Recommender System - Offering pro-level customization to any customer
Salomon is a global leader in outdoor equipment. Through decades, the brand has relentlessly stretched the limits of outdoor technology to always better connect practitioners with their body and their playgrounds. With the advent of industry 4.0, powered by technological breakthroughs in robotics, material engineering, additive manufacturing or artificial intelligence, Salomon decided to go beyond product innovation, to radically change it’s product development paradigm. Trail and running shoes would not be produced for the masses anymore, they would instead be co-designed and co-engineered specifically for every user. Thanks to an intensive use of data, advanced algorithms and a completely reworked production system, Salomon intends to durably transform consumers’ relationship to their equipment and to the brand.
What did we do ?
Salomon needed a key partner to deliver this vision to the market, to imagine online/offline user experiences, conceive best-in-class algorithms and deliver systems at scale. That’s why Salomon called MFG Labs.
Defining a mass customisation strategy
Mass customisation is an old story, often limited to simple design customisation or modular approaches to product development. Salomon’s ambition is a much bigger industrial challenge: the brand wants everyone to get access to the same customisation level as only top level athletes have, starting with trail and running shoes. A first consulting phase enabled us to:
- Define the customisation dimensions: contrary to traditional marketing study, a mass customisation approach is not about identifying a consumer’s need. It instead requires to find where needs actually differ from one person to another; then of course the dimension has to be worth it, it has to bring enough value to the user to make the investment relevant.
- Identify the boundaries of customisation: not all configurations are possible. Some components just cannot be assembled together. Some components are designed for a specific use (think of wet or dry playgrounds for example). Integrating these constraints by design makes sure the algorithm will only pick a configuration both produceable and not risking to harm the user with regards to his or her usage.
- Define the degree of freedom in the customisation process: there are many types of customisations. From a very prescriptive approach – where the user could not have a choice in the configuration he’ll get – to a complete “laissez-faire” one – where the brand only puts its capabilities in the hands of the user – up to him to do whatever he wants with it, it’s hard to find the good balance between both. It all depends on the nature of the relationship the brand wants to develop with its consumers.
- Define an algorithmic and data strategy to find the optimised configuration for every user: what data can we collect from users? How to collect this data? How to make a match between a user and one of the hundreds of thousands of possible configurations – when no shoes of the kind has ever been produced yet?
Driving an industrial mutation using agile best practices
Salomon was facing a first ever challenge: the product is not an enhanced version of another one, it’s a completely new one, or rather a gigantic number of new ones. Everything was to be built from scratch. We therefore adopted a software-alike release management process, relying on user-centric methodologies.
Our objective was to mimic, in a couple of weeks only, the entire targeted experience: data collection, configuration recommendation, shoes production, feedback collection and analysis.
We fast prototyped a minimum viable platform to:
- gather preferences, habits and measurements data from users
- deliver a recommended configuration
- automatise feedback collection and manage engagement
- collect data from activity trackers
- systematically analyse and learn from all the data collected and/or generated.
All the data was put in motion in dashboards dedicated to every aspects we wanted to refine with the test:
- the recommendation engine’s accuracy, based on users’ perception of the shoes performance
- feedback on the product’s features, based on qualitative feedback on pain points and appreciated features
- product’s resistance to tear and wear, based on the analysis of pictures regularly taken by users and activity trackers data
- product’s market positioning, based on qualitative feedback from users and A/B testing
- experience design relevance, based on focus groups for every type of users
300 users run the tests, for 3 months, paving the way for a successful market launch.
In only a few weeks, our team was able to develop a set of algorithms and demonstrate a high value use case:
- Thanks to an algorithm running on fresh data able to detect a list of fraudulous accounts twice a day.
- Our client was able to highly reduce bot traffic (divided by 5!), give confidence back to ad publishers, and securize its ad revenues.