Merchandising to the Minute
Online consumers are increasingly being bombarded with new sources of information, from Internet news outlets and email, to blogs, instant messages, podcasts, YouTube videos, text messages, and viral marketing campaigns. Information now moves so quickly and freely that the buying patterns of entire populations can change in a matter of hours. That’s why online retailers need to “merchandise to the minute” by offering relevant, personalized product recommendations to each and every consumer in real time.
Relevant and timely recommendations don’t just generate incremental sales. They demonstrate to customers that a merchant moves at the same speed they do, staying up on up-to-the-minute Internet trends. Merchandising to the minute enables retailers to build brand equity with consumers and encourages more frequent return visits. Conversely, if recommendations are static and stale, they not only fail to drive incremental revenue, they actually turn customers off to the point of driving them away from a site entirely. In today’s “information overload” world, consumers don’t have time for sites that don’t present them with the products and offers that truly capture their evolving interests at every moment.
To deliver up-to-the-minute recommendations takes some effort, but it’s worth it – because otherwise you’re leaving potential sales on the table and hurting your brand equity. Highly relevant, real-time recommendations are based on continuously changing models of customer behavior. The latest actions customers take on a site are fed back into the recommendation engine in real time, and new models with new conclusions about how customers are behaving are activated dozens of times per day. Instead of taking weeks to react to spikes or troughs in demand, new relationships between products purchased together, or new top sellers in various categories, real-time recommendations react to these changes within an hour or two. When a trend dissipates, often as rapidly as it appeared in the first place, real-time product recommendation engines adapt, rebuild, and make alternate recommendations, either for new up-and-coming products, or for tried-and-true long-term winners.
Before RichRelevance, the most common approach to Internet recommendation systems was to gather massive amounts of data from a site, analyze it for 48 hours or more, and then use the results to make recommendations. On the Internet, delivering recommendations a few weeks late is like delivering farm-fresh milk a few weeks later. It’s sour and unfit for its intended purpose.
For the merchant who wants to deliver the kind of dynamic recommendation behavior that fits today’s Internet lifestyle, there is no substitute for models that continuously adapt to customer behavior.