“Data science”, “big data” and “data analytics” currently cause a profound technological change in our society and economy. Over the next few years, both data science and artificial intelligence will also lead to changes in the insurance industry as part of the ongoing digitalisation process.
Several years ago, digitalisation was already identified as a success factor for future growth and competitiveness in the insurance industry. Consequently, new investment budgets were set up and existing ones increased. Forecasts show that in Europe alone the insurance industry’s investment in digitalisation will reach EUR 11 billion by 2020.[i] In view of the fact that European insurers are creating new boardroom responsibilities with a dedicated focus on digitalisation, this forecast may even appear too conservative.
Each and every day, insurance companies collect a large amount of data. The application of data analytics to this data forms the basis for manifold improvements along the customer journey. Processes, for example, can be designed more quickly and efficiently and customers receive tailor-made offers. Enriching this data with additional external data allows even more accurate predictions. In addition, the ongoing digitalisation will generate new customer contact points. These two effects, an improved and more efficient customer journey on the one hand and new customer contact points on the other, provide insurance companies with a high potential for profit growth—so long as the new opportunities are actually made use of. Detailed information about the customer is an important starting point to achieve this potential. Surveys show that customers are quite willing to share their personal information, provided that this is for their benefit.[ii]
The key to success is therefore not just the availability of data, but also ensuring that this data is of adequate quality and can be quickly and efficiently processed and interpreted in a targeted manner. This is precisely what data science can achieve.
So what exactly is data science and in what context do we approach this topic? Data science is the use of data analysis tools, such as machine learning, data mining and business intelligence methods. These methods are applied to extract knowledge from large amounts of stored data and to gain new insights. Such a massive amount of data is also often referred to as big data. Big data not only consists of standard company data and sensor data from various biometric and telematics-based data sources, but also of metadata and personal data. Data science is not to be confused with artificial intelligence; it is rather a basic building block or an incubator for artificial intelligence.
Research results reveal that currently only 0.5% of available data is evaluated and analysed. Due to digitalisation, the amount of available and produced data—especially personal data and metadata—is growing exponentially, therefore, this percentage will decrease even further.
To be able to use and process this amount of data, which doubles every two years, conventional methods and techniques are no longer sufficient. New technologies are required to efficiently store both structured and unstructured data and to enable their rapid availability. In addition, an adequate data quality management system as well as self-learning and self-improving algorithms have to be implemented. Machine-learning models are usually implemented ready-to-use in open libraries, therefore the main part of the work required in the data science process focuses on data selection, data preparation and data editing: up to 80% of the time is spent on these activities. Another reason for this high percentage is that the available data generally fails to meet the necessary quality criteria and is not stored in the structure required by these models. Analyses and models based on low quality data would lead to falsified results. In order to subsequently visualise the model results, further time-consuming data editing steps are often necessary, which in sum explains the considerable time required for this activity.
Another important step towards the successful transformation to a data-driven insurer is to continuously scrutinise all innovations with regard to customer advantages and benefits. It is essential, for instance, that new processes are introduced from the customer’s perspective. While this prerequisite is generally relevant for most innovations in the insurance market, the customer perspective should play a particularly important role in the field of data science. If customers fail to recognise or understand the potential benefit, they will not only avoid the new offers, but also refrain from sharing their personal data, which is, however, absolutely indispensable to the continuous development and improvement of the models.
We will address these upcoming challenges in our “Data-driven Insurance” series of six articles, in which we discuss the implications of data science for insurance companies with regard to the following key questions:
- How does data science influence an insurance company’s processes and customer journeys?
- What possibilities and opportunities does data science offer for marketing and sales?
- How can data science be used in risk management?
- To what extent does the existing IT infrastructure need to be adapted to enable the use of data science?
The answers to these key questions should best prepare insurance companies and insurance intermediaries for the opportunities and benefits that arise from the new data availability. In addition, we point out ways to significantly increase customer benefit through successful implementation. Our focus is on identifying use cases and quantifying the corresponding benefits. The next article in this series centres around the impact of data science on customer journeys and business processes.
[i] zeb.research: public statements made by insurance companies
[ii] Fujitsu: Fujitsu European Financial Service Survey, 2016: interviews with 1005 insurance customers