Case Study

Faster advice at DYI stores thanks to MLOps

Machine Learning Reply has used the MLOps concept to develop a solution for an app for a home improvement retailer, that automatically routes customer inquiries to the right experts.


The challenge

Customer contact - direct and digital

To build customer loyalty, a German DIY chain relies on its digital service platform. Among many other features, the app enables customers to get in touch directly with a specialized product advisor from their local store. With this service, the company promotes an end-to-end and personal customer experience. But the growing number of inquiries was increasingly becoming a challenge for the company. The prompt assignment of the right domain experts could no longer be reliably guaranteed. A solution that could provide a quick response to customers despite the high demand was needed.


The solution

Automation for shorter response times

Our experts at Machine Learning Reply helped the DIY store speed up the processing of customer requests received in its app. A clear plus in terms of customer service. With the help of a special machine learning model, customer inquiries are now forwarded to the right expert completely automatically. The product architecture implemented for this is based on the AWS services Lambda, MLFlow and SNS.

How we did it

More efficient models

thanks to MLOps

Machine learning models like the ones for the app of this DIY enable systems that learn themselves - and thus often have immense competitive advantages. However, it is precisely this advantage that makes the models very complex at the same time. "Classic" software is given certain rules via the code, based on which it processes the data available to it. Machine learning models, on the other hand, create their own set of rules based on the data, which they can in turn apply to other data. For example, a change in data distribution, known as data drift, can have extensive effects on the entire model. This is a dynamic that is difficult to control. In order to be able to manage the resulting complexity and make the solution as efficient as possible, the experts at Machine Learning Reply relied on central MLOps principles for the solution for the home improvement retailer.


The results

Fast, scalable and cost-efficient.

Thanks to the support of Machine Learning Reply's experts, today the customer benefits from a wide range of advantages:


Reduction in response time from minutes to seconds.


70% reduction in manual effort for service provisioning.


Reduction in deployment time of new features from weeks to minutes.


Cost reduction by combining different services on a cost-per-use basis


Increased scalability
of the system


Easier system maintenance
in future


Machine Learning Reply offers customized end-to-end data science solutions covering the entire project lifecycle - from initial strategy consulting, data architecture and infrastructure issues to data processing and quality assurance using machine learning algorithms. Machine Learning Reply has extensive expertise in data science across all key industries of German HDAX companies. Machine Learning Reply empowers its customers to successfully introduce new data-driven business models as well as optimize existing processes and products - with a focus on open source and cloud technologies. With the Machine Learning Incubator, the company offers a program to train the next generation of decision makers, data scientists and developers.