27 November 2009

Published in: Software

Searching for the unattainable?

Solvency II not only demands masses of extra reporting from insurers to be produced in much shorter time periods, it also expects this to be much more rigorously controlled and documented. How are actuarial business processes and technology responding?

European insurers are starting to look in earnest at the systems they need to cope with the new risk-sensitive world. They're being pushed to revamp the way capital is assessed and managed not just by the prospective requirements of Solvency II but also by the realization that strengthening their balance sheets in response to the financial crisis and making the best use possible of available capital makes sound business sense and can used to their competitive advantage.

But with the deadline for implementation of Solvency II now less than three years away, time is running short for insurers to upgrade their systems. A few weeks ago, IBM launched what it calls "a showcase" for Solvency II solutions at the IBM Industry Solutions Center in La Gaude, near Nice in France. There the company replicates "real customer environments," allowing you to visualize and experience insurance business solutions as though they were already in place in your own workplace. The Solvency II showcase for life insurers includes Watson Wyatt's VIPitech as its risk engine, that for non-life insurers EMB's Igloo.

The site has already been visited by a number of insurance companies and, says Walter van Benthem, insurance industry leader, IBM South West Europe, "It has been a wake-up call for many insurance companies that didn't realise there was so much to be done." He also stresses that companies are starting to realise that Solvency II "is an opportunity and a key differentiator in the market -- a new way of also acting "smarter" as the environment for insurance companies is becoming more interconnected, instrumented and intelligent."

Gavin Conn of Aegon UK on the Aegon Group's in-house model: "It's a common calculation engine that can be linked in to any liability modelling system" and "it will be our key tool for the market-value margin calculation for Solvency II purposes." 

Till now, the actuarial function has been at the heart of insurers' risk management and capital planning systems. Solvency II gives actuaries a much wider role - article 47 of the directive brings their influence even into the underwriting policy of the firm - and this will require them to become much more involved with strategic business decisions -- a major culture change for both actuaries and non-actuaries and a stiff challenge for the whole company.

There are also countless new requirements to be met so that, on top of enhancing models and handling far greater volumes of data, insurers will also have to document and control much more tightly all their capital calculations under Solvency II. An internal model will have to satisfy tests relating to its use at all levels of the company, statistical quality, documentation, calibration, validation and profit and loss attribution. The use test alone will be very demanding, requiring insurers to show that the model is widely used in day-to-day management decisions and plays an important role in risk management and capital assessment and allocation.

Among these regulatory demands, Gavin Conn, capital modelling manager at Aegon UK, highlighted two key challenges.

• Documentation of the models - "The documentation standards of Solvency II look very demanding. Documenting the model in sufficient detail is going to be crucial in order to achieve internal model approval."
• The move to real-time risk reporting - "outputting results as quickly as possible to enable continuous risk monitoring."

At the practical level, these demands will mean major changes to business processes and well as the technology that supports them. "Until now," notes Paul Hopkins, principal at Watson Wyatt, "actuaries have got away with doing their own thing. But because of Solvency II this is turning into a business process rather than just a bunch of technicians coming up with numbers."

Right now, it's hard to say what stage the industry has got to. The range seems to go from those that are still doing their gap analysis to those that have already decided on their needs and selected the systems to support them.

Upgrade or start afresh?

A lot are in the middle, either waiting for others to move first or perhaps trying to make up their minds between upgrading their existing systems, supplementing their actuarial projection systems with additional analytics, or starting afresh. Martin Sher, managing director of SolveXia says, "Many companies are waiting to see what other people are doing on Solvency II. But I don't see many, if any, clients ready to throw out their existing systems. There's more of an incremental approach."

And because of this attitude among the clients, and despite the aging technology and architecture of most mainstream actuarial projection systems in use today, no vendor is seeing sufficient demand to bring out a completely new actuarial modelling system. "It's a chicken-and-egg situation," notes Sher, "but, in time, it is inevitable that, as with all areas of technology, a new offering, possibly from a new vendor, will emerge that moves the market forwards. This is likely to happen after 2012 as the Solvency II methodology and associated modelling practices become standardized and can be commoditized."

Given cost controls at many companies (and this article concentrates on UK life companies), what is clear is that enhancing an existing system is probably an easier option for companies to accept than adopting a completely new approach.

Underlying business processes

But technology can only be used effectively if the underlying business processes are right. "There's far too much use of spreadsheets and manual intervention," points out Mark Chaplin, global head of risk and value services at Watson Wyatt. This means more potential for errors, the possibility of files getting corrupted, the difficulty of controlling access to the files, the limited audit trails for changes and other drawbacks. Overall, "the control environment is not as strong as it needs to be for Solvency II," adds Chaplin.

"The underlying business processes are very fragmented," agrees Bruce Porteous, who leads the Solvency II programme at Standard Life. "There are lots of manual hand-offs and it's very inefficient. We need to go back almost to the drawing board and redesign our processes, then fit the infrastructure to those processes."

Pearl "is considering a complete rewrite of all [its] systems and processes," according to Darren Wathall, actuarial systems development manager, who with Paul Tonkin, head of IT architecture and security, leads the infrastructure workstream of the actuarial systems transformation project at the firm. The Pearl Group owns a number of companies, major one being Pearl Assurance and Phoenix Life. The fragmentation of processes at Pearl has been exacerbated by the group's string of acquisitions of other companies. Pearl, says Wathall, has ended up with "something like 23 [Towers Perrin] MoSes models and 19 [SunGard] Prophet workspaces that contain about 3,000 products (700 independent)."

Walter van Benthem of IBM: "The actuarial and IT departments will have to work together more closely."  

Aegon also has several systems running in its operations. "Our main valuation platform in Edinburgh is VIPitech," says Conn. "Over in Dublin we use Prophet and also in some areas of the business we use [OAC's] Mo.net." He adds: "MoSes is the main valuation platform in the rest of the Aegon group. Across the whole group, we tick every model and platform there is."

The shortened reporting times demanded under Solvency II also mean less room for manual operations. "To achieve reliable results within the required timescales, you need more automated systems" says Chaplin.

These issues are part of the wider challenge of data quality, on which Solvency II documentation lays heavy stress. "The sooner you can get rid of all the spreadsheets and all kinds of errors in them the better," states van Benthem. A huge effort will be needed to retrieve and normalize the required data and then test it, document it and make sure it's auditable. IBM, says van Benthem, estimates that this task typically accounts for about 65% of Solvency II implementation. Not surprisingly, IBM offers specific assets and tools to address these problems, notably Cognos for reporting and risk analysis, InfoSphere and Insurance Information Warehouse, which consolidates, collects and centralizes information from various sources, systems and mainframes.

Quarterly reporting

With outdated practices and inadequate technology, many companies will have to put in a major effort to comply with the quarterly reporting presaged by Solvency II. Says Wathall at Pearl, "The problem, at the moment, is that the group has a huge variety of models, systems and processes that, essentially, are carrying out the same requirements across different blocks of business."

At Standard Life, Porteous says the present individual capital assessment (ICA) process takes a few months to run and for quarterly reporting, "we need to get the models and processes running much faster." The plan is to run the internal model in earnest, what he calls "the full Monty", once a year at year-end, a process that should take a few weeks, and then run a slightly cut-down version quarterly, which would still be very robust because the information will be publicly disclosed. Standard Life also has an existing monthly process, which is further cut down from the full run and which the company is planning to develop further.

Doing anything more frequently than this probably wouldn't make sense because it wouldn't necessarily add value. "The balance sheet isn't really changing on a daily basis," notes Porteous. But if a big transaction came up, for example, "we would have to do it intra-month, of course."

Pearl, after closing the Glasgow office this year and Peterborough next year, will have 120 staff in Birmingham producing the reports. "That's for half-yearly reporting," points out Wathall. "One of our problems is if we want to move to quarterly reporting, effectively we'll have to more than double the team because you have to have team A to do quarters 1 and 3 and team B doing quarters 2 and 4 because it's not possible to run all four with the same team - they'll just burn out."

Again, the problem is manual intervention. "One of the reasons why we've got so many staff involved in producing the reports," explains Wathall, "is they're having to type the same assumptions into many different assumption tables over and over again." So an obvious priority for Pearl is "to pull the assumptions out into an assumptions system - either in Excel, SAS or elsewhere - and have a single point of entry." Similarly, staff are having to pull numbers from a variety of models in many different ways to create the reports, so another obvious priority is the creation of standard reports from a central data warehouse.

Another UK company is also looking at ways to reduce the time needed to prepare quarterly reports. To get all the results from all the different business units for the full economic capital model at present takes just under three months from start to finish. One reason is that the team which runs the models has so many different reporting requirements - not only economic capital but also market-consistent embedded value, pillar 2 and European embedded value.

The ORSA and use test

Increased emphasis on risk governance in the directive also requires changes in work methods and systems, even for those UK companies which think their experience of the UK's ICAS regime leaves them with little to do to conform to Solvency II.

The ORSA [own risk and solvency assessment], notes Porteous, is "a kind of souped-up version of the FCR [financial condition report] and business planning process, but I wouldn't underestimate the amount of change we're going to have to go through on it. Projecting a market-consistent balance sheet and then subjecting it to stress I think is technically quite difficult. The models are going to need to be enhanced. Integrating the FCR and the planning models together again is going to involve some work."

The order in which you do things is also important. "The ORSA needs to be given a front seat in the strategic and planning part of our processes so that it genuinely drives the way the business is run," believes Porteous. "At the moment, I'm not sure that for some firms risk assessment genuinely drives the plan."

Ejsmond-Frey of Microsoft: "Terabytes of data moving from an insurer to a hosted site is by no means a trivial activity. The technical design of the system becomes critical."

The use test is another part of the governance challenge for companies. Documentation and evidencing of this is will need to be done, but it's still only half the story. "The documentation is the bare minimum of what you have to do," Porteous considers. "You have to genuinely get people believing that this is the right thing to do, to be really using the internal model processes to inform decision-making and to help run the business, as second nature."

Porteous, an actuary himself, says actuaries sometimes might just look at the liability side of the balance sheet. "We're going to have to change our mindset. The internal model is about the whole balance sheet, not just a part of it. So we need this holistic view of the balance sheet, not just at any point in time but also projecting it forward."

Modelling systems

The demand for improved technology and improved performance is therefore is showing itself at several different levels. Gordon Ejsmond-Frey, director, EMEA Insurance at Microsoft says insurers are seeking to run bigger and more complex models, and models with larger data sets. "We get demand from insurers for more quality and accuracy." Typically, actuaries have built the models without reference to the IT departments and, says Ejsmond-Frey, "it's no surprise how fragile some of these models and software are." However, an actuary's retort to this view was: "Actuaries have had to build models with the involvement of IT because IT didn't understand the models."

Standard Life decided it was a good time to look at "what software is out there to see if there's anything better than what we've currently got," says Porteous. "We use basically MoSes integrated with spreadsheet applications, but we have looked at some banking packages and at some of the competition [for MoSes] in terms of life insurance modelling software."

Pearl is currently assessing the actuarial software market and has been looking at all the leading systems. A formal "request for proposal" process is planned shortly, according to Wathall. "It's been a long time since the Birmingham site purchased MoSes in 2000 and things have moved on considerably, not just in terms of the modelling tools but in terms of the overall support packages available including hosted solutions and automation."

Pearl is considering building a new suite of generic models rather than start from their existing models. "Our existing models are old and largely inherited via acquisitions and they are not an ideal place to start," notes Wathall. "We expect it to be cheaper to start again rather than try and convert an existing model into a generic model. It is also much easier to reconcile to a specification rather than existing models, although this will have to be done do some extent to satisfy our stakeholders."

Another British life company is sticking with the software it bought 10 years ago for one very telling reason: it reckons that the investment put into the platform since then, including intellectual capital, totals £30-40m. Other companies are in a similar position.

Martin Sher of SolveXia: "I don't see many, if any, clients ready to throw out their existing systems. There's more of an incremental approach."  

Aegon UK looked at alternative systems such as Prophet and MoSes two to three years ago when it decided to move from VIP to VIPitech but it stayed with Watson Wyatt. A reason was, says Conn, "the ease of conversion, given our existing code had all been developed in-house and all on VIP." But Conn adds: "We do feel that VIPitech is the right platform for future risk-based modelling requirements such as Solvency II."

The Aegon Group has also built its own in-house model that is shared with country units in the form of a "DLL" (dynamic link library). The DLL uses output from the actuarial projection systems to project future capital requirements. Explains Conn: "It's a common calculation engine that can be linked in to any liability modelling system, such as VIPitech, Prophet or MoSes. It uses a shortcut which removes the requirement to do numerous runs to obtain future projected capital and it significantly speeds up the run times." The DLL projects the capital for non-hedgeable risks and, notes Conn, "it will be our key tool for the market-value margin calculation for Solvency II purposes."

Because a lot of the systems in use have been around a long time, they are in the view of Ejsmond-Frey almost customized models for each insurer. The problem is: "The people who implemented them have often now left the firm and there's little documentation."

In September Prudential, a user of MoSes since 1999, opted to stay with the Towers Perrin system but with an added interface to Microsoft's high-performance computing (HPC) solution "to centralize its financial modelling and risk management applications for all of its UK business," according to the announcement at the time. Guy Shepherd, head of business information systems at the Pru, said that the agreement will help Prudential "harness the power of the Microsoft HPC server, take advantage of the scalability of the solution and give us the opportunity to run increasingly complex models."

While it wants to bring its systems up to date, Pearl also wants to keep things relatively simple. "Actuaries have traditionally modelled non-material features -- we would look to take modelling shortcuts while preserving the appropriate level of accuracy," says Wathall. "We end up with a much simpler model that people can understand and which they are confident to change."

Although the RFP process is starting shortly, Pearl does not expect to have "the solution bottomed out with a signed-off business case" until the end of the first quarter of 2010. Pearl will be seeking approval to use some of its existing models but much development will be required. "We will need to keep some models to fill gaps while the new world is built and we may even keep some models long-term if the actuaries are happy to do so," says Wathall. "A strategic decision that Pearl have already made is that the modelling platform will sit on top of an existing SAS data warehouse," adds Wathall. 

Darren Wathall of Pearl: "We expect it to be cheaper to start again rather than try and convert an existing model into a generic model." 

SAS is Pearl's corporate data warehouse. Its job is to collect data from different administrative systems and other sources within a single data warehouse creating "one source of the truth," Wathall explains. "The data is transformed into a standardized format to remove administrative system or ex-company specifics so that generic analysis and reporting processes can be run against the data. The data is held at policy level to allow low-level drill-down to support analysis and issue identification." he adds. The warehouse will then be used to carry out the type of reports and analyses required by Pearl.

Pearl is planning to build an assumption system, though it hasn't decided yet whether it's going to be in Excel, SAS, or something else. After speaking to people in the industry, Wathall says "the feeling is that the assumption system is a difficult area that no-one seems to have got right yet." Pearl wants an assumption system in which users can enter, copy and create assumption sets and then send to the modelling system in a controlled manner.

Pearl is considering a hosted solution, says Wathall. This is because "managing complex data centres is not something Pearl wants to be doing. We have also had problems resolving systems issues between the software vendors and our outsourced IT providers when ownership of hardware and software is separated."

Pearl is seeking advice from the software vendors on how to achieve automation. This may involve using their existing enterprise offerings, building a new wrapper or using a third-party solution. "Any third-party relationship would be introduced and managed by the vendors if we go down the hosted solution route," stresses Wathall.

For Standard Life, the key is finding the right balance between control governance, integration and flexibility. Explains Porteous: "We're going to take our existing life insurance modelling capability and move it more towards the banking approach, which is generally more controlled and integrated, but not going the full way because we think we need that flexibility to be able to go in and change our models without having to go through a vendor's change procedures. But we want to improve the controls around that and make our governance even tighter."

"We've already got this in place, to a reasonable extent, but we need to roll it out to the whole balance sheet model," Porteous believes. And he adds, "We're coming to the conclusion that spreadsheets are not necessarily bad per se. They'll always be a need for some of them provided that they're properly controlled in a robust way, and their numbers are limited."

Processing power

Because models are getting bigger and more complex, data sets are getting larger and reporting times are shrinking, there's a lot of interest among insurers in solutions that offer greater capacity and increased speed for Solvency II calculations. There are also sound business reasons for this. "Some companies see it as quite an important competitive differentiator," notes Ejsmond-Frey. "You're a lot better positioned if you can see your solvency position at the end of each day."

However, not all insurers see this as a priority. Standard Life sees a huge opportunity to reduce the time spent crunching numbers by first of all improving the businesses processes. "Processes are near the top of our list." says Porteous. "Tools like replicating portfolios are too and ramping up the hardware is also certainly on the table."

Chaplin of Watson Wyatt argues: "It's generally not computing time that's the problem, although this can be improved, it's the amount of manual interventions" before and after the runs. The internal model use test and the requirement for continuous monitoring of solvency mean that a parallel, but more nimble system is also often needed. This is where replicating portfolios or free capital formulae come in. These models, if implemented correctly, give you "a real-time view of your balance sheet and an ability to do what-if analyses," says Chaplin, whose firm Watson Wyatt offers a tool for this approach called Replica. But these approximate models do need rigorous testing because, adds Chaplin, "getting proxies that behave properly in all circumstances is not always simple."

Mark Chaplin of Watson Wyatt: "It's generally not computing time that's the problem, although this can be improved, it's the amount of manual interventions." 

Aegon is introducing a replicating portfolio tool across the whole of the Aegon Group to support risk management. "Replicating portfolios will be instrumental in enabling risk management in a market-consistent context," comments Conn. Toronto-based Algorithmics is the partner for this project. Explains Conn: "The country units create replicating portfolios of their businesses that can then be re-valued under any market conditions. The infrastructure allows for aggregating these results to any level, enabling daily views of risk and value at all levels of the organization." Aegon UK already uses Watson Wyatt's replicating portfolio tool to quickly revalue the liabilities for its with-profits business.

Running your old models on HPC is, of course, one possible approach to improving performance and perhaps deferring investment in new models. Microsoft's Ejsmond-Frey estimates that with the right set-up HPC could give you as much as 8-10 times faster results on old software, but he warns, "These won't necessarily be more robust or reliable."

Pearl is considering on-demand computing, Wathall says, although he says "on demand" is slightly misleading: "You can't ring up and say you want 1,000 servers and expect to get them immediately. They will try their best but you have to be reasonable. We'll have to plan up to 12 months in advance and probably at least three months in advance in detail."

The advantage of an on-demand solution compared to a rigid data centre is, adds Wathall, "flexibility and cost. Our current data centre only has a utilization rate of 5%. With on-demand we could ramp up the number of servers available to meet changing peak demand and then release them during our quieter periods, thereby sharing costs with other companies. Also, we would worry less about hardware refreshes and there would be minimal upfront costs."

Another potential advantage to Pearl of the on-demand approach is it's scalable. "As a closed fund, our business is running off, so we need to be able to scale our costs down, and on-demand and will allow us to do that," points out Wathall. Conversely, if Pearl acquires another company, it needs to scale up quickly. "Typically, we can need another 50% or 100% power within six months of an acquisition and it gets more difficult to do that with dedicated data centres because of physical space, power and network constraints," he says. "You also end up with many different hardware types which is not an ideal platform on which to support distributed computing."

Aggregating results

For aggregating results from their models, some companies still rely on data management platforms such SAS's or spreadsheets, the latter still popular even with those companies that possess reasonable modelling systems. Aegon UK is looking at spreadsheet management solutions to try to increase controls over spreadsheets. "We want to assess the potential benefits for automated audit control and version control and for their analytic features that highlight potential sources of errors," said Conn. Among the companies offering these solutions are Cluster Seven, Finsbury Solutions and Prodiance.

Software as a service

A number of firms have looked at SolveXia, a "software as a service" (SaaS) web-based platform that allows users to automate the "end-to-end" processing of their risk analytics. It also plays to those companies which want to approach the modernization of their systems incrementally, notes SolveXia's Sher. "It supports all mainstream actuarial projection systems, analytic tools (including Excel), data and reporting processes," he says. This is coupled with a data warehouse and risk dashboard capabilities.

But there are security concerns. Ejsmond-Frey of Microsoft cautions that "terabytes of data moving from an insurer to a hosted site is by no means a trivial activity. The technical design of the system becomes critical."

Sher agrees, noting that SolveXia has built its system from scratch incorporating strong security protocols throughout and using latest technologies. He says,"We are sourcing capacity from major, high-security data centres which have the infrastructure and experience to meet the processing and storage needs of the insurance market." He adds "while we are at forefront of the move to SaaS and cloud computing within the actuarial modelling sphere, we are really riding on the back of a general IT shift forming the battleground between Microsoft, Google, Amazon and others. He believes that over the next five to 10 years "many, if not all, companies will adopt SaaS solutions in many areas of their businesses and will look back and ask what all the fuss was about."

Hopkins of Watson Wyatt notes: "Intuitively you would think SolveXia would have more appeal for small and medium-sized companies, but larger companies might not want to risk such a new proposition, given the time it would take to resolve any problems." Also, he adds, there is a relatively large cost to set up outsourcing.

Sher of SolveXia admits that the firm is seeking its first Solvency II client to join others from outside Europe, but he says he's hopeful because he knows of three or four clients who expect to decide on their road map for the directive by the end of the year.

Actuaries and IT

However the technology develops, some in the industry believe actuarial dominance of the systems inside insurance companies will be challenged. Ejsmond-Frey is not alone in believing that all the business-critical technology used by insurers needs to be centralized within the IT department. He thinks all the firm's inputs and outputs should be integrated into the group reporting system and become part of mainstream management reporting. "Actuaries have to let go of choosing hardware and software and tinkering with it," he says. "But this will be hard because it's been going on for 25 years."

Bruce Porteous of Standard Life: "Processes are near the top of our list. Tools like replicating portfolios are too and ramping up the hardware is also certainly on the table." 

Van Benthem says life is changing for the actuaries because getting rid of all the spreadsheets and putting an overall solution in place feels like a threat to them. "You will still need the actuarial expertise, but getting the right data from the right sources and making sure that they're normalized mean the actuarial and IT departments will have to work together more closely."

At the companies, there does appear to be more interaction between the two areas but the actuaries still seem to be in control. One life company has approached the issue by making an actuary head of IT. He says it made sense because in preparing for Solvency II, "the actuaries wanted more and more things in the IT domain."

Another insurer is considering bringing IT people into the actuarial area in order to help it cope with the rigorous Solvency II demands for documentation of the model.
Standard Life has IS/IT people aligned to the actuaries who form the model development team. The purchase of any new tool, for example when the company appraised banking packages, always involves IS. But, says Porteous, "I don't see IS owning the models going forward but just a continuation of our current support set-up which has worked extremely well to date."

Second wave of investment?

Whatever the eventual structure of their systems, the time for insurance companies to get them right is running short. As 31 October 2012 approaches, one consultant believes there will be compromises on all fronts. "In order to fit into the management process, insurers will have had to take shortcuts on some technical points." There'll also have to be some give and take between companies and their regulators, he thinks.

And there'll probably have to be a second wave of technology investment in capital and risk management. "Looking out 5-10 years the more visionary insurers are realising they need a fresh look at enterprise risk management," concludes Ejsmond-Frey. "It just can't be done by taking old systems and souping them up."

 

Back to top