Lapperre is committed to optimize the managed care path for its customers with machine learning insights.

In retail, Customer loyalty and repeat purchases are often driven by the regular offering of commercial benefits. Just like "eyesight", hearing can also deteriorate with significant consequences such as social deprivation, learning problems, tinnitus, listening fatigue, cognitive decline. Lapperre's starting point is that in hearing care, the customer journey after purchase should be based upon an effective managed care path for the offered solution and allow the customer to enjoy his hearing and life to the fullest. 

Based on Lapperre's 75 years of audiological expertise and daily care to its customers, the company decided to further investigate into the most effective managed care paths supported by insights from machine learning analysis. For this, Lapperre turned to Cegeka's data science expertise, which resulted in some scenarios with almost 40% efficiency improvement. 

Lapperre is the market leader in Belgium in hearing aids and operator of hearing centres. When a customer buys a hearing aid, it is not a one-time purchase. Hearing loss usually continues to evolve, and the brain adapts. After one or two years, the hearing aid can be adjusted and a few years later, the customer may need a new hearing aid. "How does an ideal managed care path look like in terms of the number of appointments to allow the customer to enjoy his hearing solution to the fullest? That was one of the challenges we had a few years ago to improve our capacity usage," says Dirk Schaele, Director Global Store Performance & Processes at Sonova. 

To answer these questions based on data, Lapperre needed data science expertise. The company found it at Cegeka. "Cegeka guided us in the research, organized the necessary workshops and asked the right questions to arrive at an analytical framework. The statistical validation forced us to continue to investigate new relationships and enrich our data model. We regularly received feedback that our processes were not sufficiently aligned and that the result was not statistically relevant. This gave us important insights into which data and processes were missing," says Dirk Schaele. Based on these new insights, Lapperre started to optimize its processes until there was sufficient statistical relevance and sustainable customer satisfaction. 

Up to 40% more efficiency 

Lapperre wanted to understand which factors influence customers' experience to renew their hearing aid. Cegeka tested no less than eighteen variables to develop a model. "Seven of these variables turned out to be significant," explains Dirk Schaele. This was not only important to make the Managed Care more efficient, but also provided interesting business insights. 

Cegeka was then able to use machine learning to develop a model that ranks Lapperre's customers according to their "care needs" for adjustment and/or the purchase of a new hearing aid. The model was trained on data from previous years and then validated with data from the last year. Finally, this year, these forecasts were used in Belgium to invite customers for a "check-up visit" to test the efficiency of the model. 

What turned out? The use of the model led in some scenarios to almost 40% efficiency improvement compared to the traditional way of working: calling customers randomly or chronologically. Everyone was surprised by this high percentage. "The customers are happy with our proactivity in managed care and the capacity of our care providers is now also being used much better," Dirk Schaele argues. 

“The customers are happy with our proactivity in managed care and the capacity of our care providers is now also being used much better”

Dirk Schaele, Director Global Clinic Performance & Processes at Sonova

Creative and collaborative process

Developing a machine learning model is a creative and collaborative process. An experienced partner support is essential in this, Dirk Schaele notes: "It's a lot of brainstorming, and if you don't have anyone guiding and documenting that process while it's going on, you miss important insights. So, we've benefited a lot from Cegeka's experience and support during this process." 

Of course, the model also needs to be updated regularly. "In the past, the battery of a hearing aid was a good predictor of customer loyalty. When the expected battery life had expired, you would normally see a customer again to replace the battery. But since the introduction of rechargeable batteries in hearing aids, this is no longer a useful variable," explains Dirk Schaele. And so, the development of a machine learning model is a continuous process of growth. Lapperre expects to update the machine learning model once a year. 

Data quality

Dirk Schaele emphasizes that it is important to have data quality in order before you start with an AI journey. And that can take some time. Cegeka began implementing a data platform at Lapperre in 2018, while the company had initiated their digital CRM roadmap back in 2013. According to Dirk Schaele, 'Evidence-based, data-driven work and decision-making are truly learning processes. It wasn't until a year or five later that we understood where it would lead us, and only then did we decide to implement a data platform.' 

Lapperre now shares the model and insights with its parent company for global use in other countries where it operates. Dirk Schaele concludes, "Naturally, there are national differences such as legal frameworks and customer habits, but I anticipate that we can maintain a common layer that is continually enriched with findings from other countries and/or supplemented with local variables."