Gaining value from data requires analysis in the midst of uncertainty, which complicates the quest for insight. Dr Michael Kelly of CSC explores the world of “known knowns” and “unknown unknowns” in data and shares how insurers can use data to become more competitive.
We want to advance our business, but we are often uncertain what questions and what data will best get us there. We need some information just to operate effectively. We also need to derive insights that will enable us to compete more effectively, and perhaps to take bold actions to leap ahead.
The “unknown” continuum
Our problem deriving these insights is depicted in the following Figure 1. Most insurance companies are knowledge poor yet data rich.
To move forward, we do not just need to consider new means for processing the data; rather, we need to use the techniques appropriately to ask the right questions, and to help us ask, and answer, the right questions.
Asking the right questions is not so straightforward, though, since we often do not even know what we do not know.
Looking at “known knowns” and “unknown unknowns”
We made initial gains by looking at “known knowns”, and simply depicting data better, and by using statistical methods to assess that data. We drew patterns from the data, even if by depicting it more visually and allowing users to draw the inferences.
We have been moving to find “known unknowns”. In this case we know the questions to ask – we know what we do not know and we attempt to find the answers. Sometimes the questions may not be relevant to what we are trying to solve, but they are directed by what we think is relevant. Techniques such as machine learning and data mining have helped us to find deeper insights than was possible from just statistical analysis.
To really make use of the all the data we have, we need to step beyond what we definitely know, and start looking at the “unknown unknowns”. We need to take this step for many reasons, including:
• Too many variables to draw conclusions;
• Sheer quantity of data;
• Inability to assess data fast enough.
Although some companies think this level is needed only for handling what is being labelled as digital insurance, and not for current insights; they fail to grasp the importance of “unknown unknowns”.
For instance, companies have already started using these techniques to go through the many variables that could affect claims for house insurance – they were able to determine that factors (such as number of bathrooms) that were not considered previously were actually most important for predicting future claims. This information could then be fed back to actuaries to enable them to define the product to account for this variable.
Certainly as insurance companies move to increased use of digital insurance, and our customers adopt these channels, we need to derive approaches that require greater individual insights. Techniques associated with defining “unknown unknowns” will enable us to customise our products, offers, market, and service approaches to the “individual”.
Driving insight in insurance companies
So how are insurance companies using data for different purposes, and what are some ways we are using data to become more competitive?
Let’s look at the options under the headings of “operate”, “operate more efficiently and effectively”, and “change the game”. (And let’s not forget that the basis for all of this insight is a properly constructed logical data warehouse, as I discussed in my last article in Asia Insurance Review
April 2016 issue. http://goo.gl/Im8AUa
Management in all companies need to understand how well the different activities of their organisation function compared to established goals. Historically they have used printed reports periodically generated, and much of the analysis has been left to the reader. Managers have used their experience to compare more current data against past data and the desired targets, then draw conclusions.
Fortunately, visualisation tools have been changing the way this data is presented to managers so that they can more readily discern patterns and key results.
Tools such as Qlik, Tableau, and PowerBI have changed the way that users can view data, even enabling them to drill down into areas and filter the data they see. These tools have also allowed the display of data combinations as performance dashboards, displaying data in the form of KPIs, according to the user’s role.
However, the tools used to present the data up to the visualisation layer are generally still the tried and true spreadsheets (eg, MS Excel) and sophisticated statistical and mathematical modelling tools such as SAS and SPSS. Thus we can see that to operate business as usual, companies need to deal primarily with the “known known”.
In addition to generating internal operations reports, insurance companies need to organise data to report to external entities, especially regulators. Generally, the data for these reports comes from multiple internal systems, including policy admin and financial systems, combined with processed data related to risk and re-insurance. Generating these reports can be lengthy, and can involve substantial processing of data pulled from the different systems.
Operate more efficiently and effectively
Of course most insurance companies want to grow and provide better services at the same or lower cost.
To achieve this end requires asking serious questions to determine how to achieve this growth and efficiency. Moving to “known unknown” analysis provides an information base to decide steps to take, or at least to try to see their effect.
This involves finding answers to questions such as:
• “What customer groups are purchasing what products?”
• “What products are most likely to be purchased by customers who have bought another product?”
• “What agents should I notify to target specific customer groups with the information derived from such analysis?”
• “Based on historical analysis how can we identify potential fraud to save money?”
Digital insurance in particular requires obtaining and acting on considerably more data than traditional insurance. For instance, to sell to or service directly with customers via internet, mobile, or social media, the company should track the interaction journeys, and sometimes automatically vary the interactions based on responses.
To do so requires a consideration of user options during the interaction, then tracking the user experiences, assessing them, and providing certain responses during the interaction. Thus we see the need for rapidly finding answers and automatically taking actions based on the information obtained.
Insurance companies need to be more proactive with even their internal data, as well as in combining their internal and external data from customers.
Assessing agents and brokers becomes possible using predictive analytics, machine learning and tools such as Watson to drive new insights.
Likewise, companies can assess their claims data to find out more about the potential for fraud, or can evaluate their use of reinsurance and claims risk to determine how to improve their asset use.
Change the game
Some insurers are using analysis to significantly improve their competitive positions. Using the previously described techniques provide a solid foundation, and exploring the “unknown unknowns” provides the additional insight to change the game.
Tools to derive this insight include link analysis and MapReduce on top of the machine learning – and of course specialists such as data scientists working with business people to use the tools effectively and use both predictive and prescriptive analytics.
For the wealth of data we can obtain from wearables, usage based auto insurance, and other Internet of Things (IoT) offerings, we have no choice but to use more advanced analytics. How else can we find what is important in the volumes of data generated by these devices? We need machines to help us sift through the data and identify those factors that may be important, and then explore how they may be important.
But do not think that analytics for “unknown unknowns” is only for the IoT – insight into products, distributors, claims, risk, and asset utilisation may all be obtained by using new techniques even on available data.
What’s your move?
So, how are you doing on dealing with the data deluge? Where do you find yourself in the chart on Figure 2? Are you in danger of sinking? Are you simply using reports to maintain your current situation in the face of increasing competition? Are you using analytics to obtain better performance – to improve your competitive position or your business value?
Or are you starting to use analytics tools to give insight even when you do not know the right questions to ask – to create excitement and change the game to drive ahead of your competitors?
Dr Michael Kelly is a Consulting Partner with CSC, based in Singapore.