We live in a world of Big Data. Petabytes, Zettabytes, Yottabytes of data. The Internet of Things (IoT) increasingly connects day-to-day appliances, machines and equipment with each other. Billions of devices constantly recording data: sensors, cameras, microphones, thermostats, pressure gauges, RFID chips, attached to everything from mobile phones to industrial hardware.
Contrast this with early Business Interruption (BI) insurance in the 1860’s when obtaining reliable records to quantify losses would have been difficult, long before the advent of modern accounting and reporting standards. Most records were manually created until as recently as the 1990’s when transition from paper to computer accelerated. Even then a 3 1/2” floppy disk only stored 1.44MB and joining multiple Excel sheets was not possible.
Nowadays datasets grow rapidly, to the extent they cannot be easily managed, stored, shared, analysed or queried with previously applied methods, and require new approaches.
Despite these previously unimaginable changes, the fundamental principles of BI insurance have not changed significantly over the past 50 years. But what potential effects can increased data have on the insurance industry?
Could big data mean less losses?
GE, for example, remotely monitors vibrations, temperature, and pressure on thousands of turbines around the world 24/7/365. The data interpreted by its proprietary software Predix, allows operators improved efficiencies and enables preventative maintenance before units malfunction, mitigating both uninsured and insured downtime.
Could big data provide assistance for underwriters?
Could underwriters follow telematics car insurers’ lead and harness industrial operational data to specifically price a product for companies which provide access to this data?
Could big data mean faster settlement of claims?
Typically files provided on a large energy claim can run to hundreds of megabytes, if not gigabytes. This can include refinery Linear Programming, petrochemical production data or hourly power plant data containing detailed statistics such as: ambient temperature; availability; generation; flow rate; heat rate; fuel usage; or utilities usage. Whilst arguably not in the realms of big data, it is sufficiently large to require a forensic accountant to mine and distil down for Insurers consumption.
Enterprise Resource Planning (ERP) systems allow extraction of specifically tailored datasets of Production, Sales and Inventory to assess BI losses, reducing the time spent by the Insured manually preparing reports, which would require verification back to source records anyway.
Time can be better spent on analysis issues relevant to the quantum of loss, improving response time for updates, interim payments and settlement figures.
Could big data mean more accurate loss measurements?
Increasing volumes and granularity of data allows quantification of losses which would have been unsupportable historically, for example:
- Analysis of heat rates and fuel burn to identify turbine efficiency losses following non-catastrophic damage
- Change in product mix following a partial loss
- Identifying loss of capacity on a specific turbine
- Flaring of hydrocarbons in the aftermath of an incident at a petrochemical complex
- Correlation of availability and output to ambient temperature during an outage
Increasingly complex models can be constructed reflecting interdependencies of operational units allowing improved analysis of issues, and more accurate loss measurements.
Is there a down side to big data?
As IoT connectivity and data volumes increase, so does the potential for something to go wrong. Companies are under threat from cyber-attack, be it disruptive (e.g. data breach) to causing catastrophic damage (and resultant BI).
The need to extract, handle and analyse large volumes of data, means it is important for underwriters and claims handlers to work with forensic accountants who know the data to request, how to analyse this data, and distil the data into a user-friendly assessment, which stands up to scrutiny.