One
of my favorite e-commerce retailers is Williams-Sonoma. Williams-Sonoma is a publicly traded company with
sister chains including Pottery Barn, Mark and Graham and West Elm brands
(SeekingAlpha, 2013). This post explores
analytics practices as they apply to the Williams-Sonoma brand.
In
2013, Williams-Sonoma CEO, Laura J. Alber, announced that store sales were
down; however, their e-commerce segment grew 20% and was responsible for 49% of
their second quarter’s revenue (SeekingAlpha, 2013). So how did Williams-Sonoma grow e-commerce
sales while realizing in-store losses?
Williams-Sonoma enlisted a Revolution
Analytics product called Upstream.
Upstream was based in the medical industry as a predictive modeler of
potential patient expiration or treatment outcomes (Dusto, 2012). The underlying methodology was converted to
process marketing online and offline marketing data to product recommendations
to match the best channel to each customer and to determine how best to target
campaigns to those customers to drive conversions both in-store and online.
Online and Offline Data Sources
Williams-Sonoma provided
online data consisting of web analytics logs, email send data, search data,
sales files and offline data including sales files and catalog mailing lists
and schedules to be fed into Upstream’s engine.
Next, the retailers typical
uptick in sales for specific holiday periods were added, followed by the names
of individuals who carry a the Williams-Sonoma branded Visa Card. The introduction of seasonal and
customer-specific factors was captured so that they could be considered when
allocating customer dollars across the marketing treatments. “Marketing treatments” can be thought of as
“customer touch points” that influence a specific customer to purchase
(Revolution Analytics, 2013).
Know Your Customer at a
Micro Level
Many marketing analytics
packages aggregate customer behaviors to make recommendations; however, the
team behind Upstream tracks customer activity as the micro level. This includes each click-through from an
affiliate site, each click-through from a Williams-Sonoma generated email, each
catalog that is mailed to a home and each in-store promotion (Revolution
Analytics, 2013).
Why is this important? The theory is that more than one marketing
treatment may be responsible for each conversion or sale. For example, Williams-Sonoma mails a catalog
to a customer’s home. The customer pages
through the catalog and finds a Breville Countertop Oven that interests
them. It is $250 and Breville is a
respected name so they decide to go to the web site to find out more. On the web site they watch a video
demonstrating the product but they do not purchase. Later, the customer receives an email (based
upon Williams-Sonoma’s web analytics and CRM integration) with a free shipping
offer. The customer decides to
purchase. Was the e-mail responsible for
the purchase or did the catalog and web site interaction help drive the
conversion? According to Upstream’s
logic, all three treatments were responsible for the sale in various
percentages and thus a portion of the sales dollar is attributed to each
treatment proportionally.
Partial Residual Theory and
Decay of Treatment
The partial residual theory
asserts that each marketing treatments influence decays at a different
rate. The rate of decay for an email or
an Internet search is much more rapid than a catalog (Revolution Analytics,
2013). To determine the allocation of
spend across each treatment combines this decay of treatment scale with the
recency of consumer touch point.
Returning to the example of
the $250 Breville oven purchase. The
customer received a catalog and subsequently visited the online site; however,
they did not immediately purchase. Time
passes and the customer receives an e-mail to offer them free shipping for the
product by redeeming a promotion code during checkout. The e-mail would receive a portion of the
sales due to its recency; however, without the catalog enticing the customer to
visit the online store, the custom email would not have existed thus a greater
share of the sale is attributed to the catalog based upon the slower decay of
effectiveness of the catalog.
Additionally, the online visit would also receive a portion of the sale.
Figure 2. Attribution of sales across marketing treatments. Courtesy of Revolution Analytics (2013)
Results
All this may sound
fascinating, and it is even to an advanced mathematically challenged individual
like myself, but what does this extra number crunching tell Williams-Sonoma? Glad you asked. The output of Upstream provides Williams-Sonoma
with improved ability to:
· “Identify the most profitable channels for
every customer and the most profitable customers for each channel” (Revolution
Analytics, 2013)
· “Target the right customers at the right time
with the right message” (Revolution Analytics, 2013)
· “Understand how the spend in each marketing
channel impacts sales in order to properly budget marketing dollars for each
channel (Revolution Analytics, 2013).
Early results substantiate
Upstream’s claims of being able to deliver marketing improvements by taking a
scientific approach. In the words of
Mohan Namboodiri, vice president of customer analytics for Williams-Sonoma “We
have seen our ability to target with the catalog improve using these techniques
on a scale that we have not seen with any sort of small technical improvement”
(Dusto, 2012). Additionally the
“qualitative improvement in our [Williams-Sonoma’s] ability to target the right
type of customer with the right type of messaging” has greatly improved their marketing effectiveness (Dusto, 2012).
Room for Improvement?
I did not uncover any
existing metrics pertaining to web site performance or those used as a basis
for the analytics provided to Upstream; therefore, my recommendation for
improvement is one based upon my exploration of the Williams-Sonoma site.
One of the main selections
that a visitor can explore is a Recipes
section. If a visitor finds a recipe
that they like, they are able to print it from the web or they can save it to
their own personal recipe box [after setting up a site profile]. Additionally, the majority of recipes are
presented with products that can be used to create the recipe. I believe this section could be used to
increase conversions by generating an email to the visitor who saves a recipe
to the profile.
If the recipe box owner is
a customer, Williams-Sonoma can review purchase history and make product
recommendations based upon products used to create the recipe that the visitor
does not own. An email is then generated
to the customer and provides a discount or free-shipping opportunity for a
product. Marketing tactics pertaining to
the conversions based upon targeting recipe box customers can be easily
monitored by Google Analytics and Next Analytics for Excel.
The addition of short
technique videos to accompany difficult or advanced steps within a recipe may
also increase conversion rate. These
videos may also be published (and/or linked) to the Williams-Sonoma YouTube
channel. YouTube Analytics will report
statistics including number of views, number of repeat views, number of
subscribers, shared videos as well as how far the viewer watched the video
before leaving.
The
output of the YouTube Analytics, Google Analytics and the Next Analytics for
Excel would then be an input to the Upstream application to further defines
target market messaging treatment strategies and the attribution of revenue
across Williams-Sonoma’s many marketing channels.
References:
Dusto (2012, April 14). Marketing Technology -
Williams-Sonoma
targets e-customers with a “treatment” approach - Internet
Retailer.
Retrieved February 22, 2014,
from
Revolutions (2013, April 29). How UpStream uses R for
Attribution Analysis.
Retrieved February 22, 2014,
from
http://blog.revolutionanalytics.com/2013/04/upstream-attribution-analysis.html
SeekingAlpha
(2013, August 28).
Seeking
Alpha (2013). Williams-Sonoma, Inc. (WSM): Williams-Sonoma Management
Discusses Q2 2013 Results - Earnings Call Transcript - Seeking Alpha.
Retrieved February 22, 2014, from
http://seekingalpha.com/article/1663232-williams-sonoma-management-discusses-q2-2013-results-earnings-call-transcript?page=2
YouTube (2014). Views reports Help.
Retrieved February 20, 2014, from
YouTube
(2014). Engagement reports Help. Retrieved February 20, 2014,
from https://support.google.com/youtube/topic/3029004?hl=en&ref_topic=3025741
