Go Back

Can your Grandmother underwrite yet?

Part 1

In 1989, following an IT apprenticeship in banking, I went to work for a major London Market reinsurer. Walking around the underwriting room, I was mildly surprised to note that there was not one computer in sight. Their chief underwriter's stock, and somewhat indignant response to any suggestion of computerisation in his area was that, if he got one in "to do his job", he may as well get his grandmother to do the underwriting. The contrast between that operation and the bank's exchange dealing rooms, just 100 yards away, was staggering. £400 million premium income in a year and not a penny of it written with the assistance of a computer lest Granny take over the show. That stunningly misguided view would nearly have been funny had it not been widely representative of the London Insurance Market at the time, in which computer systems were predominantly used for filing risks, tracking claims and handling the volume of technical and financial accounting.

So, what has changed in the last decade? Soft market conditions have dictated that efficiencies must be exploited to the full, underwriting mistakes must not be made and corporate capital has moved into Lloyds and brought with it corporate views on analysing data. Underwriters themselves have become a lot more specialised and their average age is reducing as fast as their computer skills are increasing. The view that the well-informed underwriter is the better underwriter is becoming well established and the demand for software to assist in the underwriting process is now stronger than it's ever been. The problem was that, having shunned the computer for so many years, there simply wasn't the choice of off-the-shelf software available to satisfy the demand. 10 years of development should have put that right by now, you might think. So, can Grandma now load EasyWriter2000 for Windows on her laptop and write a hundred-million pound book with a 25% ultimate loss ratio? Luckily, things aren't quite that simple.

We are focussing here on what software is available specifically to the underwriters in the London Insurance Market, falling into some broad categories:

- Pricing, Risk Analysis and Statistical analysis applications - OLAP databases & MIS tools - e-commerce systems - Artificial Intelligence systems - and, of course, the embedded facilities provided in the major underwriting packages

In Part 1, we will concentrate on what is available to "better inform the underwriter", in other words the Information Systems.

Let's deal with largest group first: The market is still dominated by the big "strategic" packages from suppliers such as Rebus (Lines, IRIS, Genius, Writarsure), Intech (Open Box, Open Co.), Room (Subscribe 2000), Eurobase (Synergy, Synergy at Lloyds), Olympic (Visual Writer, InsuranceWorks, IMACS), Sherwood (Senator, Sceptre), PMSC (Re+, I+) and CSC (Sics/NT). These packages offer a wide variety of functions to the whole business and are designed to run an entire (re)insurance operation from a single suite and, in the majority of cases, that is perfectly feasible.

With only a few exceptions, most of these have grown up from a mini-computer heritage dating, in some cases, back to the early 1980's but the majority of them now have been redesigned to offer modern user-interfaces. That some of the older designs still have significant user-bases is remarkable and a great credit to both their original designs and current support. They were never equipped with the graphical user interfaces that we take for granted now, nor were they "modular" - the component parts of the system being very tightly bound and interdependent. The underwriting information systems, such as they were, tended to be very basic, inflexible and inseparable from the package. With the growth in demand for information presentation and analysis, however, that position has been changing.

Software vendors are now moving towards providing much more powerful, stand-alone information systems that work with or, theoretically, without the main package: Rebus (Knowledge Server, Lloyds Confidential), Eurobase (Business Objects), Sherwood (Eureka), PMSC (Amis) and longer-term plans to do the same being mooted by Intech (Crystal Reports). Another change, apparent on closer inspection, is that some package providers are beginning to use other people's software to provide components of their suites, such as the reporting & information systems and databases. This new approach represents a considerable change in the underwriting software industry: it may not be so very long before purchasers will be able to pick and mix the best components and get their friendly systems integrators to glue them together. Successful integration, however, will always be a development-heavy exercise until there is some general consensus upon a "standard" interface between modules. We are some way away from that Utopia with the fledgling concept of business objects still far from general acceptance.

By and large, when it comes to information systems, the "newer the better" rule still tends to apply. Great leaps have been made in the past few years in database technology and data analysis and presentation techniques. If the user requirements are for a graphical interface, graphs, cubing, cut-and-dice, drill-down analysis and so on, then the older packages are quite unable to deliver on their own, being still shackled to their non-graphical interfaces and single-dimensional, relational databases. So buying an older package on the strength of its information services might not be a good move. But what of the many organisations who already have one - why replace something that works just to satisfy one specific area of user demand? Replacing these systems is a very expensive business. The market is full of smaller organisations, writing say £10 to £20 million a year, who are still using "green screen" packages bought in the 1980's. They are mainly running on midrange hardware, like IBM's AS/400 that, as processing power and storage becomes cheaper by the day, can be easily replaced to keep up with the burgeoning data and throughput requirements without any major changes to the software. For them to consider spending many tens of thousands of pounds to acquire the latest package technology is inconceivable. What they need to do is add value to their existing investment, not to throw it all away and put the business to major upheaval and risk.

This is where specific, targeted applications come to the fore. Pricing tools, statistics tools, risk analysis and exposure analysis tools represent an excellent way to add value to the older systems in which users have made considerable investment and whose databases contain years of accumulated information that is impossible to value but without which it would be impossible to trade. Staying focussed on the information systems category, there is really not that much available at this level, which is quite surprising given the demand. TriSystems' product, Experience, is one of the few examples at present, being an underwriting statistics analysis system that has actually designed for underwriters as opposed to actuaries or accountants.

Next come the OLAP (on-line analysis program) databases, data marts and data warehouses. There are many so called "multi-dimensional" databases, such as EssBase, SAP, Amis, Pilot, SQL Server V7, and so on, available on a wide range of platforms. These are not specifically targeted at insurance or any business sector for that matter but, correctly implemented, they can fulfil some of the underwriter's data analysis requirements. Cut-and-dice and drilldown enquiries can be carried out very easily and quickly.

As always, though, there are a number of caveats: firstly, and this applies to any information analysis systems, if the underlying system is badly "coded", having poor data quality in terms of statistical analysis classes, etc., then no amount of technology will make it valuable to enquire upon. Secondly, and of considerable importance, is the need to design the data models that these OLAP databases use and this can take a lot of expensive consultancy. There is a growing trend for software providers to sell these databases with pre-defined insurance analysis models. Although this approach could be interpreted as lacking flexibility, it does ensure that the end-user is getting value from day one. Another consideration with any generic database is that it is likely to be supplied with a generic enquiry tool-set. Forrest & Trees, Showcase Analyzer, and other such off-the-shelf OLAP enquiry tools will work with most of the OLAP data-engines. The problem is that they were designed with management accountants in mind. Every training course that has ever been given on these tools has been of the "widget unit sales by territory by salesperson" variety - simple crosstab type reports that even Excel can do standing on its head. Underwriters need specific types of report that do not fit easily into this simplistic paradigm. Doing a multi-currency development triangle showing gross and net loss ratios using an OLAP database and a standard cube navigator is not quite so easy. Again, an underwriter-friendly front end to a multi-dimensional database is a considerable asset and Experience from TriSystems is one such tool that can do this.

With any add-on database or function, since they are nearly always sold separately from the main package, they will have to be integrated into the existing infrastructure. This almost inevitably mandates development of a middle-tier interface layer to handle the flow of data between the two. Middleware is the subject of a whole article in its own right but tools such as IBM's MQSeries are an ideal way to automate this process in a controlled and robust manner across a wide variety different hardware platforms. This, at last, allows us to choose the most appropriate platform for the main systems and not be bound to use that for the information systems as well. It also means that companies who already have a stable "heritage" system on a midrange or mainframe computer can keep it for what it does best and farm the data out to a modern data-mart on an up-to-date server to keep the power-users happy and well fed. Underwriters as power-users; how times have changed.

In summary, then, there is now a reasonable choice of software available from the major package vendors over a variety of hardware platforms that enable underwriters to be better informed from the data locked-up in their own corporate databases. Some of this software is, however, tied to the main strategic packages from the major vendors who dominate this market sector. Although more of these systems are now being offered as stand-alone, non-partisan solutions, it is somewhat surprising that few smaller software vendors have been willing or able to sell specifically targeted, "tactical" information products into this environment.

So, how about Granny? Thankfully she's still at home knitting oversized jumpers but in Part 2, where we'll be looking at the other types of underwriting software, artificial intelligence and the ubiquitous e-commerce, things may start to change.

Part 2

Welcome back to our fin de ciécle look at underwriting software for the London Insurance Market. Last week, we looked at what underwriters could use to get better informed from the data held within their own corporate databases by using the information technology part of Information Technology. This week, we're going to have a look at some of the other available out there, along the lines of:

- Pricing - Exposure and Risk Analysis - e-commerce - Workflow/IDM - Artificial Intelligence and Expert systems

Pricing software has always been a bit of a paradox. The "simpler" the business is, the easier it is to price and the easier it is to write pricing software for, to the point that there is no point in writing it at all. Maybe this is where Granny gets a look in: leave her alone with a simple book and a calculator that multiplies a sum insured by a fixed premium rate and she could do it blindfold. Except that little of the business is that simple and this is where the pricing software needs to be smart and the person using it needs to be an expert. Diverse classes need diverse pricing parameters so a generic, all classes point-and-push pricing engine is some way off, if it is at all possible. Many of the major packages (covered in Part 1) have their own pricing tools built-in and Rateabase (from Citrus) is one of the few stand-alone pricing systems now available. Many underwriters, however, have developed their own pricing systems based, for example, on Excel/Lotus or ResQ models and so have tailor-made the tools to suite their own particular underwriting practices.

In the fields of Exposure and Risk analysis, again the major packages and the stand-alone information systems can, and do, provide the means of analysing exposures that, provided the data has been captured and coded with some clarity, should satisfy the underwriter. Mapping systems, such as those available through companies like ISL or even the more rudimentary systems like those provided within Excel, can display territorially analysed data in a quite dramatic fashion. One major broking firm, several years ago now, had a system that could simulate a flood, treating the viewer to a chilling preview of one of global warming's less desirable symptoms, while their major areas of exposure were highlighted, and swamped. Cecil B. Demille would have been proud but that level of virtual reality is perhaps a touch overdone for most underwriters.

Risk analysis is a growing field. Most companies are able to use historical statistics to derive a view of risk using similar techniques that they would employ for exposure analysis. Another, complimentary way of doing it is to purchase access to private or public data for the purposes of assessing risk against known factors. For example, databases have long been available that detail chemical spillage, water abstraction points, road accident rates, police crime statistics, and so on. The problems were always too much data, unsure quality, patchy coverage, inadequate access indexes, no cross-referencing and no convenient delivery mechanism. This is beginning to change: databases are now being cleaned, compressed, cross-referenced and packaged on CD-ROM by some third-parties (and may well also become available via the internet) and with relatively simple enquiry tools to specify, say, a location and a radius, details of all associated risk factors can found. The value of this for certain types of cover is self-evident.

What is still a little way off is the sort of comprehensive "integrated underwriting desktop" concept that a foreign exchange dealer 100 yards away would take for granted – i.e. to look at the risk being presented, look at past performance statistics, at market conditions, at exposure, assess the risk, etc. in a single application suite to gauge the proposition and derive a price. It may well be that the final solution to appear will be a pick 'n mix approach that technologies like IBM's business objects concept will allow - multiple software vendors providing the basic Java elements to the desktop suite with the systems integrators gluing them all together in a comprehensive package. What the stand-alone system will always lack, though, is the link to the outside world. Out there are the risk databases, repositories, market information databases, the millions of gigabytes of historical data held by the bureaux, and so on. No underwriter is an island.

Underwriters have been using e-commerce for years: the LPC and LPSO have conducted millions of transactions a year with the market using their own networks since the late eighties. The General insurance world needs, and has now begun to implement the consumer-facing web-sites, on-line quotation systems, jolly graphics and secure credit card systems that they need to deal directly with consumers. The London Market however, being more parochial, has quite different business-to-business requirements and as the business class spread is very broad it is consequently hard to quantify in generic terms like mandatory workflow practices and fixed data-elements. Most players in the market gave Electronic Placing the thumbs-down some while back, despite the end-to-end benefits it could bring to the business and since then, apart from some e-mail initiatives, EDI for technical and settlement processing reigns the unchallenged king. Quite rightly so; it's a dull batch file transfer system, completely bereft of any form of gloss and it works just fine. Its success, however, is in large part due to the underlying business practices which, in the bureau markets, is very clearly defined - prescriptive business practices are always much easier to support with e-commerce techniques than those with a more open architecture, especially those which involve any degree of negotiation. It has to be presumed that no amount of technology could ever supplant a talented broker placing a complicated risk with a skilled and experienced lead underwriter; ditto their counterparts in the claims departments. Supplant; no. Support; yes. This is they key: e-commerce will need to support the underwriting process, provide the information delivery, invoke the analysis tools, access the market databases and create the data-backbone upon which the automated technical and financial accounting procedures will follow. But it must still retain the flexibility and, where required, the face-to-face negotiation.

What EDI has never done is exchange large quantities of textual or multimedia data and it has always been the lack of cost-effective network bandwidth holding this back. This is where the repositories come in to hold the data centrally and allow access to it on-line à la internet? It's early days for the repositories but the future certainly looks good.

With all the e-commerce dot.com hype being rammed down our collective necks of late, it would be surprising to find the major providers holding back in this area. Many of the software suppliers to the London Market are now able to offer web-enabled versions of their software, or will do shortly, (Room, Intech, Eurobase and CSC) and operations such as Terra Nova/Octavian, Catex and Enterprise are all working on various e-commerce facilities. Web-enabling existing applications allows them to be operated via an internet/intranet connection through a web browser and this has major benefits in the rollout of the software. This does not necessarily equate to e-commerce though.

Without a dictionary definition of e-commerce that more than two people on the planet could agree on, defining the e-commerce requirements of London Market is a tad tricky. On thing is certain though: market initiatives are notoriously slow in coming to agreement and internet technology is moving very, very fast. The underlying feeling is that there is no real consensus within the market as to what e-commerce needs to deliver. This leaves software providers with a problem because someone is going to have to invent a new business practice (or seriously rework an existing one) in order to sell e-software - out there is a major opportunity but for the problem that the business users are not sure what to use it for. Until they can decide, the software companies can not write code to support it. This is not the way it usually works and, being terribly speculative here, it might be that the big push into e-commerce will not come from the established vendors reacting to user-demand but, perhaps, from a relative newcomer or even a company from a different market altogether, like banking or manufacturing, both of which have relatively mature e-commerce systems in place.

Back to the underwriting room now, where we find an underwriter buried under a ton of paperwork and e-mails. Enter the workflow and document management systems. Variously known as Image Processing, IDM and Workflow systems, this is not particularly new concept but a number of factors have made it more accessible of late: the falling cost of data storage, the increase in network speed and the arrival of the market data repositories. The concept is straightforward: any document, slips, wordings, endorsements, advices, closings, you name it, be they scanned from paper, faxed, word-processed, etc. can be accessible from a central document database, duly indexed in every way to relate it to quotes, policies, claims, settlements, etc. as appropriate. As documents can be shared by everyone, no file contention will ever occur and no documents will ever be lost or become defaced. The workflow systems will pass the document from person to person as it wends its way towards the final resting-place secure on a CDROM and a backup tape. This, and a lot more besides, is what they do. Stand-alone Workflow/IDM systems are now available from Filenet, OpenCo and Documentum with specialist consultancy and integration expertise coming also from Linkhand and August.

A final word on Artificial Intelligence and Expert Systems that are beginning to be used mainly in reinsurance recovery systems and services, such as those from ReClaim, Rebus (Elgar), Insurance Market Solutions (Excelsior). This is an area which one would imagine would find more applications in the underwriting process, especially if allied to actuarial modelling systems, pricing systems, etc. Although there are some Internet services springing up which profess to use such techniques for weather trends, etc. little else seems to be available at present.

So, back to the original question: can Granny do the job for you while you sip Martinis by the pool? Not a hope of it. Software to help the expert is becoming much more widely available but to imagine that it could actually replace the expert remains a dream or a nightmare, depending upon your point of view. Neither Granny nor the computer knows the first thing about underwriting and I doubt that either will have the time or the capacity to learn it in our lifetimes. As it happens, there's a good deal more chance of your cat learning to do the job passably than your computer - it was only last Friday night that I finally convinced my PC to recognise the year 2000 !

By Jeff Ward, Director of TriSystems Limited

-   Insurance Technology