Skip to the content

Behaviourally augmented search

Schneider Electric – Search

Schneider Electric is a global specialist in energy management and automation with operations in more than 100 countries. They offer integrated energy solutions across multiple market segments.

Schneider Electric is a global specialist in energy management and automation with operations in more than 100 countries. They offer integrated energy solutions across multiple market segments.

Problem / Opportunity

Schneider’s Website Search is the pivotal knowledge discovery mechanism for Schneider’s customers. Schneider has one of the largest products ranges in the sector with several million products across different regions each with their own documentation. Moreover, Schneider has very different and distinct customer types. These include small electricians replacing in-home lighting right through to lead architects designing new airports. Their legacy search engine was unable to navigate this complexity in a performant manner and return relevant results to their customers. They approached us to find a solution to keep their vast knowledge base searchable.

Innovation Team

We formed the innovation team around Schneider’s Search Owner, Global Customer Support Director and our search and machine learning experts. The whole team then attended a workshop in which different approaches were discussed and consensus on the correct approach emerged. After the workshop, we documented the approach including key milestones, release timeline and success criteria. We then began building the system through a series of feedback driven iterations often using video conferences for planning meetings due to the global structure of the team.


There were two key challenges to making the search accurate. Firstly, we needed to identify the user’s customer segment so that we could tell which range of products and depth of documentation would be most relevant. We tracked all user interactions with the site through a JavaScript integration which sent data to our API including search terms, page visits and success criteria. We were able to detect successful search journeys based on a complex set of ambient criteria as well as the user explicitly marking information as helpful. Using this data, we built a machine learning model which identified patterns in the relationships between user segment, search terms, search journeys and success criteria. The second aspect of the problem was how to index the content in a semantically-rich but behaviour-aware format. Essentially, we wanted to combine the model’s understanding of relevance with a deep natural language search. In this way, we were able to augment and filter results from the natural language search with the insights into what similar users had found relevant.


At the beginning of the project we identified three KPIs, click through rate, 0-search results and successful search rate. After rollout, we saw a 48% increase in click through rate, a 96% drop in 0-search results, and a 65% increase in successful searches.

Methodology SE Search Created with Sketch. Search Behaviourally Augmented ALPHA Hack and Craft Schneider Electric Team members PROJECT ROADMAP 2. Development 1. Ideation
Dev SE Search Created with Sketch. integration Smart search results SE SEARCH RESULTS SEARCH-USAGE LEARNING ENGINE 6 5 4 3 2 1 Search results without learning engine Schneider Electric

All case studies



Science and technology are the principal drivers of human progress. The creation of technology is hindered by many problems including cost, access to expertise, counter productive attitudes to risk, and lack of iterative multi-disciplinary collaboration. We believe that the failure of technology to properly empower organisations is due to a misunderstanding of the nature of the software creation process, and a mismatch between that process and the organisational structures that often surround it.