, created by Florian Baur
Game changing technology needs that personal touch
Recommendation Systems are transforming the way we choose products and services but an ‘off the shelf’ approach is unlikely to be the way forward.
‘People who bought this also bought the following..’ - sound familiar? Anyone shopping online, particularly at Amazon, has experienced this type of recommendation. Behind this seemingly simple marketing tactic is a game changing technology which has been quietly growing and evolving. Having already taken huge steps forward, Recommendation Systems (RS) are destined to progressively impact our decisions on what to buy, watch, eat or do. But how will they develop to become increasingly more relevant in the retail arena?
Early systems relied on rules and heuristics to develop recommendations – for example, suggesting the most popular items in a category to shoppers buying any product from that category. Next came content and attribute based systems focusing on specific elements such as ‘premium’ and this data often came from segmentation analysis and/or panel studies.
Pioneered by Amazon’s ‘collaborative filtering, algorithmic approaches were made possible by the increase in computation capabilities which support the creation of more sophisticated solutions and provide accurate results in real-time. Current RS models typically follow a generic algorithm and which lumps people into groups rather than fully personalising the results.
A totally customised environment
People are unique individuals and no single approach to recommendation will work for everyone across all categories. True personalisation really cannot be addressed with a ‘one size fits all’ approach.
The answer is to customise by introducing a significant percentage of intelligence which is driven by fundamental shopping behaviour. Then the underlying system can learn to accommodate the differences between, for example, buying books or milk: few people buy the same book twice but will repeatedly buy one or more brands of milk. Understanding this is the key to relevant and personalised RS.
Therefore, the most advanced capabilities will emerge by integrating personalisation into the broader RS landscape to create cohesive systems. Everyone can receive bespoke recommendations, product offers and incentives completely relevant to their own specific needs – even to the point of varying the percentage of a discount based on individual shopping behaviour. An even more effective layer can be added by introducing real-time to the recommendation process using, for example, APIs, which allow several channels to be serviced via the same interface.
And moving forward even further?
At emnos, we are exploring ways in which RS could ultimately take us into uncharted territory where factors such as time, mood, location, sleep patterns and even social media history are taken into consideration.