The key to cracking the first challenge, was to understand what card should be shown first to a potential new customer, which would be most appealing. We solved this by modelling the uncertainty of a card’s performance. This allowed us to balance the use of known “bestselling” cards and of uncertain but potentially “good” cards.
We solved the second challenge by building a recommendation engine that automatically (and dynamically) ranks the cards. Here, it is important to keep track of how many people view a card and how many of them end up buying it. This allowed us to create a statistic model for each card based on its performance.
As a result of the project, the conversion rates increased for many of the categories of greeting cards and most people are able to find their card substantially quicker than before. Moreover, products are now managed and automatically ranked daily.
Read our (free) whitepaper about Recommendation Engines. Achieving one-on-one contact with your customers through personalization technology.