GIS even though Gordan was working for X, but it was no longer subject to the resource constraints of GIS. There was an ongoing discussion, supported by the Post Implementation Review, on the relocation of the development team to London or Essex because some felt that this would improve support and communications. Despite misgivings amongst the developers about the viability of such a move and mixed messages about communications, part of Jane's reorganisation of the team involved a small team being based in London. The team in the North of England also expanded with the intake of a number of consultants.
Keith remained involved in London but was sidelined in many of the new developments; he was no longer necessarily first port of call on the system for the underwriters with a development team based in London. He was not part of the mangement group controlling the Multinational development and his training role seemed to be diminishing.
S.8.4 Multinational phase III
Following Phase II, there were a variety of strands to the development of the core of the Multinational system which were loosely known as phase III.
Arising from phase II and the data audit there were many small enhancements to the system (and some errors continued to emerge). These included further work on re-insurance and tightening up on validation. In all there were over 150 requests and these were prioritised to be delivered in releases in March and June 1995. In addition further work was done on the system manual and documentation of the transfer mechanism.
The download of claims information continued to cause problems post phase II. Eventually in autumn 1994, Holland managed to establish a procedure for periodically downloading claims but in the UK there were still data problems that prevented some claims from attaching to policies. Attempts to download claims in the US were abandoned because the quality of the data was so poor.
A reasonable number of standard reports were included in phase II but the users still felt they needed to produce their own reports as had been possible with the Q&A system. To enable them to do this, the team had planned to install the Q+E product initially in London followed by Services and Holland. Although Q+E was a standard product the team wanted to create some special views, test performance and do some training before releasing it. Initially, Q+E was scheduled for release in London in July 1994 but in practice it was not implemented in Holland until late in the year and had still not been implemented in London in January 1995. The delays in implementation of Q+E appeared to be due to difficulties over providing appropriate views and because the work was not given a high priority.
The Multinational system was installed in Germany on a standalone basis in July 1994 with the aim of linking them into the network once their cases had been properly loaded and they had moved offices. A large part of phase III, beginning in early 1995, was the rollout of the Multinational system to many other offices.
One of the most successful developments in Summer 1994 was a separate sub-system for McDonalds, X's biggest client. This system was based on a cut-down version of the Multinational system running on a standalone PC. It allowed claims information in five countries to be collected and emailed to Ireland for analysis by McDonalds. The system was easy to develop and showed how the system could be used for specific clients. One of the reasons the system was successful was that this was a new client and so there were no data quality problems as there were with the main Claims download.
During Summer 1994, Stuart worked with a number of X's other important clients to identify ways in which additional information (mainly claims related) could be collected and provided to clients.
In parallel with this, Jim produced a report on process automation. This looked at the gaps in the new system for Services; areas where they were still using other systems or IT could be used to improve their efficiency. The main recommendations were for a diary, scheduling and control system, a fax/email gateway and the installation of Q+E. The developments were put on hold until the new year because the users did not want further automation until they were satisfied that the system was performing correctly.
One of the consequences of the problems of phase II was that the developers became much more cautious over the development approach. Initially, Multinational was developed using evolutionary prototyping but by the end of 1994 this approach was out of favour for all but small developments with prototyping being viewed as a technique for requirements capture in other cases. Of course one key difference was that they were much more familiar with the technology than when the Multinational system was started.
One of the developments of phase III was the introduction of new standards linked with the CASE tool already in use in other parts of GIS. Much of the work was done by external consultants but supported by Jane and Colin and following recommendations in the Post Implementation Review. One of the objectives was to reverse engineer the data so that they had at least a physical model of the database. The CASE tool was also used to develop a model of the design of the Multinational system. This was being developed gradually as enhancements were done to parts of the system. The CASE tool was used merely as a design tool, there were no plans to use it to produce PowerBuilder code. As a consequence of this work the development approach became much more traditional.
To support the roll-out plans to other countries the team were planning to use the Replication server capabilities of the Unix version of Sybase with a new server in Australia for locations in the Far East and Australia. In addition they were upgrading their communications technology from ISDN to frame relay. Sybase were providing consultancy advice on these new developments but Jane was also heavily involved.
S.8.5 Financial Accounting sub-system
When the Financial Accounting project team was formed in April 1994, the aim was to develop an integrated Multinational Financial Accounting system initially for Holland but with a view to installation in other European countries. At this stage Gordan was aiming to make X independent of the XNetherlands system but the Finance Director clearly thought that X should be aiming to become independent of all mainframe systems. The Financial Accounting sub-system was seen as a relatively simple development that should be implemented by July 1994.
By June 1994 the developers had realised that the accounting side of Multinational was more complex than they had envisaged and had developed a mockup of the screens to use in a workshop to help with the writing of a formal specification. At this stage it seemed as if the complexity lay in specification not development because all the debtor/creditor aspects were to be provided by Sun Accounts; they were hoping to have a system in by the end of the year. During this period they began to consider using the accounting system eventually in the UK and so the UK started to monitor developments. In practice the business specification and projected timescales were presented in August. A two phase approach was proposed delivering the Sun General Ledger system in November 1994 and the integration with the Multinational system by June 1995. However the number and complexity of the screens to be added to the Multinational system was questioned by the UK and despite the early work it seemed as though much of the work could not be done in Sun Accounts.
At this point, in November 1994, Colin took over the project. Colin organised the project with a steering group involving both Holland and the UK and produced a plan, which showed implementation in Holland by November 1995 and this was only achieved by stripping the system down to the bare essentials. Colin produced a new specification in January 1995, which he asked the users (steering group and directors) to sign. One problem which took time to resolve was the links with the Multinational system because if underwriters changed fields once a policy was live it would be a 'nightmare' in accounting terms. One of the difficulties that Colin was going to have to manage was holding the requirements of the Accounting sub-system constant when the Multinational system was changing. Although Colin was derogatory about previous attempts at specification, the specification agreed in January still appeared to be high level and did not have an identifiable data model.
Colin was proposing to develop the Financial Accounting sub-system in a traditional way and had already produced a note outlining how the programming was to be organised which was similar to how things were organised when Colin joined as a trainee programmer in 1977. His argument for this approach was that accounting was a very precise problem so there was no room for prototyping although he did plan to keep the users involved with the development as much as possible.
Colin's plan was to produce a system at the outset that would be appropriate for the UK although it would be installed in Holland, France and Germany first. This meant that it would be the end of 1996 before the data had to be exactly right in the UK. Although it was expected that each country would need minor modifications there was a real sense that they were defining the accounting policy for X through the system.
S.8.6 Marine system
There continued to be no link between the Marine and Multinational systems except possibly in an overall information system. The system was using the same kind of technology as the Multinational system but was developed by a package supplier. Implementation started in June 1994 with a delivery date in 1995. It was a big project with some difficult conversion problems but Gordan was very confident in Manny who was leading the team.
S.8.7 Post audit data quality
Despite the findings of the data audit and the changes in management structure, data quality problems continued to bedevil the Multinational system. The 40% error ratio revealed by the audit and the continuing phase II problems meant that they missed the end of year deadline for getting all the data sorted out in the UK. There seemed to be mixed views amongst the users about whether Neville's appointment would solve the problems but the signs when I lasted visited the City (in January 1995) were not good: there were still some cases that had not been updated at all on the system, management were not using reports produced by the system to control data quality, little had been done to improve attitudes to data quality following the data audit and changes in underwriting policy were not being reflected in changes in the system. One of the problems in getting the data accurate was the consistent handling of awkward cases within the system.
The Claims part of the system was only really working effectively in Holland although there were signs that the situation with Claims was improving in the UK. There seemed to be continuing problems over understanding how the electronic transfer of information was supposed to work and the Claims part of the system was still very slow. There were still old Q&A systems containing claims information prior to the start of the Multinational system which needed to be loaded into Multinational.
In January 1995 the decision was taken to relaunch the system in America because the data was so poor and the original underwriters had all left. The relaunch involved developing new training materials and a user manual and Jane and Jenny training each of the underwriters. A plan was agreed for loading the data in America.
In rolling out to Europe far more emphasis was placed on training than previously. Moreover, in February 1995, George charged the development team with managing the data quality initiatives.
S.8.8 Future
By February 1995 there was a clear view on the role of IT within X although the IT strategy had not been rewritten partly because they had 'gone out of fashion'. The role was far wider than Multinational but then X encompassed more than Multinational by 1995 and IT was seen both as a driving force and in a supportive role within the business. The new management structure of IT seemed to be promising a higher profile for Multinational and there were achievable plans for rolling it out to other countries. However, there was not clear evidence that the 'hearts and minds stuff' of X acting as one organisation and sharing information had really been solved.
S.9.1 Conception of Multinational system
Officially the Multinational system arose from the recommendations of an IT Strategy study conducted for X in late 1991 and early 1992 by Group Information Systems (GIS). X had just been created as a separate organisation focussing on multinational insurance and needed to consolidate its world-wide activities which were using a variety of IT systems in the different companies in the X-Group. As the development of Multinational began, a group was formed in Essex to provide back office type servicing to the multinational underwriters.
The aim of the system as presented in the IT Strategy was for an information system to support multinational underwriting but the scope broadened during development to include operational aspects. By 1995, the Multinational system was regarded as the core of an IT system for X. The relationship between the Multinational system and the other systems used by X varied accordingly.
The IT Strategy and Requirements studies established an architecture for Multinational based on a client-server architecture running on PC LANs with Unix servers. The Sybase database server was used with the client application developed in PowerBuilder. The system was actually developed using PC servers, but when Sybase proved to be unreliable in this environment, had to be converted to Unix in late 1994.
S.9.2 Construction of Multinational system
An evolutionary prototyping approach was adopted for phase I. Part of this approach was a strong emphasis on the business and David, the manager of Multinational Operations and Gordan from GIS, jointly managed the project. This partnership was regarded as very successful. Phase I focussed on establishing the infrastructure in the first three countries to get the system (UK, Holland and USA) and developing the data entry screens. It turned out to be much larger than planned and there was a delay of about two months in its full installation.
There were flaws in the database design of phase I and the client application was very slow. One of the main tasks in phase II was to re-engineer the database. In practice the majority of the system was rewritten in phase II. In phase II analysis began on a Financial Accounting extension to the Multinational system. Initially this was regarded as a relatively minor development for Holland but the scope of the application grew to include the UK and its development time increased substantially. This became part of phase III along with a number of separate development initiatives covering adhoc reporting, client services including a sub-system for McDonalds, and further process automation.
Apart from Sybase problems, phase I had some errors but very few in comparison with phase II. Phase II had so many errors that the users lost confidence in the system. Most of the errors had been fixed within four months but this delayed the start of many of the phase III developments.
Phase I benefited from a concentration of senior management input but the team changed at the start of phase II. George, the Finance Director was appointed in X with responsibility for IT and he replaced the MD's involvement. David left Multinational Operations and Jenny who was less senior replaced his role in development. Gordan's number two, Scott, left and was not replaced and Gordan was allocated an additional large project for Marine. Gordan was further diverted when George appointed Kevin as X's IT manager. Some of the problems of phase II were attributed to this loss of leadership and it was eventually addressed in phase III. Gordan was promoted to have budgetary responsibility and had two new project managers working for him. Jenny's boss was made responsible for the business aspects of the system with express authorisation from the MD.
S.9.3 Delivery of Multinational system
Data for phase I was converted in Holland and the UK from the existing PC system. Unfortunately this data was not very accurate and some errors were introduced by conversion. When phase II was installed the data was converted from phase I and a number of further errors introduced.
In the UK the plan was for underwriters to check and add additional data to programmes as they came up for renewal. Thus it would be a year from the installation of phase I before the UK data was updated and Services had to continue to process accounts that had not yet renewed on the old PC system. As a consequence of the problems of phase II and a reluctance amongst underwriters to use the system, the data had not been updated eighteen months after installation of phase I.
Holland had a smaller number of programmes and checked the converted details immediately. The underwriters were also rather more enthusiastic about using the system. However an audit of data in UK and Holland carried out fifteen months after installation revealed data inaccuracies and omissions in both UK and Dutch data.
In the USA the data had to be entered from scratch and the resistance to using the system was such that it had to be re-launched in the USA eighteen months after going live.
The problems with data quality impeded the smooth running of the Claims and electronic transfer systems. However there were also technical problems, problems of understanding and of ownership that had an impact.
S.9.4 Conclusion
Although there were technical failures in the development of the Multinational system, notably the database design of phase I and the errors in phase II, the development was generally undertaken in a professional manner. Despite this eighteen months after the first phase was installed the data was still so poor that the system was not useful as an information system.
© Clare Tagg 2000