Often there is a lack of a uniform data view

Dr Stephanie Friedrich has focused on insurance fraud for years. In this issue of .experience, she spoke with us about the changes Big Data and analytics are bringing to the industry.

How are data and analytics changing the insurance market?

As in all other industries, data and data-based analyses are playing an increasingly important role in the insurance industry.

Processes in the companies are optimised and automated by analytical procedures. This makes insurers more efficient and faster. However, these optimisations also allow for more differentiated considerations, e.g. in the context of customer potential. However, all these steps are not possible without a sufficient data basis and the corresponding analytical know-how.

Insurtechs in particular are taking advantage of the opportunities offered by data-driven business models and are therefore, in my view, an important driver of this transformation. In my view, the established insurers are observing these trends very closely and are reacting by taking up these topics and intensively driving them forward in their companies.

What are the most visible trends today?

The changes for the end consumer are particularly evident in the area of claims processing. This area had long led a shadowy existence, but most companies have initiated automation processes in recent years. The starting point and basis has usually been automated fraud pattern recognition based on a large amount of data available in the claim. While the first systems tended to be rule-based, ML processes are now also being used, particularly in larger companies. In addition, unstructured data is increasingly being included.

A more recent trend is above all process optimisation within claims handling. It is no longer only conspicuous losses that are recognised, but in particular also those that are identified as inconspicuous and suitable for quick settlement on the basis of statistical models. The use of expert reports is also partly based on this.

In addition to claims settlement, other processes are also becoming the focus of data-based decisions. In particular, external data can be used in underwriting and pricing and represent a competitive advantage. In addition, customer value, up- and cross-selling, and lapse models are used to better manage customers in customer management.

Is the situation different when you compare the German and Swiss insurance markets?

Certainly, there are some specifics when comparing the two markets. First of all, it is noticeable that the Swiss market is significantly less fragmented than the German market. Therefore, prices in the German market seem to me to be subject to even more significant competition, so that the pressure to optimise processes should actually be higher. Interestingly, this is not necessarily reflected in the activities. On the contrary, I see at least as much willingness to invest among Swissinsurers, especially in claims management but also in the area of data and analytics. And in the area of BVM, Swiss companies are, in my view, much more professional and consistent.

In my experience, there is a greater tradition in Germany of including external risk data in underwriting. In my view, this is not yet as widespread in Switzerland.

What are the big questions coming from customers today regarding the use of their data?

I think it is important above all to deal transparently with the use of data and to communicate the benefits clearly. A customer definitely has a clear advantage if damage can be repaired more quickly because the relevant data is available and the algorithms declare the damage to be inconspicuous within seconds. An insurance company that consistently optimises processes will also be able to offer lower premiums in the long term, which is in the interest of consumers.

Insurers have a large amount of sensitive information about their customers’ marital status, household sizes, property, occupation, net income, education. How do you deal with the issue of data security?

In my experience, the issue of data security and data protection is a particularly important one for insurance companies, whose business model is, after all, the security of their customers. In case of doubt, even if it were legally possible, it would be preferable not to use the data than even run the risk of coming into conflict with data protection or data security.

The topic is therefore relevant in almost all projects, and in practice we very often also involve our lawyers.

What do you think are the biggest opportunities with Big Data in insurance?

It is difficult to name a single topic. I think that there are still many options, particularly in the area of pricing, but also in customer management and even claims management, to manage more efficiently and in a more risk-adequate manner. However, our customers often lack a uniform view of data, since insurance companies have a lot of data at their disposal, but it is not necessarily possible to access it centrally in the sense of Big Data. Therefore, the prerequisites for comprehensive analysis and use of the available data often have to be created first.

And what about the challenges? Do you agree, for example, with the analysts from Gartner who say that it’s not the technologies, but the people?

I wouldn’t say that for every company. The technologies are available on the market, of course, but insurers often still work with very old legacy systems that often stand in the way of using modern applications. So there has to be a transformation on the technology side as well, which is happening in a lot of companies right now. Of course, you also have to take the users with you, and that is certainly a challenge in some cases, but from our experience it can be mastered quite successfully. It is often advisable to take smaller steps without losing sight of the goal.

What should companies look for when designing their Big Data strategy? What is Experian’s approach in this area?

First of all, the goals should be clearly formulated. Big Data is not a strategy in itself. It must be clear in advance what you want to achieve with Big Data.

Then, in our view, it is essential to build the strategy on 4 pillars:

• Data
• Analytics
• Software
• People

So we need the necessary data for the goal and based on this, the appropriate analytical models. Modern ML methods are not always the means of choice; sometimes it is also of particular interest to be able to transparently understand why a decision was made. A BVMexpert, for example, needs a clue as to why a loss is conspicuous in order to be able to continue working. A score alone will not help him. A very important point is to be able to implement the results of the analysis operationally, with the right decision engine. Decisions often have to be made in real-time, so it doesn’t help if the model takes 1 hour to calculate. And then the employees in all areas must also recognise and support the opportunities in the project. Only then can projects be implemented successfully.