Big Data: The practical considerations for the road ahead - Q&A with Geraldine Gallagher:
Geraldine is DWF LLP’s Head of Business Intelligence, and is therefore an expert on the practical and technical challenges that businesses can face when attempting to extract and analyse data. We spoke to her about the particular problems that insurance brokers may have to overcome to adapt to the likely impact of IA2015, and to keep or gain a competitive edge in the marketplace.
How can businesses use a data pool to establish trends and patterns?
The key technical tools are predictive analytics and data-led risk scoring. If possible, benchmarking of data against other data sets is also a very powerful way of drawing conclusions from a particular data set. Organisations would need a deep and wide data pool and to get this, it might be necessary to create a historical data pool using prior years’ policies. The potential difficulty is that the data now required may not have been captured at the time. That problem could be overcome by:
- Backfilling the data – unfortunately, this is very labour intensive and not always possible.
- 'Fixing Forward' (creating the data pool from current and future policies) – the downside is that the ability to establish trends, patterns and therefore analytics will be delayed for a significant period of time, and good foresight is needed to make sure all of the right data gets captured from the outset.
- Collaboration with other organisations to share data and create a common and open access data pool – here, the downside would be the difficulties of maintaining data security and potential risk of losing competitive advantage.
On this last point, if organisations find it commercially impractical to create an open access data pool, then it might be that the only way for organisations to gain large amounts of data quickly would be by acquiring other organisations. This in turn raises the prospect that brokers with well-structured data collection practices and a good pool of placing data might be a better acquisition target than a broker with a similar client base but less good data.
Is technology enough on its own?
Technology such as predictive analytics will never be enough on its own to gain meaningful insights from the data. It is essential to apply specialist industry (and sometimes legal) knowledge to the data analytics to identify the factors which have the most impact on potential outcomes. An experienced broker and possibly even an insurance lawyer would need to work with the data specialist to be able to demonstrate what factors really make a difference.
Also, there is a requirement for continuous testing and adjustments of analytical assumptions. This is essential to negate the risk of circular, self-fulfilling outcomes and to aid early identification of changing risk patterns in the market to enable counter-strategy design and deployment. In other words, specialists in insurance broking and in data analytics will always need to think carefully about the trends emerging from the data, and make adjustments if necessary, to prevent the data from becoming meaningless.
How easy is it to work with data that has come from disparate sources?
Sometimes it is far from easy! To take the example of two sets of data coming together when two organisations merge, the two data sets could use different naming conventions and data architecture, and there could well be different levels of granularity of data collection and storage. There could also be differences in how much of the data is structured and how much is unstructured. ‘Big data’ tends to be unstructured, so if it is essential to work with it, there is a need for additional technical tools for 'data scraping', and/or intensive manual input to get the data into a format that can be analysed.
Another impact of different data sets being drawn together is that the data cannot simply be used from source. The data has to be cleansed, converted and maintained into a consistent and robust data warehouse.
What skills and resources are needed to make sense of the sort of data collected by insurance brokers?
The best place to start is with an analyst team which is not only technically skilled in structuring, extracting and delivering data, but which can 'read', interpret and answer the commercial and practical 'so what' questions which arise out of the analytics. The team also needs the skill and experience to be able to identify which parts of the raw data have the potential to offer the greatest insight. That helps to minimise the cost involved in designing a data schema.
This information is intended as a general discussion surrounding the topics covered and is for guidance purposes only. It does not constitute legal advice and should not be regarded as a substitute for taking legal advice. DWF is not responsible for any activity undertaken based on this information.