In part 2 of our findings of the ChoiceStream 2009 Personalization Survey, we reported a significant decline in consumers’ perception of recommendation quality between 2008 and 2009. According to our survey, the number of online shoppers who received poor quality recommendations in 2009 was 59%, which is more than 30% percent higher than the 45 percent reported in 2008.
These findings aren’t really surprising to us. Today’s consumers expect more from recommendations than they did even a year ago. They expect them to be accurate and on target, so when they’re not, shoppers notice.
The question is, then, what should retailers do about it?
Improving recommendation quality should be job 1 for every online retailer. Losing sales to poor quality recommendations is just unacceptable when you consider that by following some basic rules of merchandising, recommendations can be accurately tuned for each shopper.
Rule 1: Pay Attention to Context
If a customer walked in to a sporting goods store carrying a new, expensive tennis racquet, you wouldn’t you try to sell her another racquet, would you? No, you’d try to cross-sell her a similarly expensive bag to carry it in, or some fancy clothes to wear while playing.
So why is it, then, that when consumers shop online, retailers often completely ignore this basic rule of merchandising and make recommendations to shoppers that are completely out of context?
For example, online shoppers are bombarded with product recommendations that try to sell them more of what they already have in their carts. In a small percentage of cases (e.g., consumables like canned foods or paper products), it might make sense to try to increase order size by recommending more of what a consumer is already buying. But, for most online purchases, once a shopper has made a purchase decision — i.e., she put an item in her cart—product recommendations should no longer include ’similars’ or substitutes. Instead, they should immediately switch over to complements or cross-sells.
It sounds simple, but few recommendation providers take context into account when delivering recommendations, leading to consumer frustration with the recommendations they’re shown.
Rule 2: Learn from Past Shopping Activity
Good bricks-and-mortar retailers know their customers. They know who’s brand conscious and who’s not. Who only buys on sale and who doesn’t care about price and just wants the latest and greatest.
They know this because they know their customers past shopping behaviors and they take them into account when figuring out what merchandise to recommend.
Online retailers need to do the same thing. They need to ‘learn’ from their customers’ past shopping activity so that they can understand each shopper’s persistent behaviors and use that knowledge to recommend products for the next purchase(s).
For example, suppose you have a budget shopper who consistently buys lower-end or sale merchandise. But, in her current session, she’s looking at expensive, status-label products. A recommender that only looks at the current session data would label her a luxury shopper and recommend only high-end products to her. A recommender that also incorporates knowledge about this shopper from her previous sessions knows her better and might recognize this as a gift purchase or a splurge and provide a more balanced set of recommendations.
There are recommendation providers out there, like ChoiceStream, that follow these rules; but there are many more that don’t. When you evaluate recommendation vendors, be sure to ask about context and past purchase behavior. If a provider doesn’t factor these into their product selection process, you should probably move on and look elsewhere. If you don’t and you end up providing poor quality recommendations, your shoppers might just move on instead.
As an avid professional basketball fan, I am watching the frenzied free agent market and teams looking to put the right pieces together to make a run at the NBA championship next year. Disclosure: I am an avid Lakers fan since birth so I wish all of the other teams the best of luck.
