Machine Learning in Insurtech – a fuzzy and probably approximate science!
We have looked previously at the huge opportunities provided by the rise of Insurtech within the insurance industry ([i]). Insurance companies have been notoriously slow to embrace the lucrative changes provided by the rise of technology, with 74% of leading worldwide insurers agreeing that the sector as a whole has been lagging behind in the digital revolution.
While Insurtech is wide and far reaching, one particularly poignant area is in the rise of Big Data and the opportunities that speedy interpretation of this can provide.
A new concept?
Of course, collating, organising and interpreting data to get accurate measurements of risk is by no means a new science. In fact, some 17th Century Underwriters specialised in this field, analysing statistics to quote insurance premiums for merchants and tradespeople of 1690s London.
To quote Agile Risk Partners MD, @James Poole’s blog ‘The Big Data Paradox’ (
Data is no more important now, than it was then. However, what the rise of technology has done to information, is create more and offer the potential for almost instant analysis. Statistics that took weeks to study can now be done in seconds, thanks to the rise of machine learning and Artificial Intelligence. Data scientists are now able to program computers to capture data from a vast range of different sources; from meteorological conditions, to geographical and historical data; as well as person specific data captured from social media or the Internet of Things – and instantly interpret this information. This can lead to swift conclusions for multiple areas of insurance; such as risk analysis, risk reduction, insurance product pricing, detecting fraud and offering targeted marketing.
Swift and accurate (approximately!)
The speed of these processes is vital to the advancement of the industry. Being able to analyse risk and give up-to-date responses in real time, offering instant and accurate results is vital – both for staying ahead of competitors, and for catering to the “instant” world of the millennials. “Time is money” as they say, and that is certainly true in this case, as quick analysis of Big Data allows companies to save money by pricing policies correctly, reducing pay-outs for fraudulent claims and being pro-active in reducing accidents before they occur.
Of course, the computerised system cannot entirely delete risk. In fact, the systems use what is known as “fuzzy logic”; the term “fuzzy” not being one usually associated with a field as mathematical as computing. It accepts that the binary 0 and 1 model cannot be applied to risk, that by its nature is uncertain – with few “definite” true or false values. Fuzzy logic produces machine learning known as PAC (Probably Approximately Correct) learning, which – as the name suggests – results in data that is as close to accurate as is possible within the scope of the information.
A hunt for new talent
In spite of the PAC learning, the margin of error is far slimmer than human error. The experts in this field - data scientists - are critical in the continual push for better and more efficient algorithms, able to produce faster and more accurate data analysis. When top worldwide insurance executives were asked about the most difficult obstacles they face in their push for innovation, it was “talent” that topped the list, at 87%.[ii] The highly sought after nature of this industry (particularly the niche career path of a specialism in both underwriting and data science), makes this an ideal profession for entry level IT specialists with a desire for a shooting career trajectory.
Specialist recruitment agencies such as Aston Charles will be vital in the search for and recruitment of talented data scientists with a specialism in underwriting. The wide range of candidates available, from new graduates to experienced industry professionals, will give insurance companies the opportunity to employ some of the best data science minds to help move their company forwards in the era of machine learning.
Can we beat 17th Century Underwriters?
The big question is, can data science of the 21st Century beat the innovative approach to data analysis of the underwriters of 17th Century London? The answer is yes; as long as machine learning is fully embraced by the industry – with a constant impetus to invest and move forwards. While the insurance sector lags behind other industries and doesn’t push forwards to reach new digital heights, the industry is actually falling behind their 17th century counterparts, who were pioneers of the then-advanced use of statistics.
But is it possible to get ahead now, at this stage in the digital revolution? Innovative Insurtech start-ups would say “Yes.” Agile Risk Partners has been developing a machine learning-based data-capture and risk-pricing platform for the industry. Data science company, Mango Solutions’ () second sprint on the build of the system has produced a mind blowing 200,000 risk-locations in 10 seconds. Insurtech companies such as this can herald massive advancement for the industry, in terms of their reputation for technological innovation.
It’s time for the insurance industry to put down their ink pens and lead the machine learning race.
[i] Willis Towers Watson. 2017. Insurers under pressure to go digital – Willis Towers Watson survey - Willis Towers Watson. [ONLINE] Available at: . [Accessed 11 June 2019].
[ii] Raconteur. 2018. Rise of the data scientist in insurance - Raconteur. [ONLINE] Available at: . [Accessed 11 June 2019].