Design
For
a company that is almost completely dependent on conversions on its website to
generate revenue, the design and functionality of the site is paramount. From
the layout, to the checkout, every step must be user-friendly and seamless. The
industry-standard measurement of e-commerce checkout rates is studied by
basically every company and classified as an abandonment rate. Overstock
clearly studies that rate (which will be addressed later as a key performance indicator,
or KPI) looks to constantly make performance-related adjustments. “Overstock’s
site design staff discovered shoppers were adding to carts only X% (the company
declines to reveal the number) of every 100 times the shoppers viewed product
pages. ‘We thought it should be X plus 3%,’ [CEO Patrick] Byrne says. Using
analytics data, Overstock realized that product pages forced shoppers to scroll
down to the bottom half of pages to get the information they needed to make a
purchase decision-a design that caused many shoppers to move on without carting
products” (Demery, 2007). Using web analysis in this case helped Overstock
realize that they had a simple design flaw in their process. By seeing that
they had a significant problem in the consumer funnel towards a successful sale,
Overstock was able to address the issue of the layout and design of the page in
order to more actively cater to their consumers. One of the most basic tenets
of the aforementioned consumer funnel is the ability to see where consumer
conversions hit a snag and cause consumers to leave. In this instance, it
appears that Overstock has done a good job in isolating that funnel snag and
adjusting.
Predictive analysis
The concept of predictive analysis is certainly not
singularly a web analytics tool. Many companies use the concept of predicting
where or what a consumer would want and then try to cater to that need. What
makes Overstock different is their use of predictive analysis based on
attributes that the consumer provides. In this case, Overstock is able to use
their prior sales records and commerce results to create a more accurate
suggestion of potential product choices for their customer. By using a “Bayesian
attribute-based technology”—which basically means predictive analysis, the
company is able to employ mathematics to help consumers find more aptly-suited products
to their needs in Overstock’s GiftFinder application. “The application asks
consumers to specify the occasion, the age of the gift recipient, the relationship
to the recipient, and the interests of the recipient, and then recommends
specific products. Overstock’s CEO notes that the GiftFinder drives 2.5
times the average purchase revenue for
the site compared to when customers don’t use it” (Demery, 2007). Again, here
is another example of the Overstock group’s ability to employ a metric like prior
sales and synergize it with web analytics to help improve their sales results.
Key Performance
Indicators
The
ability to study results and target customers for future sales is a great use
of on-hand resources to try to generate more sales. By studying this data and
others, Overstock is constantly monitoring their own key performance indicators
and efficiencies. From the growth of revenue as a result of the GiftFinder
application to even using event-driven email marketing to customize offerings,
Overstock is able to constantly monitor their own data to try and drive larger
order sizes and more consistent conversion rates. All of this falls under the
concept of trying to reduce abandonment and create sales. Furthermore, it also
underscores the dedication to true synthesizing of analytics data for sales
purposes by Overstock.
Suggestions
While
it is clear Overstock uses behavioral analysis to create recommendations for
consumers, it is unclear if they use it for targeting new consumers. By
potentially employing this same level of analysis to referring websites and
upstream and downstream visitors, Overstock might be able to see exactly who is
bringing in the most traffic to their site. In the previously addressed concept
of abandonment, Overstock would be able to not only see where the consumer left
the conversion funnel, but also where they went to complete the sale and why.
If, for example, via downstream analytics the Overstock group was able to
deduce that consumers were leaving the checkout process at the shipping subtotal
page and going to Walmart’s site, one might imagine that Walmart might be
offering a better shipping alternative. While more analytics would need to be
run to confirm this, perhaps testing a free shipping (or at least reduced-cost)
alternative might be beneficial. Additionally this would also help them in
studying which visitors are converting with a higher propensity and where
visitors are going after abandoning Overstock’s site. All of this analysis
might certainly be already occurring within Overstock’s operation, but the data
is clearly there and must be synthesized correctly to accomplish the goals of
the company.
Overstock
is clearly not the first company to employ these techniques—nor the last. The
key is clearly in using the data available to the marketing team and trying to
figure out the best way to make that data work for you. Furthermore, testing
each model against a baseline of data is paramount in understanding if there is
something positive happening for the company. Behavioral modeling and
predictive analytics are certainly great forays into that equation.
References
Davenport,
T. (2009). Realizing the Potential of Retail Analytics; Plenty of Food for
Those with the Appetite. Babson Executive Education White Paper. Retrieved from
http://analytics.typepad.com/files/retailanalytics.pdf
Demery,
P. (2007, February 22) How analytics supports site design
at Overstock.com. Internet Retailer. Retrieved from http://www.internetretailer.com/2007/02/22/how-analytics-supports-site-design-at-overstock-com
