About Loop54

Our leading proprietary AI helps eCommerce leaders overcome the barriers of innovation – swiftly implementing powerful personalisation to create truly relevant customer experiences.

+97%

revenue

Fashion

+42%

value per search

Smart tech

+30%

conversion rate

DIY

Our story

In 2011, Loop54’s founders Mike Odin and Joel Kall realised that existing recommender systems such as Netflix’s Collaborative Filtering Algorithm were powerful and personalised – yet unattainable for many businesses. The algorithm required users with lots of interactions that intersect each other, while regular businesses often had less traffic and a data sparsity problem.

The team saw an opportunity to bring the concept of ‘content discovery’ to eCommerce. All without big data. They successfully developed a product discovery algorithm that was purpose-built for e-tailers.

With this stroke of innovation, Loop54 morphed into a standalone product search and navigation engine built on a completely new type of machine learning foundation.

In 2021, Loop54 was acquired by FactFinder, making it one of the first major consolidations of leading eCommerce SaaS solutions in search, merch and personalization.

Today, Loop54’s powerful eCommerce personalisation technology platform is trusted by leading eCommerce businesses.

We provide one-to-one personalisation on both mobile and desktop, delivering speed and relevance as a complete product discovery platform.

The mission

What do you get when you combine a Mathematician and a Software Engineer? A new approach.

While others engaged problems of on-site search as solely Engineers (focusing on efficient database indexing and query alteration), Mike and Joel took a mathematical approach.

Combining their expertise, Loop54 modelled relations between products using different techniques ranging from Graph Theory to complex Neural Networks.

What Loop54 can do:

Supports exploratory process of mathematical modelling & rapid innovation
Creates isolated code that allows plug & play functionality, without affecting anything else
Lets Mathematicians work and optimise independently of Backend Dev
Enterprise-level personalisation with state-of-the-art machine learning models
Developer-friendly

How it works

The evolution of our models and algorithm

2012 Topic Model	Topic Modelling	Determined underlying structure of the catalogue by reducing the complexity of a products meta-data to abstract 'topics'. Created the initial ability to find Related Results.
2013 Batman	Behaviour Augmented Topic Model and Association Network (Topic Modelling)	Added graph theoritical elements to reinforce the topic model. Improved robustness of 'topics'. Added behaviour-based learning and initial personalisation.
2015 Bean	Behaviour Enhanced Association Network (Graph Theory)	Created a completely new way of representing the catalogue structure inspired by social network analysis and graph theory. Improved reliability and ease of configuration. Reduced model training time by 90%.
2016 Golem	Generalized Organisation by Layered Expanding Maps (Neural Networks)	Created a multi-layered abstraction of the product catalogues structure - analogous to the structure of a well mechandised physical store. Increased conversion rate by 8% compared to BEAN.
2018 Phat Golem	Generalized Organisation by Layered Expanding Maps (Neural Networks)	An evolution of GOLEM where we used swarm intelligence algorithms in model training. Significantly reduced training time and increased robustness of generated model.
2020 Gas Golem	Generalized Organisation by Layered Expanding Maps (Neural Networks)	Greatly increases model adaptability to changing product sets and handles streams of mixed-data. Attempts to extract humanly readable insights from the abstract model.