Insurance operations are complex with a data management burden that is significantly higher than many other industries. The breadth and variety of risks in a Specialty portfolio combined with the requirements to create multiple perspective views for time development, business segmentation and the valuation of assets and liabilities, drives the need for a holistic portfolio data strategy.
In terms of systems and tooling, customisable off the shelf underwriting and policy admin systems only go so far and miss the mark in supporting the breadth of portfolio management activities.
Although a commonly heard statement, the claim that the Insurance industry does not invest in technology is not necessarily true. From experience, it has been the case that the industry has been poorly serviced by the expensive software and consulting options that exist in the market.
A Specialty insurer will have an extensive array of special purpose tools that handle the complexity of the portfolio model. In an attempt to deliver a consolidated portfolio view, there will be a host of manual processes, spreadsheets, bespoke systems and IT integration solutions that collect this information and get the data to where it needs to be.
With so many data components and demands for alternative views on performance and risk information, both the burden and expectations of data professionals are high. The complexity of insurance is continually underestimated by new entrants and seasoned veterans alike, and unfortunately, the industry has a habit of learning from the same mistakes.
As this series of articles develops, they will cover each of the mentioned areas in more detail and discuss the methodologies, tools and technologies that verge360 is employing to address some of these challenges.
The management team at verge360are combining our years of on-the-ground practitioner expertise to deliver solutions and consulting services to support a data driven portfolio strategy. By applying our accumulated knowledge, utilising operational engineering techniques and employing the right data technology choices, verge360 is offering a solution that re-engineers the management of the portfolio for existing and future demands.
Theverge360 framework lowers the cost of operations and maintenance while extending the life of legacy solutions through abstracted data integration. Through targeted data management driven by automated workflow, and a flexible underlying insurance data model, the verge360 architecture releases employees from data movement and manipulation tasks to focus on value-add activity. And by embedding the business perspective model in the core of the portfolio management strategy, the verge360 approach delivers the right perspective views of data, in the correct format and shape wherever, and whenever, it is required.
An exercise in balance and appropriateness
For the modern Specialty insurer and reinsurer, portfolio management is a broad concept with wide operational scope. The concept of the portfolio is one that is used throughout the business model. From the basics of managing underwriting, policy administration and operational accounting right through to risk assessment, capital management, executive reporting and regulatory filings.
As a business function, portfolio management is traditionally limited to underwriting performance, exposure management and portfolio optimisation activities. However, the portfolio itself lies at the heart of every insurer, informing and supporting the multitude of decisions made across all functions of the business.
The portfolio is a foundational component in the development of the business plan, and a critical yardstick for measuring the current success at, and future likelihood of, achieving that plan. More importantly, it is a tool that should ensure that the business model can respond and adapt to changes in conditions, and identify where the underlying assumptions of the plan no longer hold true.
Experience suggests that many insurers struggle to develop a platform upon which they can holistically consolidate and manage their entire portfolio. Fewer still, can use it as an effective tool to consistently inform every day decision making across all functions of the business.
What makes up an insurance portfolio?
The portfolio logically connects to almost every part of the insurance business model. As a container of information, the portfolio stores, collates and reports upon the assets and liabilities of the insurance organisation.
From the front-office, the portfolio captures information about each risk underwritten, the premium due and collected, the losses expected and incurred and the costs associated with acquiring business.
From the back-office, it collects the expenses of running the business, the assumptions upon risk and their impact on operating performance, the instruments of risk transfer and the assets and income of investment strategies.
A simplistic view of the portfolio data flow
In many organisations, the norm consists of a fragmented system of portfolio views, with data collected, manipulated and transformed to represent singular business perspectives. This is often the result of deploying specific solutions for a targeted purpose and isolated development of business units.
In this environment, datasets are generated to address a specific problem; standardisation with other business units and functions is limited; perspectives are constrained to snapshots of time movement; and transformations are applied that do not easily reconcile with the alternative views of the portfolio used throughout the business.
There are notable exceptions, several Bermudian property reinsurers have developed maturity in managing their portfolio of exposures incorporating a rolling consolidated portfolio view into their underwriting and pricing platforms that are used to support risk management and reporting activities.
These platforms can be relatively expensive to maintain, and offer a singular view of underwriting risk limited to catastrophic events captured by third-party vendor models. Although applicable to property reinsurance portfolios, it is not a model well suited to a modern global insurer, where capital is distributed amongst multiple markets with product lines and geographic coverages that are not covered by these models.
Elsewhere, internal models, and specifically underlying capital models, address a more consistent and wider view of the portfolio. These models, which have become more common under the Solvency II regime, offer a capital focused view of risk based on both historical snapshots and forward-looking projections of the portfolio.
Unfortunately, the utility of capital models has been limited due to poor data accessibility; a mechanical necessity to aggregate information; the speed and frequency of execution of the models; and a lack of common understanding of what lies beneath the numbers. The capability to interrogate the data is restricted by cumbersome user interfaces, and is reliant upon expensive key-employees to unwind the complexity of the model to provide the answers that are needed.
When it comes to using these answers to make critical decisions, the information provided is often stale, and seldom comes with adequate health warnings regarding the assumptions and quality of the underlying data.
Designed and implemented well, a portfolio data strategy provides a grand unifying view that represents the complete business model of an insurance operation.
Why does having a holistic portfolio view matter?
Effective management of an insurance portfolio is an exercise in balance. On one hand, it is necessary to provide a product that is attractive and affordable to the consumer. While on the other, ensuring that enough premium is collected to deliver an adequate return to investors.
As part of that balancing act, it is important to develop an underwritten portfolio that is well diversified, so that a concentration of risk does not overly expose the insurer when the wind blows, the ground shakes or financial markets tumble.
Controlling loss through appropriate risk selection and efficient deployment and protection of capital are all key to delivering satisfactory returns. As is deploying investment strategies that maximise investment income, but do so without significantly contributing volatility to the risk profile.
With all these levers, assumptions and inter-linked dependencies, a holistic portfolio data strategy is an essential tool to defining and effectively navigating the path to profitability.
Plan B – what plan B?
The business model of insurance is one that is predicated on assuming risk. As such, the management of that risk is of paramount importance to delivering optimal profitability.
The activities of business planning and portfolio optimisation are a series of forecasts, guestimates and assumptions on what likelihood of those risks developing could be, the best options to protect against and mitigate those risks, and how to develop a portfolio that generates a return on those risks.
The reality of executing against even the best laid plans, is that changes in the environment do occur, and decisions need to be executed in short order to effectively respond to these changes. Without an evolving view of the portfolio, these decisions are made purely on instinct and good judgement, using subjective observation of the developing risks or opportunities.
To make informed decisions, the details are all important when measuring the impact of both macro and micro-environmental changes to the risk-reward environment. Be it market rate movements, higher than expected loss development, or deteriorating business retention. A change in conditions, or the wrong decision, may not only impact the individual portfolio segment, but could also significantly impact the profitability of other business lines and the effectiveness of risk transfer and mitigation strategies. The lack of available insight increases the likelihood of capital inefficiency and missed opportunity, ultimately leading to sub-optimal performance of the organisation.
To support proactive decision-making, a good portfolio data strategy needs to be able to define where an organisation was yesterday, is currently today and could be, in a range of possible tomorrows. It supports segmentation and decomposition at an appropriate granularity to answer the questions the business may have of it, and it will do so consistently and reliably for the many consumers of its information. Addressing the complexity of good portfolio modelling is all about being able to serve up the right perspective, at the right time, and for the appropriate level of detail.
Having the information to react to changing situations before they become a problem is clearly beneficial, but being prepared with alternative options and strategies ahead of time is priceless.
Maintaining a wide landscape with alternative perspective
Each function of the business relies on some view of the portfolio to manage day-to-day tasks and produce the reports, metrics and artefacts that quantify risk, identify opportunity, and support decision making.
Many portfolio data solutions are likely to exist within an organisation, be it a policy administration solution, a financial ledger, a contract data warehouse and dashboard, built-in application reporting, a repository of spreadsheets or a data analyst with a SQL prompt.
It is rare however, to find an organisation that has easy and repeatable access to all of the necessary data in the format, shape and timeliness required.
Operationally, the fragmentation in data extraction, migration and translation leads to unnecessary costs, with highly skilled individuals spending their time on low value tasks that involve manually moving, processing and reconciling data. It has been a common experience to see these time-intensive tasks periodically repeated with only slight variation, re-working or producing data that either already exists in electronic format in the organisation.
From a governance perspective, the management and control of manual data flows introduces additional overhead in an already inefficient and costly workflow. Furthermore, the flexibility of controls and processes, often lead to a breakdown in both control mechanisms and the confidence in the information.
The operational expense behind the activity of portfolio management is driven by this inefficiency. Shadow IT organisations are prolific in the industry, with experienced Excel and SQL users propping up day-to-day operations and generating non-reusable legacy.
Due to the constrained perspectives and snapshot nature of the information produced, it is rarely used for other purposes beyond that for which it was intended.
A good portfolio data strategy can replicate and automate the repetitive elements of these activities while supporting alternative perspective views and variations in analysis. The aim of information management is to deliver data where it needs to be, at the right time, in the expected format, with as much hands-free processing as possible.
Uncertainty - What’s in a number?
The complexity of the risks in an insurance portfolio, and the hazards to which they are exposed, introduces a vast amount of uncertainty.
Especially where the portfolio contains high degrees of volatility, such as a catastrophe exposed book of business.
A typical Specialty insurer employs a high number of skilled and experienced underwriters and actuaries to build models, or purchases expensive third party platforms from industry specialists, to make judgement calls on each risk in the portfolio. The more the variety in the risk profile, the more specialised and complicated these models become.
No model is perfect, with the data underlying, and being fed into it, even less so. Whether you view it is an art or a science, the constructs of the portfolio are a compromise of imperfection built on assumptions, with varying degrees of data and statistical quality that are full of uncertainty. The balancing act is finding the correct compromise that instils confidence and supports business activity, yet won’t constrain and cripple the value or availability of the portfolio through an unnecessary pursuit of precision.
Uncertainty is not a weakness. In insurance understanding and managing uncertainty is a strength that provides a competitive edge in the market.
The confidence, or uncertainty, in a number is itself an important measure, identifying where the decision maker needs to spend their time analysing the potential risk or opportunity presented. Reliance on a single or restricted set of metrics, without understanding the quality of both the risk and the applicability of the models used, can lead to poor decisions and surprises in the portfolio.
That being said, turning an underwriting or portfolio management dashboard into something that resembles a 747 cockpit is counter-productive. Augmenting hard metrics with visual indicators to highlight uncertainty, exceptions and confidence issues, is an important facet to add additional information without crowding the key statistics upon which the business operates.
A good portfolio data strategy can not only operate upon assumption, uncertainty, and handle imperfection, for a given perspective view it can quantify and qualify the uncertainty behind the metrics.
It’s not just a simple matter of timing
Despite the cost, delays and process impact of Solvency II, it has led to a noticeable improvement in the maturity in which the market assesses portfolio risk.
Regulation has driven a somewhat standardised view of risk and, in many cases, more frequent and detailed generation of the complete portfolio. The emphasis on stressing the assumptions supporting a portfolio view, and identifying risk development with a forward-looking assessment, has strengthened the understanding of underlying potential risk in the portfolio.
Unfortunately for many firms, even though this view of the portfolio has been projected forward, the vintage of the data supporting this view is often several months old by the time the numbers are presented. Informing decisions only after thresholds have been breached, risks have developed and/or opportunities have come and gone.
Where supported by a portfolio consisting of available forward-looking data, geared to the mechanics and workflow of the organisation, a dynamic portfolio data strategy can support a prospective view of risks and opportunities before they fully develop. By incorporating the identified stress conditions and operational thresholds defined during the business planning and risk assessment process, an adaptive portfolio can highlight where assumptions no longer hold true and identify when alternative options may be available.
Of vintage and managing adjustments
Lloyds of London is a great example of a traditional insurance market that has managed to drive product innovation over the centuries. The nature and flexibility of the syndicated model allows for an eye-watering array of insurance coverage with exposure profiles from the mundane to the novel and highly bespoke.
Lloyds has a long history of re-invention and expansion, from moving out of the coffee shop and developing a global distribution network, to introducing electronic exchanges for automated placement and claims processing, and more recently the Refresh and TOM initiatives.
It is a market that transforms by the consensus of its constituent members, and as such, transforms slowly, with compromise and varying degrees of success. As a result of maintaining this flexibility, while preserving some of the traditions, it is a market that also suffers from inefficiency and high operating costs.
One particular challenge is the inherent lag and latency in the provision of contract, exposure and claims information as risks are quoted, contracts underwritten, premium collected or claims raised and settled. Before any electronic exchange of data takes place, manual entry and re-entry of data is a common bottleneck in the information flow. This is driven by the bespoke nature of the risks underwritten, the involvement of intermediaries and agents, and the lack of standardisation across the platform.
When it comes to defining a portfolio, only having part of the picture at any given time poses a particular problem. Do you manually adjust and accrue a position or roll-forward an old one? How do you manage data when it catches up, and what you reported yesterday is different from what you reported today? The vintage problem is not just a question of “how old is the data”, but also keeping track of what changes as business process catches up with reality.
A good portfolio data strategy must not only allow for reproduction of prior positions, but also support abstraction and analysis of the adjustments made, so that it is possible to easily reconcile these differences and explain any incongruence between views.
Forward-looking projection and time dimensionality
The development of a forward-looking position focuses on the projected changes to the constituent parts of the portfolio, adjusted over one or multiple time dimensions.
Such transformations include premium payments and earning development, claims development and reserves, technical provisions, valuations of assets and liabilities, debt aging and seasonal capital consumption.
Each function of the business has a different perspective view of time development, adding to the complexity of building a singular portfolio data strategy that can support each possible time dimension and development methodology.
Regardless of function, the forward projection and allocation of each valuation and time development methodology should originate from a single starting position, a base portfolio of underlying data. Oft quoted as a “single source of the truth”, even this base position will need to support multiple dimensional views to provide positions for a given “as-at” or time series positions, segmented by business and risk profile, as well as requiring additional data to support the valuation or projection methodologies employed.
Tracking the complexity of the time dimension needs to factor in not only the current position of data, but also project over multiple time horizons. For example, viewing a 12-month view vs. a development to ultimate position, managing the multi-year elements of underwriting cycles and co-ordinating overlapping mixed-duration assets and liabilities.
In most cases specific tools will have been developed, or acquired, to perform the necessary transformation and generate a dataset representing a particular perspective. These derived programmatic datasets will be generated and stored in a range of mechanisms application databases - such as the reserving system or financial ledger, - via transformation processes as part of a data warehouse, or within highly complex and inter-linked spreadsheets.
Unfortunately, many of these data solutions will have been built to serve a singular purpose and will generate a serial snapshot view of data. It is more than likely that views will have been aggregated to a lower resolution of data than is required to re-use this information for other purposes. By these design constraints, the lack of broader accessibility, and an inability to re-purpose and integrate the information with other business perspectives, lessens the value this information could have across the business.
A good portfolio data strategy will not only serve up the base portfolio data, but also provide a perspective model for the transformations applied to this portfolio data that can be consumed across the business model.
To extract the value to the business, this transformed data needs to form part of a wider data landscape, with feed-back and integration to other data sources, and enriched with additional data perspectives. To manage the dimensionality and alternative views based on projection methodology, it is necessary to build a portfolio strategy that can incorporate the perspective view and handle different granular views of data.
Why stop at your own data?
During the start-up and initial growth phases of an insurance organisation, data collection is generally focused on capturing the base information to support policy administration and financial reporting. As the organisation matures, this view of the portfolio develops to capture information required to support risk management and portfolio optimisation activities. However, the focus remains on capturing data for prescribed purposes, such as financial reporting, pricing and risk management, and is limited to underwritten policies and experienced claims.
As part the quotation process, substantial amounts of information is processed as submissions are investigated and quoted. Regardless of whether the policy is actually underwritten, effort is expended and data is collected through-out the underwriting process.
Often judged as not being important to managing the current underwritten portfolio, these data mementos are subsequently discarded. While the data may not have been standardised, cleansed or validated, even as a dirty source of information it has merit to inform and benchmark the current portfolio, and help identify future market opportunity.
This data documents a world of risk outside the underwritten portfolio, providing insight into why business was not acquired or retained, and developing a repository of intelligence into potential target risks. Even where potential business falls outside current risk appetite, as rates harden or internal appetites change, it may become attractive business in the future.
A good portfolio data strategy extends to the world outside the current book of business, capturing information to deliver insights into other opportunities and profile the success and failures of the underwriting channel.
Not all data of importance relates to the terms of the contract, the characteristics of the insured exposure or the loss details of the claim. Information regarding the market and the organisation relationship within it, is available from the external interactions of the organisation with the outside world. This information carries significant importance when determining future development of conditions, or a changing view of risk not evident by just looking at historic values.
This data can be critical to help observe developing market trends, highlighting possible undocumented change in risk profile, or deteriorating relationships with intermediaries and clients. All key elements to achieving a target business plan, that often get left to subjective and informal tracking without substantive metrics or alignment to actual performance indicators.
Customer Relationship Management, “CRM”, tools are now becoming a staple investment for several insurers as they aim to close this information gap. That being said, the implementation of these solutions focus heavily on the front office functions such as sales channel and claims management. Other than simple pipeline projections, this data is rarely integrated into the business planning and portfolio management process.
These soft indicators provide insight to forecast and explain changes in relationship and reputational driven metrics such business conversion/retention rates, as well as capturing anecdotal information to validate and improve market assumptions, such as predicted rate movements.
A good portfolio data strategy utilises this wealth of information to support re-positioning of the prospective view of portfolio, and deliver additional insights into the future performance and risk assumption.
Open data initiatives by governments and organisations, and an increase in third-party data service providers, has created a new wealth of data that can be applied to the underlying exposures of an insurance portfolio.
From easing the burden of data collection, formatting and cleansing to augmenting and enhancing the risk profile of the portfolio, the increase of data availability is a clear benefit to the industry.
The opportunity to use machine learning to highlight the unknown risk drivers, and further classify risks in the portfolio has historically been held back by the incomplete and inconsistent methods of data collection when underwriting risk and processing claims.
Understandably, an insurer wants to minimise the number of details a customer needs to enter when developing a quote to reduce friction in the acquisition of business. The resulting impact is that data points are collected for known drivers of risk to support pricing. To make use of machine learning, this needs to expand to indicators that have not yet been identified.
As the resurgence of artificial intelligence gathers pace and technology platforms become more accessible by small-to-medium enterprises, the requirements on collecting data points will increase to support data mining activities. With pressures to improve customer experience, alternative sources, such as open and third-party data services will need to fill this gap.
As a source of data, the rise of internet connected devices, telematics, drones and social media also herald a new age of insurance capability with a mass of data available to support profiling of risk. Whether or not privacy issues will restrict the scope of the information that can be collected, or that any real meaningful insight can be garnered for anything other than niche demographics, is still yet to be seen.
One thing is certain though, the industry currently struggles to manage the data it has been collecting to-date. Once these initiatives move from a position of research and development to requiring full operationalised production support, they threaten to drown existing data teams and IT departments.
A good portfolio data strategy should provide an inexpensive method for augmenting the trusted and authoritative view of portfolio data with third-party data, and incorporate machine learning and data mining capabilities to allow for generation of an enhanced view of the portfolio.
Paul Owen heads up Analytics and Technology and is a Co-Founder of verge 360.
To find out how verge360 can work with you to address these challenges in your organization please contact us on