Assessing the Value of Product Recommendations

Sep 28, 2009 8:28 PM  By

Las Vegas – If you’re making a purchase at a grocery store, you may be tempted to buy a candy bar or a gossip rag when you check out. But what about an e-commerce purchase?

Internet merchants can recommend related products at the checkout, too. But what to offer and how to do it can be a challenge.

Personalized product recommendations on Puma North America’s e-commerce site have helped grow its online sales, according to Julian Chu, head of e-commerce for apparel merchant.

Chu detailed Puma’s experience Sept. 23 at the Shop.org Annual Summit at Mandalay Bay Resort and Casino.

When it switched personalized product recommendation service providers in January as a part of its Web redesign, Puma North America ran a 120-day A/B test with a 50-50 split between the new and old providers. The new service showed clicktroughs on recommended products rose 91%, conversion rates on those products jumped 48%, items per order were 4% higher and demand from responders (customers who made a purchase) rose 184% over the old service.

Chu said personalized product recommendations helped Puma convey to its customers that the company was not just about soccer, but was a lifestyle brand. Puma also includes apparel and shows for golf, running and sailing, as well as fashion designed exclusively for the brand.

But Chu said Puma also wanted to drive incremental sales without increasing labor and opportunity costs, and knew it could do so by cross-selling.

“Some of the choices are not at all what a merchant would choose, but this approach allows serendipitous discovery,” Chu said. “The customer would be browsing items he might not have known about.”

Chu said Puma is using personalized product recommendations on its product pages, and category pages as well.

It also uses newly arrived merchandise as a filter for personalized product recommendations, as well as geographic segmenting, browsing behavior, past purchase history and popular items.

So how do personalized product recommendations work? Meyar Sheik, CEO of personalized product recommendation service provider Certona, described it very much like search engine optimization. You tag your online catalog based on the categories you would like to have show up among the recommended products (in Puma’s case, it uses related products, recently viewed, and matching product line).

Then the merchant would use a dashboard to determine attributes to search for, such as color, price range, and size.

For example, Sheik says you wouldn’t want to offer something at a higher price point than what it is the shopping cart, since you should consider the recommendation to be an add-on to the purchase.

And size does matter, to a certain degree, even though offering products by size is a rule of thumb in sales.

“I can tell you behaviorly that if someone is looking for a size 6, they may not be just looking at other items based on size because sizes across manufactures aren’t consistent,” Sheik said.