CUSTOMERS.COM® RESEARCH FROM THE PATRICIA SEYBOLD GROUP
RichRelevance Recommendations
Dynamic, Automated Optimization of 40 Recommendation Strategies
By Susan E. Aldrich, Sr. VP and Sr. Consultant, February 4,
2010
NETTING IT OUT
Recommendation
engines are a way for content owners—such as merchants, marketers,
and publishers—to present the most interesting content to each customer
at each step in the interaction. Recommendations were popularized a decade
ago by Amazon’s famous “other people who looked at this bought
that” style of recommendation. Today, recommendation solutions are
available from a variety of sources, including software-as-a-service providers
such as RichRelevance.
If you are in retail ecommerce and looking for a recommendation solution, or
a means to personalize interactions, RichRelevance should be on your short
list.
RichRelevance’s focus is recommendations and personalization for retail
ecommerce. It serves more than 200 million recommendations per day to its 40
or so clients, all of which are in retail ecommerce in North America. Customers
include Walmart, Sears, Kmart, The Vitamin Shoppe, Burton, Bass Pro Shops,
and Wine.com.
RichRelevance is successful in part due to its retail roots: its founders include
David Selinger, who led R+D for Amazon’s recommendation technology,
and Tyler Kohn, Overstock.com’s VP of Technology and Analytics. The
key strengths of RichRecs (the engine for delivering recommendations on a
merchant’s Web site) span technology, customer relationships, and operations.
OVERVIEW OF RICHRELEVANCE
RichRelevance
RichRelevance was founded in 2006. It is based in San Francisco, California,
and has 54 employees. David Selinger, its CEO, was formerly head of Amazon’s
personalization R&D. RichRelevance serves more than 200 million recommendations
per day to its 40 or so clients, all of which are in retail ecommerce. RichRelevance
claims that its clients experience a sales increase of 5-15 percent upon
implementing RichRecs, the engine for delivering recommendations on a merchant’s
Web site. RichRelevance’s market has been North America; it has plans
to expand to Europe in 2010. Customers include Walmart, Sears, Kmart, The
Vitamin Shoppe, Burton, Bass Pro Shops, and Wine.com.
RichRelevance is in the services business. It delivers its technology as software
services and accompanies the software services with comprehensive professional
services. RichRelevance clients get ongoing support, monthly site reviews,
and quarterly business reviews to improve their business results and their
merchandising skills.
RichRelevance Product Family
In the two years since it released its enRICH recommendation engine (November
2007), RichRelevance has produced a family of products to address the customer
lifecycle and the key customer touchpoints. See Illustration 1. The product
family is comprised of these software as a service (SaaS) offerings:
•
RichRecs. The engine for delivering recommendations on a merchant’s Web
site and personalizing any aspect of the interaction. RichRecs automatically
tests and optimizes among its more than 40 recommendation strategies to achieve
the merchant’s specific goals.
•
MyRecs. Dynamically builds a profile of each customer’s activities and
displays multiple pages of recommendations that contain likely products and
categories of interest—based on past viewing, search and purchase history,
and other customer information.
•
ClickSee. An interactive grid display of products which changes as a shopper’s
behavior expresses his preferences.
•
Fashionista. A collaboration with Zugara that combines augmented reality, motion
capture, and personalization to enable shoppers to virtually model clothing
via webcam.
•
RichMail. Personalizes each email with product recommendations based on a customer’s
shopping activity. Recommendations are selected when the email is first opened.
•
RichReach. The advertising application that allows clients to deliver personalized
ads to shoppers when they are on other Web sites.
•
RichAgent. Delivers recommendations in the call center environment. Not yet
released.
•
RichMobile. Delivers recommendations in the mobile environment. Not yet released.
•
RichPromo. Provides content targeting and promotional offer generation on the
merchant’s Web site. Not yet released.
RichRelevance
Product Family
© 2010
RichRelevance, Inc.
Illustration 1. RichRelevance’s product family shares a common
set of services which RichRelevance calls the enRICH personalization platform.
The products, which are all SaaS offerings, address multiple touchpoints.
By providing services that bring customers to a site, and that support
the call center, RichRelevance is expanding its customer lifecycle coverage
past Select and Buy phases.
EXAMPLE: WINE.COM
Wine.com offers thousands of vintages to millions of customers, including collectors,
casual shoppers, and business buyers from Fortune 100 companies. In an effort
to mimic the corner wine store experience, Wine.com had created its own set
of recommendations based on top sellers. These recommendations were not dynamic
or personalized, so they fell short of what Wine.com-–and its shoppers—were
looking for. By implementing RichRecs, Wine.com’s recommendations now
take into account the shopper’s geography, browsing behavior, and opinions
of the community; and they leverage the successful purchase patterns on the
site. At every step, RichRecs selects from recommendations offered by its
many algorithms to optimize the experience for shoppers and results for Wine.com,
deciding whether to show what others have purchased, or top sellers, or someone’s
wish list. RichRecs now drives roughly 10 percent of sales on the site, and
has increased average order value by 17 percent.
SUMMARIZING THE RECOMMENDATIONS EVALUATION FRAMEWORK
Recommendations are hot, solving problems from order size to search to personalization.
My guess is that recommendations will be ubiquitous within the next three
to four years. I’m always optimistic on these guesses, but since recommendations
are widely available as a service, rollout can be very swift. Through customer
interviews and research, we have identified the requirements and evaluation
criteria. These criteria are set forth in our evaluation framework1 which
we will be using in 2010 to analyze a handful of the leading recommendation
solutions, culminating in a detailed comparison.
The framework describes requirements in seven categories: guidance and advice,
recommendation structure, managing recommendations, integration, operations,
vendor’s development and maintenance, and product and company viability.
The evaluation requirements are presented in Table B.
RICHRELEVANCE PERFORMANCE AGAINST CRITERIA
Bottom Line
If you are in retail ecommerce, RichRelevance should be on your short list.
Table A summarizes our evaluation of RichRelevance RichRecs in each of our
evaluation categories. Table B provides our detailed evaluations.
Strengths
The three key strengths of RichRecs span technology, relationships, and operations.
In the technology arena, RichRecs’ broad range of recommendation strategies
makes it possible to present a recommendation in any situation and explain
it to the consumer. Another element is the automated optimization that selects
the best strategies to meet the client’s specified goal.
In the operations arena is the demonstrated availability—99.99 percent
to date; and scalability—200 million recommendations per day, and 5 billion
in a week during 2009 holiday shopping, without exceeding 60 millisecond response
time for end-users.
Finally, RichRelevance has a strong customer relationship model, with a sales
and deployment team giving way to an account manager and executive sponsor
for ongoing care. That care includes ongoing site audits, a structured quarterly
business review, and yearly personalization roadmaps, as well as ongoing
strategic consulting and daily support to optimize performance against client-defined
KPIs. Merchants working with RichRelevance can learn how to be effective
with recommendations.
This report continues...
**ENDNOTE**
1) “Recommendation Evaluation Framework,
Version 1,” January 7,
2010, by Susan E. Aldrich, http://dx.doi.org/10.1571/fw01-07-10cc
**ENDNOTE**
(Back to top)
|