Making Sense of Big Data

Introduction

Even if you’ve never heard of Big Data before, it’s affecting your daily life in some way. If you checked Facebook this morning and saw ads, those were chosen to target you based on your internet browsing history. If you use an iphone or any other smartphone, the program that corrects your typos bases these corrections on the words you use frequently. If you used Waze to get here, the route it provided was based on traffic information obtained from other users, such as their location.

But that sort of thing is not just for Internet-age companies. It also has applications in other industries – including insurance.

So today, I’ll talk about what Big Data is; the state of analytics in Philippines; how Big Data could be useful to insurers; quick ways to harness the power of Big Data; and what the industry could look like.

What is Big Data?

What is Big Data? It’s the use of technology to gather and analyze large amounts of information to look for patterns and generate insights that businesses can act on to improve their bottom line.

What makes it different from standard data analysis is that it deals with datasets that can’t be handled by normal tools, such as Excel. We can look at it in terms of three V’s: volume, variety, and velocity.

The first V is Volume. In 2012, 2.5 billion gigabytes of data were created every day. Sources of this data are everywhere: cellphones in our pockets, computers in our homes and offices, RFID tags – some as small as a grain of rice – in merchandise being sold to us in stores. Even the cards we use to access our offices or hotel rooms have RFIDs. During every interaction with one of these devices, data is captured and transmitted for analysis.

This data comes in many forms, which brings us to our next V, Variety. Data isn’t just figures in a spreadsheet. Handwritten accident or medical reports, phone conversations with customer service representatives – even qualitative input can be stored and analyzed. For example, documents can be scanned using Optical Character Recognition technology so that they can be searched using a computer. Phone calls can be transcribed into text that can be mined for data, such as the frequency of certain key words.

Lastly, the third V that characterizes Big Data is Velocity. Not only is there a lot of data in multiple forms, it is constantly created every time we interact with devices.

In other countries, insurers are becoming increasingly aware of the competitive advantages afforded to them by Big Data. Based on a survey conducted by PricewaterhouseCoopers in 2014, 70% of insurance participants reported that Big Data changed their decision-making. In a 2012 study by IBM, 74% of insurers reported a competitive advantage because of Big Data, a notable increase from only 35% just 2 years earlier.

The state of analytics in the Philippines

In the Philippine non-life insurance industry, there’s a lot of untapped potential with regard to harnessing the power of data.

As of April 2013, there were 31 licensed actuaries in the Philippines. 24 of them worked directly for life insurance companies, and most of the others worked as consultants. The numbers have increased since then, but there still aren’t that many actuaries working for non-life insurers.

Additionally, during meetings with other national reinsurers from the region, I observed that our industry data is published around a year later than theirs. They also gather more information: one notable example is Japan. I’ll share examples from there and from other countries in the next part of my talk, how Big Data can be useful to insurers.

How Big Data can be useful to insurers

There are several possible applications for Big Data, including marketing and sales, but today, I’ll focus on examples related to the NRCP’s experience as a Reinsurer: Claims, Pricing and Reserving.

Claims

Based on statistics from trade associations, claims fraud adds $77 to the average policy in the UK and $167 to the average policy in the Netherlands. Undetected fraud in the UK was estimated to be worth $3.24 trillion in all.

So, how can an insurer use Big Data to improve their claims systems and their bottom line?

In Japan, the General Insurance Association established a fraud prevention office to collect and analyze data regarding fraudulent claims. In 2014, it introduced systems to 1) share information among insurers and 2) to visualize connections among fraudulent claimants and detect organizations repeatedly involved in false claims.

Similar efforts have taken place in the US. Consultants helped companies look at internal and external data, such as historical claims and fraud patterns, accidents, connections on social networks, and medical and criminal records; previously, up to 75% of this data was unstructured and could not be analyzed by older systems. Analyzing this data helped companies prioritize suspicious claims, making the process more efficient. One SAS client reported that the system paid for itself in less than five months. It also helped companies spot things they may not have seen otherwise. According to an SAS survey, 57% of insurers using analytics increased the amount of fraud they detected year-on-year increase by more than 4%; only 16% of insurers that didn’t use analytics obtained similar results.

Based on another study by the Boston Consulting Group, one North American insurer used a similar system to improve their fraudulent-claims detection by around 30%, reducing their auto claims payouts by 2–3%.

Pricing and Reserving

Aside from improving claims systems, Big Data can also help underwriters lower their loss ratios, while also mitigating the risk of insolvency. This is particularly important because under-pricing and under-reserving are among the things likeliest to cause an insurer’s failure, according to a study by Standard and Poor’s. For example:

• In 2002-2005, Mutual Risk Management Ltd, Trenwick Group, and several other international insurers and reinsurers failed, mainly because of insufficient reserves for casualty lines following a period of industry-wide underpricing, compounded by weak risk management.
• In 2008, AIG almost collapsed due to growth that was too rapid, and in some cases, outside its areas of expertise. This was exacerbated by governance and enterprise risk management issues. According to S&P: “in our view, AIG found it difficult to develop an appropriate valuation approach for certain parts of its portfolios, and this had implications for reserving and capitalization.”
• In New Zealand, some insurers did not adequately model areas that weren’t considered earthquake-prone; thus, they were unable to anticipate the impact of the 2010-2011 earthquakes, and were unable to meet their obligations because of inadequate reinsurance and their portfolio’s geographic concentration. Western Pacific Insurance was placed into liquidation, and AMI Insurance had to be bailed out.

So how can insurers avoid this sort of thing using Big Data?

The General Insurance Rating Organization of Japan, or GIROJ, collects data from its members, and analyzes this to provide them with advice, including advisory premium rates that underwriters can incorporate into calculations to ensure that the final rate accurately reflects the level of risk.

Because of the tools at its disposal, GIROJ can customize the benchmark rates for each prefecture. For instance, areas in Kyushu are prone to typhoon damage, while Hokkaido and other seaside prefectures are vulnerable to heavy snow; this information is factored into its recommendations. Its members can also avail themselves of catastrophe models to help them assess their total risk exposure and factor it into their reserving decisions.

In terms of automobile accidents, GIROJ’s analysis showed that a lot of incidents involved elderly drivers. So, it worked with the government to disseminate safety advice to them. It also introduced rating tables/systems that help insurers account for age when adjusting premiums. Similar tables exist for assessing clients with a history of accidents.

By giving underwriters more information for their decisions, Big Data can help increase a company’s profitability through improved loss ratios in the short- to mid-term, and improved stability in the long-term.

How to build Big Data capability quickly

Harnessing Big Data is not without challenges. The volume and variety of data and the velocity at which it is obtained and needs to be processed requires investment in tools, business processes, and people. Furthermore, it entails risks such as data security.

A quick and easy way for an insurer to build its Big Data capabilities – and with less risk — would be to work with a reinsurer, who deals with a lot of data owing to the nature of its business. While the insurer hires and trains staff, it can partner with a reinsurer to run analyses using the reinsurer’s tools and business processes, as a “try before you buy” strategy towards building Big Data capability.

In one case study from AIR worldwide, a reinsurer worked with its cedants to examine the valuation of the properties in their portfolio. The results helped both parties make better reinsurance decisions and improve their overall risk management. The company with the best valuation processes reduced their reinsurance premiums by about 10%, while companies with poor valuation processes increased their premiums by up to 15%, because of their improved risk assessment.

On our part, NRCP has hired, and is continuing to hire, licensed actuaries and other people with the necessary skills needed to think in terms of Big Data, including people with backgrounds in mathematics, statistics, and economics. We’ve also acquired licenses for a number of programs that we share with our partners, especially those interested in exploring the possibilities of Big Data.

For example, to assess their total risk exposure in the event of an earthquake in Makati or a super typhoon in Manila, an insurer could partner with NRCP to run catastrophe analytics using RMS’s Philippine earthquake module and AIR Worldwide’s typhoon model.

To improve its enterprise risk management, insurers could work with NRCP on reinsurance optimization and capital adequacy analyses using URS Risk Explorer, and on reserving analytics using URS RES-solver and Anaplan.

This “try before you buy” strategy could help insurers determine if these tools are something they would want to use in the future, without having to spend $600,000 a year in licensing costs.

The future

Using such tools is something we can do in the short to medium term. Where could things be headed in the long term?

Faster and more comprehensive data collection within the industry is something worth considering, and also something NRCP can help with because of the Big Data capacities we are currently building. More industry statistics at a faster rate – for example, estimates of losses due to claims fraud – could benefit everyone’s decision-making. Information sharing among insurers along the lines of GIROJ model is also worth considering; imagine being able to cross-reference claims reports with a country-wide database.

In an article called “Insurance at a Tipping Point”, Jamie Yoder from PricewaterhouseCoopers postulated that, within the next few years, there could be a shift in the philosophy underlying underwriting.

The increasing use of sensors and connected devices offers more real-time and predictive data, which has the potential to change the emphasis from “what has happened” to “what could happen”. This could open up opportunities for insurers to gravitate from reactive claims payer to preventative risk adviser.

Using Japan as an example again, the general insurers’ association uses data on automobile insurance claims to determine the most accident-prone areas, which are indicated in a map that’s updated annually. Something like this could even be taken a step further; for example, insurers who had access to such data could make recommendations to the government on safety measures, such as better signs or lighting for dangerous roads.

In conclusion, Big Data has a lot of potential benefits for insurers. With the way the world is changing, data will become even more important in the coming years, and if we want to remain competitive with other countries, we have to begin exploiting this potential.