DELA dives deep into returns on marketing with predictive modeling

An insurance firm for funeral services, Dela boasts 4.000.000 customers in Belgium and the Netherlands. In search of more effective marketing communications, DELA Chief Marketing Officer Marysia Kluppels and her team asked delaware to develop a customized data model capable of accurately predicting returns on its campaigns.

“We wanted to get a more predictive view of the expected outcomes of our marketing campaigns,” Marysia explains. “Doing so would enable us to simulate the impacts of changing certain variables in our marketing mix, empowering us to choose the combination with the highest predicted return. In other words: we’d be able to get definitive answers to the question of ‘which marketing-mix creates the most customers and new contracts?”

Establishing new traditions

Like most marketers, Marysia and her team previously approached new marketing campaigns using what worked – and discarding what didn’t work. “We took what we had learned and tried to improve on it – but it was essentially guesswork and often meant spending our limited budget without guaranteed success. We needed a new approach.”

Using DELA’s contract and payment data combined with historical marketing campaign data, the analytics experts at delaware created a statistical model with an impressive 82% accuracy rate. “This model allows us to spend our budget on those channels, messages and approaches that create the highest value, which puts this money to the best-possible use.”

“This model allows us to spend our budget on those channels, messages and approaches that create the highest value, which puts this money to the best-possible use.”

DELA Chief Marketing Officer Marysia Kluppels

The more variables, the richer the predictions

The data was obtained from the Selligent marketing cloud as well as DELA’s Sitecore website and contract database. It was processed using Microsoft Azure’s built-in machine-learning tools to create the predictive model.

For the model to work, it was crucial for the data pool to contain both past campaign characteristics (online, offline, channel, timing, message, target group, etc.) as well as results. “Using these ‘soft’ characteristics of the campaign was really key,” Marysia goes on to say.

Beyond simple marketing outcomes

Predicting marketing outcomes has obvious uses, but Marysia’s team can use the model to do even more. “Now that the predictive logic is in place, we can change the questions we ask it. For example, if we ask ‘what is the optimal marketing mix needed to realize a 10% growth in customer base?’, the predictive model will return the best combination of target groups, timing, channel mix and budget,” explains Marysia. “This approach combines both the art and science of marketing.”

Ready to harness marketing communication insights before even starting the campaign? Our team is ready to help you design the predictive analytical model best suited to your business case. Get in touch.