Recomendation Engines

Recommendation engines (the magic by which Amazon always seem to know exactly what your heart desires) are among the most valuable analytic tools that any retail business can possess.

Some important factors to consider when building a recommendation engine,

  1. It should be a ‘stand alone’ service, rather than being hard coded into a process.  By this I mean that you should be able to ‘call’ your recommendation engine so that it can be integrated into the process that builds your customer emails, your website’s product pages, your website’s shopping basket page, your customer service staff’s desktops, etc…
  2. It should be both ‘customer centric’ and ‘product centric’.  You must be able to provide the engine with a customer ID and have it return the next best n products for that customer.  Likewise, you should be able to give it a product ID and have it return the next n products to be recommended.
  1. It should be commercially orientated.  The model should include the product profitability as well as the likelihood of a successful conversion so that recommendations are made that maximise the likely profitability of the recommendation.
  1. It should have a random element in the recommendations it makes to ensure that the recommendations don’t become self fulfilling.

If all of these elements are in place, then your product list pages will be ranked for maximum likely profitability, your product pages will recommend related products, your customer pages will recommend the best cross sell products, your emails will be full of relevant recommendations, your customer service and sales staff will have tools available to them to make real-time recommendations, your home page will carry customer specific recommendations – pretty powerful stuff no?

We have extensive experience building recommendation engines.  In every single case it has created demonstrable additional profit.  If you’d like to find out more about how we can help you to develop a recommendation engine for your business, please contact us.