Log in

If You Liked This DVD, You Might Also Like…

21 Jul, 2003 By: Holly J. Wagner

Retailers employ a variety of strategies for building customer loyalty in the new realm of online DVD sales and rentals.

Strategists at relative newcomer Walmart.com and No. 1 online rentailer Netflix both say customers come to them looking for value and a broad range of titles — often titles outside the mainstream — but they take different approaches to service.

Netflix — which was the first to bring the concept to market and has the largest base of subscribers, with more than 1 million — puts a lot of emphasis on service and its Cinematch recommendation engine.

“There are three main ways that the engine drives rental,” said Neil Hunt, VP of product development at Netflix: One, the customer asks for a recommendation; two, Netflix presents the information in a customized order to appeal to a specific customer (“It is basically the same thing as a customer going into the store — we rearrange the shelves to put the things that customer is going to like best on the shelves”); and, three, Netflix directs the customer to movies that customers with similar tastes liked (“If you liked this movie …”).

The Cinematch engine, which is updated monthly, is a key element of the Netflix experience and one that executives credit, in part, for the site's success.

“I believe somewhere in the vicinity of 40 percent to 50 percent of the movies that people rent from the site are in some sense mediated by data coming from the recommendation system,” Hunt said. “It's based on collecting ratings from members on movies that they have watched with Netflix or elsewhere. If you imagine a giant matrix where the rows are the customers and the columns are the movies, you put numbers in the cells representing each customer's rating for the movie. If I am trying to predict a blank cell for a particular customer for a particular movie, I'm going to match that against customers with similar ratings patterns.”

The site was due to get a new feature this week that explains how Netflix generates the ratings, Hunt said.

Walmart.com uses only a search engine, not a recommendation system, to generate suggestions.

“What we do is, when you select a particular movie, there are two to three other movies that are identified as being similar,” Sevick said, adding that the site, like its competitors, uses a proprietary algorithm. “It is subjective and is applied more at the category [genre] level than by title. They are identified by human action [i.e. a data entry person].”

The site focuses more on price, selection and convenience than suggestions. The company is adding rental distribution centers within existing corporate facilities — the site has six to Netflix's 20 — to speed delivery times.

“Our theory is consistent in the way we run the Walmart.com site. Our customers are focused on ‘Give me the great Wal-Mart deal online,’ said Matt Sevick, Walmart.com's DVD rental manager. “When we've talked to the Wal-Mart customer, their level of comfort with online recommendation engines is low.”

As with Netflix customers, though, Walmart.com renters are more interested in finding obscure titles than brick-and-mortar renters.

“They are pushing more for what's beyond their local video stores,” Sevick said. “The demand is generally fueled by word-of-mouth from friends who saw a film at an art house, or smaller releases, or a published review. Then there are some that are just harder to get. Imax movies have been very popular among our customer base.”

Sellthrough e-tailer Amazon.com splits the difference, with a recommendation engine that offers suggestions and competitive pricing.

“You can tell from the very prominent placement that the personalization features are very important to our Web site. If you go to the DVD home page, you'll see it right at the top of the page,” said Matt Round, director of personalization for Amazon's Web properties worldwide. “Home page recommendations will not be locked into any category. It's like the corner store, where you come in and the clerk recognizes you and says, ‘I have a new book for you.’

Round would not disclose the mechanics of the engine, but said it uses a customer's purchase history and some unique algorithms.

“We think we have a very strong offering in general. We offer pretty unusual and interesting titles, in part because of the breadth of our catalog,” Round said. “We're not limited to talking about top sellers. We can recommend fly-fishing DVDs for specific rivers in Tennessee.”

Sevick and Round would not discuss user counts, churn rates or conversion rates based on recommendations. Round did note that recommendations within the DVD store are confined to DVD, but recommendations from the home page may cross product categories. Netflix's Hunt, however, offered some insights.

“Compared with Amazon, we think we have made some important improvements because we are dealing explicitly with movies, not with other products like Amazon is doing. Certainly the more constrained area helps,” Hunt said.

“The other thing that helps is we have managed to collect a substantial number of ratings, perhaps a quarter of a billion of ratings,” he added. “We have between two and 10 times as many ratings on our site as [Amazon partner] IMDb does. The volume and the movie specialization would be the strengths.”

Add Comment