Home » Case Studies » Case study: Online Retailer

Case Study: Online Retailer

Every day, an online retailer receives several hundred inquiries new and existing customers via various contact forms and e-mails.
The inquiries are from various areas. Questions about products before and after purchase, services or return requests.
The retailer’s Business Intelligence department found that many inquiries were received after users came to the site via PPC campaigns, some of which were expensive.

Customer Service

Customer Service Topic Modeling
The user requests were processed by different teams, some of them geographically separated, and escalated to higher levels if necessary. In this decentralized system, communication with customers develops in different directions and sometimes in different technical systems.

In order to generate a uniform picture of user comments, we have standardized texts from various systems and formats and identified various subject areas for the retailer using topic modeling.
Within the individual subject areas, focal points could be extracted which regularly lead to customer enquiries.

Based on this information, the self-service areas of the website could be revised accordingly. In future, topics that are frequently asked about will be presented more prominently and specific questions will be dealt with in greater detail.


The effectiveness and optimization of marketing activities depends on the quality of the available data.
The web analysis solution in this case only covered the successful sending of a contact form, not the subject matter. Until now, only structured data was available to the retailer for optimizing PPC campaigns: The number of impressions, clicks and conversions.

It turned out that a large number of the alleged lead conversions had different topics than expected. If a user submitted the contact form to send a support or after sales request, this was counted as lead conversion.

By using a neural network, all incoming messages are now classified into different topics and these topics are transferred back into the web analysis solution as a tag.

The marketing team now has not only click conversions and the way to conversion available to optimize the PPC campaigns, but also the topics of the conversions, thus enabling targeted optimization with regard to the user’s intent.

feedback loop