• <noscript id="ikikm"></noscript>
  • ? Myrzik & Jarisch

    Consulting for enhanced data analytics and pricing processes

    Advanced solutions for pricing sophistication, portfolio steering and smart pricing

    In this fast-changing environment, we act as your strong partner. Our experts combine their experience in primary insurance with their knowledge in data science and advanced analytics in order to offer support from A to Z throughout sound and modern pricing techniques.?

    Pricing sophistication

    Digitalization plays a tremendous role?in industry.?In comparison to only a few years ago, a multitude of?structured and unstructured?data is available. This brings the need for enhancements in infrastructure and computational power to store and handle this amount of data as well as sophisticated and modern algorithms for their analysis.?

    In insurance, the challenge is to be strong in pricing sophistication in order to stay competitive. Proven and established actuarial models need to be combined with modern machine learning models?to exploit the power of big data.

    By supporting in modelling process and deriving a technical price, we help to create a strong competitive advantage by pricing risks adequately. In turn, this reduces anti-selection in the portfolio and increases profitability.?

    Based on historic portfolio data, risk models enable a proper prediction of the expected claims cost by considering the influence of relevant rating factors. These can be identified using generalized linear modelling (GLM) and other machine learning techniques, a multivariate approach considered the de facto standard in primary insurance’s technical excellence.??

    Deployed into business, this forms the basis for strategic ratemaking decisions: it enables the identification of mis-priced customer segments, as well as yielding an improved risk selection through deeper understanding of the historical risk.?

    Retention rate as well as conversion rate models are gaining more and more importance in the insurance sector. On a technical level, demand models predict a policyholder’s estimated probability to renew or to write a policy. Deployed into business, the main area of application is to explore and measure the impact of price changes. This ultimately leads to effective scenario testing and potentially to price optimization approaches. Furthermore, the model results can be used as a basis for strategic marketing decisions and customer lifetime value propositions.??

    Although generalized linear models still represent a well consolidated technique, we have taken advantage of tree-based machine-learning algorithms (XG Boost and Random Forest) to improve the estimation of the customer elasticity.?

    The predictive risk cost models are a basic and fundamental part in the ratemaking process, but they cannot be considered exhaustive. The behavior of the customers at the point of quotation or renewal should be taken into account for a profitable growth.?

    Making use of advanced cost models and demand models we help first in identifying the position of the company in the expected projected profitability and retention space. Based on the client’s strategy, we then propose a set of pricing scenarios aimed at meeting the company’s target, simulating the renewal process and optimizing the cap and floor structure (soft optimization). Should the market allow us, we can run the optimization at policy level (hard optimization) thus identifying the optimum premium per each customer which maximizes the target function (volume or profit) under some?constraints.?

    Portfolio analysis & portfolio steering

    Relying on a sophisticated pricing process is one of the most important factors for a successful insurance company. Nevertheless, a key factor and starting point for all actuarial modelling is a deep understanding of the company’s portfolio. The descriptive analysis of the company’s historic data and a constant monitoring of actual numbers and KPI forms the backbone to stay competitive in the market.??
    To have sound understanding of the status of the portfolio today, it is crucial to investigate the historic data of an insurance company. We help to identify, structure and clean the data necessary for this step. We combine information of single policies with the clients’ claims data base. Based on the combined data set, the first step is to visualize insurance-specific KPIs in a univariate view: we show trends and differences between the modalities of the provided rating factors or variables. This exercise can be done on aggregated historic data as well as separated by calendar,?underwriting?or financial year to show development of certain trends over time.

    Measurements to improve the pricing process are important. However, their impact on the daily business needs to be ensured and reviewed constantly. A sound monitoring system helps make sure that initiatives in the pricing process have the impact desired. Equally important, such a system ensures continuous control over all important KPIs, reflecting the health of the portfolio.?

    We help in establishing monitoring systems on several levels. We are able to provide primary insurance companies with a fully-fledged monitoring dashboard customized to their needs. Equally, our consultants can assist to enhance existing monitoring systems and conceptualize add-ons accordingly.?

    Smart pricing solutions

    The digitalization process has caused a deluge of? structured and unstructured data, which can and should be used to enhance the underwriting and pricing process.?

    Working closely with our clients we are continuously improving the set of tools, procedures and techniques to stay ahead of the curve. Additionally, we liaise with out unique data-hunting units, which allow us to retain our competitive advantage to other players, and enable solutions that would otherwise remain untapped.

    Clients can leverage on a tailored geographical categorization of? the risk carried in their portfolios. Geo-spatial smoothing?is based on?the assumption?that after adjusting for?all?other risk factors, the risk of policies in adjacent regions is more similar than the risk of distanced policies. Therefore,?insured risks?that are?physically?located close to each other should get similar?technical rates, provided?all other risk factors are?held constant.???

    This technique?can be used to detect spatial patterns?and?model deficiencies. When applied?in pricing. It can?additionally?be used to?improve geographical risk segmentation.??

    The vehicle-related data is among the most important risk drivers in the motor business. The vehicle is usually described by many variables e.g. vehicle age, fuel type, cubic capacity but ideally just a distilled rating factor could carry most of the relevant information. We help reduce the complexity of the database, identify the independent vehicle-related variables and find the more significant vehicle groups to increase the model predictiveness.?

    One of the tasks of technical departments is often that of offering novel approaches to marketing campaigns or, more generally, to identify unique selling points for clusters of customers. Cross and up-sell opportunities fall under the same broad umbrella term of “segmentation” The goal of these analyses is to find the next best product to propose to the final customer.??

    Another, more traditional, approach is that of moving from the known recency-frequency-monetary (RFM) segmentation of the customer base, to a multidimensional one that uses a bottom-up approach which finds structures (or clusters) in the wealth of unlabeled data. This approach leverages on unsupervised machine learning techniques.?

    Finally, if previous information on the behavior of customers confronted with offers or campaigns is available (in other words, if labelled data are available), a top-down approach is also possible. This supervised approach can identify the drivers of positive decisions, as well as calculate the probability of positive outcomes for each policyholder for similar, future campaigns.?

    Technical sophistication is also driven by the identification of new rating factors. A novel, promising approach is to develop an isolation index. This “remoteness” score is based on the actual travel times between the written risks of a portfolio, and a selection of points of interests (POI). The set of POI is fully customizable, once the risks have been geo-located accurately. This analysis fosters the development of new, tailor-made rating factors.?
    Contact our experts
    Massimo Cavadini
    Massimo Cavadini
    Head of Section
    We use cookies on our websites to improve your experience as an internet user, and to optimise our online services. They comprise cookies that are required for technical purposes, and without which the website functionality could not be guaranteed. We also employ cookies to carry out statistical evaluations of the reach of our websites. These evaluations are anonymised. You can find further information on the cookies we use, and ways to object to the use of cookies for statistical evaluations, in our cookie guidelines.