All of your recommendation needs, one platform

What can it do

We challenge our customers and partners to boil down their issues into recommendation problems and we are growing the selection of use cases day by day. Here are some of the most popular ones;

  • eCommerce optimisation (recommending the next product to buy)
  • AI augmented customer interactions (recommending the right information to the customer)
  • Next best action nudges for customers (e.g. recommending where a customer could improve their personal finances)
  • Mining for latent customer insights (recommending what is the truly interesting information to know about a customer
  • Process optimisation (recommending to a customer or employee the next step they should take to complete their action)
  • Proactive churn management (recommending which customers might churn and importantly what to do about it)
  • Etc

Recommendation Conversations

Introducing RecBot, your AI agent.

Streamlining internal communications so organisations can focus on more complex tasks.

Predictive, self-learning conversations

Powered by artificial intelligence, questions are interpreted and relevant knowledge base articles, online resources or subject matter experts within the organisation are recommended.

Predictive, self-learning conversations

A personalised checkout journey for each customer

All of your customer are different, treat them that way.

Tailor their experience with your business to be the best it can be.

Tailored check-out journeys

This video shows our engine creating tailored check-out journeys for real customers based on the products our Recommender Systems suggest they are most likely to buy.

While this example is base on 1 million real flight records, though it could as easily work on financial products or eCommerce transactions.

At the end of the day, data is data.

More models for more accuracy

We reduce model failures by running multiple models in parallel

RecommenderX build hybrid models

Hybrid models are effectively multiple different models being run in parallel and then comparing the output of those models against one another and choosing the most accurate of them or combining them to maximise efficiency.

Most of the inaccuracies that arise from non-hybrid recommender systems are due to the underlying model hitting an edge case that it is not equipped to deal with and therefore failing, by running the models in parallel as one model hits an edge case the other models are able to compensate.




Why hybrid models matter

The main reason that some companies are reluctant to adopt automated systems is because they break too often; a customer will present an edge case to the model and the entire journey will break down. The company has now frustrated the customer and destroyed much of the satisfaction they were working to build.
In fact, a poorly thought out approach to automation can actually reduce customer satisfaction rather than improve it.

This is why hybrid systems are so important, they reduce the number of failures an individual model would encounter by many orders of magnitude.

The difference in failure rates between standalone and hybrid models is so significant that they are almost incomparable

Contact Us