Cases 13.01.2023

Fennia Mutual Insurance Company makes extensive use of advanced analytics to enhance marketing performance

Data-driven leadership Marketing effectiveness Marketing optimization Data Science Sales modelling

Linking value and attitude-based segments to the customer base

Segmentation provides Fennia’s business units insights on customer values ​​and attitudes. Customers can be grouped so that they can be effectively approached at every touchpoint of the entire customer journey. Segmentation is often used as a basis for strategic planning.

By extending the modeled value- and attitude-based segmentation results to the entire, potential customer base, a comprehensive view of the current customer base’s priorities and opportunities is gained. Big differences in customer amounts, for example, compared to the existing market segments, may reveal hidden customer potential.

This lookalike model is based on a classifier that tells the segment to which the non-customers or non-users belong based on the common features of the selected segmentation sample and the entire customer base. The insights of the segmentation can thus be extended to a larger audience without querying the entire customer base.

The results of the extended segmentation are used in implementing different marketing strategies while the planning is constantly complemented and updated with the new customer potential from the model findings. These may include, for example, underrepresented customer segments, particularly prominent products, or other specific considerations such as customer turnover and churn rate or particularly long customer relationships.

Attribution modeling of digital marketing

The attribution modeling of digital marketing actions is in practice the analysis of the different touchpoints throughout the customer journey that led to a conversion or did not. What actions have been taken in which part of the journey that lead to a conversion? How do non-conversion journeys differ from those that led to conversion? When these issues are analyzed by machine learning methods, the effect of each action on conversions can be determined.

After the analysis, it is possible to state how much, for example, continuous display or organic search has brought conversions within a certain time period, most typically a month. When the investments on the different actions are divided by the corresponding conversions, the average cost of the conversions (CPA, Cost Per Action) brought by each action can be measured – and the actions can be ranked. The CPA as a performance indicator guides the optimization of the aggregate marketing budget and its possible allocations in the most effective way.

Continuous monthly attribution is more than the sum of its parts

The one-off attribution modeling gives only a little indication of how effective the different channels are, and in what direction it is advisable to change channel-specific digital marketing media investments to improve CPA. When the attribution modelling is done every month, the significance of random variations in a single modelling decrease. Continuous attribution modelling also helps to detect channel saturation. With continuous measurement, each channel will receive an estimate of the cost breakeven point after which the monthly investments will no longer produce conversions with improved cost-efficiencies.

Fennia’s continuous attribution modeling was introduced and implemented six months ago. The attribution modelling has helped digital planners to adjust both the overall budget and its allocation between different digital media channels. Several trials using continuous data-driven attribution were performed before a reasonable balance was reached between prospecting and re-marketing in programmatic display advertising. At the same time, it was discovered that conversions no longer occur with a similar CPA – so it was time to try and expand the digital media mix with tactical advertising in social media. In due time, modeling will show how well the new channel will perform. The modeling also increased the understanding in the length of conversion paths and the time it takes to convert from the first ad impression. This has contributed to the optimization of the bought impressions and thus led to further improvements in media performance.

When the CPAs of each media begin to stabilize between measurements, it is still wise to continue with the attribution modelling. Experiments can be extended to testing different search engine advertising budgets and, hence, evaluate how organic search conversions correlate with the varying SEM budget. This, in turn, can indicate the need for search engine optimization or even better, the ability to enhance the best performing keywords with strong SEO. As a result, marketing performance continues to grow.