Join thousands of weekly readers and receive practical marketing advice for FREE.
MarketingSherpa's Case Studies, New Research Data, How-tos, Interviews and Articles

Enter your email below to join thousands of marketers and get FREE weekly newsletters with practical Case Studies, research and training, as well as MarketingSherpa updates and promotions.


Please refer to our Privacy Policy and About Us page for contact details.

No thanks, take me to MarketingSherpa

First Name:
Last Name:
Jun 20, 2006
Case Study

Three Advanced Ecommerce Tests

SUMMARY: What should you test to improve ecommerce conversions when your site is already doing exceptionally well?

Check out this Case Study for three test ideas, two of which in particular we suspect we'll see popping up on many more e-retail sites shortly:

- "Hidden" chat request form
- Tweaks to automated cross-selling displays

Yes, includes creative screenshots and results:

You’re a highly established direct merchant. You’ve been mailing a catalog for over 17 years, online since 1995 and you’ve sold a gazillion units. After using paid search, partners, affiliates, etc., where can you turn to next for growth?

Personal Creations, a multichannel merchant selling its personalized teddy bears, garden gnomes and the like, faced exactly that dilemma.

The company’s Web site was already optimized and doing a fine job of converting browsers into customers. Plus, they had a low single-digit shopping cart abandonment rate (that’s unheard of in etail).

How could the marketing team raise results further?


Ecommerce Director Tom Dennison and his team decided to focus on improving browsers to add-to-cart conversion.

Here are three key tests they undertook:

Test #1. Revamp the cross-selling and upselling system

Personal Creations includes three suggestions with the heading "You might also like …" on the right-hand side of its product pages. Previously, the suggestions were automated but based on a combination of best-sellers, related items and gut feeling.

Dennison's team ran an A/B split, with half of the visitors seeing items recommended based on the old way and half with a new algorithm based on two behaviors (creative sample below):

-The universe of shoppers who bought the item
-The user's behavior who is browsing for the item

This is significant because a user’s history in a single visit, as he or she clicks on various categories and subcategories, is very specific. By taking that behavior and comparing it to similar behavior of other visitors, the recommendations can be that much more targeted.

Test #2. Hidden live chat test

Dennison’s team also tested a live chat feature, but they didn’t want to shove it in people’s faces. Instead, they used a DHTML floating ad that appeared while users browse.

Shaped like a Rolodex™ card, the ad said, "Click here to learn how to save 10% on today's purchase." It includes an image of a product appropriate to the season. (Example: a picture frame for Father's Day.)

When users clicked on the ad, another window opened and informed them that there was a representative standing by to offer product suggestions and special promotions for the day (creative Sample below).

The floating ad itself did not specifically say, "Click here for live chat" because, as Dennison says, "that's sort of showing your hand." Instead, it reads:

I am gathering today’s online specials to share with you now. Press SEND below to learn about Today’s Special Savings and Offers!

The live chat representatives were not Personal Creations employees but outsourced and compensated based on performance. They're trained to increase average order value by making appropriate suggestions and helping people find particular items.

After the chat ended, visitors were asked to fill out a three-question survey on whether the chat was helpful.

Test #3. Personalized preview pane

A few weeks before Dennison arrived last summer, Personal Creations’ ecommerce team tested a "personalized preview” window that allowed customers to view 15 of the site’s best-selling items at the time with the personalization on it. They picked July since it’s traditionally the company's slowest month because of the gift-giving nature of the products.

After customers entered the message that would be added to the item, they could click "Preview." The preview pane would show the item complete with personalization in a way that looked real.

"Whether it's wrapped around a ball or around a mug, or whether it's on a T-shirt that's kind of wavy, it looks wavy," Dennison explains.


Dennison was very happy with all three tests.

The automated cross-selling and upselling feature took a few weeks to gain "knowledge," but now, "we've seen pretty astonishing numbers,” he says.

The conversion rate is 4 percentage points higher for visitors seeing the new automated recommendations, and both revenue per visitor and average order value is nearly 3.5 percentage points higher.

Regarding the live chat feature, Dennison says 55% of people who click on the floating ad engage in a proactive chat. It "definitely increases ROI," and both average order value and conversions are higher.

Dennison’s team also figured out that they didn’t have to automatically give the 10%-off offer to every shopper. It depended heavily on the nature of the chat that takes place. Some shoppers may be looking for more information, while others may be looking for discounts. However, if the person asks for it, the offer is always given.

Increases varied for the Personalized Preview pane, but in general, conversions from the product page to adding the product to the shopping basket for customers who used the feature increased 30%.

Useful links related to this article:

Creative samples –- automation suggestions and live chat offers:

Scene7, Personalized Creations' vendor for Personalized Preview:

Certona, vendor for automated recommendations:

Personal Creations:

Note: Tom Dennison will speak at July’s AD:TECH in Chicago. For information about the show, visit

See Also:

Post a Comment

Note: Comments are lightly moderated. We post all comments without editing as long as they
(a) relate to the topic at hand,
(b) do not contain offensive content, and
(c) are not overt sales pitches for your company's own products/services.

To help us prevent spam, please type the numbers
(including dashes) you see in the image below.*

Invalid entry - please re-enter

*Please Note: Your comment will not appear immediately --
article comments are approved by a moderator.

Improve Your Marketing

Join our thousands of weekly Case Study readers.

Enter your email below to receive MarketingSherpa news, updates, and promotions:

Note: Already a subscriber? Want to add a subscription?
Click Here to Manage Subscriptions

Best of the Week:
Marketing case studies and research

Chart Of The Week

B2B Marketing

Consumer Marketing

Email marketing

Inbound Marketing

SherpaStore Alerts


We value your privacy and will not rent or sell your email address. Visit our About Us page for contact details.