Technology

Darwin & Goliath has developed a unique concept and technology to provide your users with the best possible recommendations. Darwin & Goliath is the first and only recommendation-as-a-service provider that uses these technologies.

Multi Recommender-System Framework

Darwin & Goliath is built on top of multiple state-of-the-art recommendation frameworks combined with our own proprietary algorithms. This allows us to tailor our RaaS to your needs, i.e. your type of data and users. For instance, if your platform has many users with many interactions (e.g. ratings or purchases) on relatively few items, collaborative filtering algorithms are used (Apache Mahout).  If you have linked data, graph-based recommendation algorithms are used (neo4j). If you have lots of textual information, content-based filtering algorithms are used (Apache Lucene/Solr). And if your data and users match several criteria, then we simply use all frameworks and pick the best framework and algorithm individually for every user and item on every request with our novel micro recommender system technology.

Micro Recommender System Architecture

Darwin & Goliath has developed a unique technology that we termed “Micro Recommendations”. Our Micro Recommender System automatically A/B tests a plethora of algorithms on your platform and deep learns when each of the algorithms performs best. Once the learning phase is complete, our AI Engine predicts for every single recommendation request, which recommendation algorithm to use.

The idea behind micro recommendations is based on our research showing that there is no single-best recommendation algorithm that performs well on every platform. Even on a single platform, users, items and contexts are too different for one algorithm to perform well all the time (even the most sophisticated deep learning algorithm). We found that, among others, at different times of the day, for different ages groups, and for different amounts of recommendations, different algorithms performed best.

Consequently, Darwin & Goliath might use, for instance, content-based filtering for user A who visits your website in the morning; collaborative filtering for user B regardless of the time of the day; and most-popular recommendations for user A in the evening. Darwin & Goliath uses hundreds of features to identify the ideal algorithm, every single time recommendations  are requested for one of your users.

User Interface Optimization

Users´satisfaction does not only depend on the relevance of recommendations. Therefore, Darwin & Goliath focuses not only on optimizing recommendations themselves but also on the presentation. Darwin & Goliath automatically creates user-interface variations (e.g. slightly different colours, font sizes and numbers of recommendations). These variations are A/B tested over a few weeks. Once the A/B tests are completed, the best performing user interface is chosen for your platform.

Private and Public Inventory Repositories

By default, Darwin & Goliath recommends only your own inventory to your users and only to your users. However, if you like, you have two additional options.

  1. You can activate external inventory to be recommended. External inventory comes from our content partners. For instance, in the domain of digital libraries, we cooperate with the open-access project CORE. CORE has indexed more than 100 million open access research articles. If you choose, these articles can be recommended to your users in addition to your own inventory or exclusively. This means you can even use Darwin & Goliath to recommend items to your users if you do not have your own items to recommend.
  2. You can make your own inventory public. In this case, your inventory will be recommended to our clients who activated external inventory recommendations. However, this option is subject to a quality review of your inventory by Darwin & Goliath.

GDPR Compliance & Data Privacy

Darwin & Goliath is aware of the importance of your users´ privacy, and the regulations introduced in Europe as part of GDPR. Therefore, Darwin & Goliath optionally offers a proxy solution that offers the highest possible data privacy for your users. This proxy is between Darwin & Goliath JavaScript Client and the API. All requests from the client are sent to the proxy which filters users´ personal data including IP addresses etc. In addition, all our servers are located within the EU.

Find out more about our state-of-the-art recommendations API