Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
#EpicFail! A/B Test Results vs. UX Best Practices (Boston UPA 2012)
1. #EpicFail!
A/B Test Results vs. UX Best Practices
Susan Rice & Kirk Doggett
Global UX Center of Excellence
UPA Boston Conference 19/12/2013
2. Overview
• Introduction
• Examples
– Pop Up Windows
– Progress Indicators
– Homepages
– Buttons
– Website Purchase Flow
– People Don‘t Read
• Summary
• Questions?
UPA Boston Conference 29/12/2013
3. Who We Are
Vistaprint provides high-impact personalized
products and electronic services that empower
micro businesses and consumers at an affordable
price.
Company Facts
• 25 localized websites
• 3,700 employees
• 3 state of art manufacturing facilities
• 13 offices worldwide
UPA Boston Conference 39/12/2013
4. ―Test Before You Invest‖
We have dozens of multivariate
―split run‖ site tests live at any
given time during our 3 week
release cycles.
• Scalable:
Most site changes are measured
before rollout
• Low Risk:
Multiple tests can be run
simultaneously with a sample
of the population
• Accurate:
Test sample is representative
of user base
4
Random
Sample
Price
Control
Price
Test
UI
Control 5% 5%
UI
Test 5% 5%
User
Population
UPA Boston Conference9/12/2013
5. UX & A/B Testing
• UX designs the UI and
helps inform the test plan.
• Test runs and is analyzed.
• Layer the quantitative with
the qualitative to create a
holistic picture to inform
future design decisions.
UPA Boston Conference 5
But what happens when A/B testing results challenge tried
and true usability principles and best practices?
9/12/2013
7. What the experts say…
• Avoid popup windows: [1]
– Extra windows keep users from getting to your site content, and even if such
windows contain valuable information, users are likely to dismiss them
immediately as ads.
– The other drawback to popup windows is that they go away — once the user
gets rid of them, they're gone, so users often can't find the information again
even if they want to.
• Do not have unsolicited windows or graphics ―pop up‖ to users.[2]
– Users have commented that unsolicited windows or graphics that ‗pop up‘ are
annoying and distracting when they are focusing on completing their original
activity.
• Selective attention is very powerful, and Web users have learned to stop
paying attention to any ads that get in the way of their goal-driven
navigation. [3]
– “pop-up purges mean that users close pop-up windoids before they have even
fully rendered; sometimes with great viciousness‖
[1] Jakob Nielsen, excerpt from 113 Design Guidelines for Homepage Usability, 2001
[2] US Government, Research-Based Web Design & Usability Guidelines, 2006
[3] Jakob Nielsen, Top 10 Mistakes in Web Design, updated 2011
7UPA Boston Conference9/12/2013
8. What the experts say…
• You’re going to lose my business if your web site
experience is annoying.
• According to preliminary information from the survey…
major peeves about commercial Web sites were: [1]
– Pop-up advertising (34.9 percent)
– Registration log-on pages (16.7 percent)
– Software installation (15.7 percent)
– Slow-loading pages (9.1 percent)
[1] John P. Mello Jr. http://www.ecommercetimes.com/story/44966.html, July, 2005
8UPA Boston Conference9/12/2013
9. In 2007 we added a pop-up…
Control Test
UPA Boston Conference9/12/2013 9
10. Results of adding the pop-up
• Significant increase in conversion rate across
most countries and across most products, from
5% to 11%.
• More business cards were created
Most Popular
Most Popular
Beauty/Massage/Spa
Beauty/Massage/Spa
Construction & Trade
Construction & Trade
Flowers & Candles
Flowers & Candles
Music & Entertainment
Music & Entertainment
0
5000
10000
15000
20000
25000
3700 3701
Sum of Documents Created
Feature Value Id
Category Name
Control Experiment
Tested in August, 2007
10UPA Boston Conference9/12/2013
11. Today:
In 2008, Nielsen declares the lightbox as the
interaction design technique of the year [1]
[1] Jakob Nielsen, 10 Best Application UIs, Alertbox, 8/12/2008
UPA Boston Conference9/12/2013 11
13. What the experts say…
• Show the steps of the process [1]
– Recommended for checkout
• Guideline: Provide feedback to let users know where they are in the
Web site. [2]
– Feedback provides users with the information they need to understand
where they are within the Web site, and for proceeding to the next activity.
• In this example, note the right-pointing arrows. Nielsen discusses
his changeover from colons to arrows. The rationale for the arrows
instead of colons or other indicators was because ―the arrows seem
to be slightly easier to understand as an indication of moving deeper
and deeper into the site.‖ [3]
[1] Nielsen Norman Group, E-Commerce User Experience, Vol. 4: Shopping Cart, Checkout and Registration, 2011
[2] US Government, Research-Based Web Design & Usability Guidelines, 2006
[3] Beth A. Martin, Breadcrumb Navigation, November 2006
13UPA Boston Conference9/12/2013
14. We tried removing the progress indicator
• Removing progress bar from Options improves
conversion
Test
Control
Results:
• We observed a
significant increase
of +0.47% of the
conversion rate
overall.
• The test
significantly
increased the
number of sessions
that make it to the
Cart.
14UPA Boston Conference9/12/2013
15. We tried adding a progress indicator…
• Adding a progress bar to the design pages hurts
conversion:
15
Test condition sessions CR CR Change
Control 50175 7.33%
Display Nav 49362 7.06% -3.69%
UPA Boston Conference9/12/2013
16. Progress Bar visual design…
Control
Test
Visual redesign hurts
conversion.
Results:
• CR is down 0.9% for
New Customers
16UPA Boston Conference9/12/2013
18. What the experts say…
• “Clearly designate one page per site as the official
homepage.” [1]
• For a decade, one of the primary homepage usability
guidelines has been to designate a single page as the
one and only official homepage for any given website.
Users are confused when several pages are referred to
as ―home.‖ [2]
• [3]
[1] Jakob Nielsen, 113 Design Guidelines for Homepage Usability, 2001
[2] Jakob Nielsen, Overloaded vs. Generic Commands, 12/19/2011
[3] US Government, Research-Based Web Design & Usability Guidelines, 2006
18UPA Boston Conference9/12/2013
19. Two Home pages?
• It worked for HomeRuns in 2001:
19
First Time Home Returning Home
UPA Boston Conference9/12/2013
20. A decade later…
• For returning customers, conversion rate is significantly higher for the
Returning Home page than from the First Time Visitor home page.
First Time Home Returning Home
20UPA Boston Conference9/12/2013
22. What the experts say…
• ―Ensure that a pushbutton‘s label clearly indicates its
action.‖ [1]
• ―The text on the button should begin with a verb. [2]
• ―Trigger words are the words and phrases that trigger a
user into clicking. They contain the essential elements
to provide the motivation to continue with the site.‖ [3]
• Fitts’s Law: ―The time required to move to a target is a
function of target size and distance to the target.‖ [4]
UPA Boston Conference 22
[1] US Government, Research-Based Web Design & Usability Guidelines, 2006
[2] UX Booth, Good Call-to-Action Buttons, 2009
[3] Jared Spool, The Right Trigger Words, 2004
[4] Universal Principles of Design, Revised and Updated: 125 Ways to Enhance Usability, Influence
Perception, Increase Appeal, Make Better Design Decisions, and Teach through Design, 2010
9/12/2013
24. Cart Buttons
UPA Boston Conference 24
Significant decrease in CTR to ShippingSignificant increase of CTR to Shipping
Control
Test 1 Test 2
• We tested larger target areas with improved visual
hierarchy and verbs with security info like Staples.
Increased cognitive friction?
9/12/2013
26. Checkout Buttons
UPA Boston Conference 26
Could simpler buttons actually decrease
cognitive load and prevent overthinking?
• We tested distinctive trigger words on our checkout
buttons and similar language to Amazon.
CR decreased significantly in FR
CTR decreased significantly from
shipping to confirmation in top 5 locales
Shut off after 9 days in testing
Control Test 1
9/12/2013
27. Product Page Buttons
• We tested rollover states, more prominent call-to-actions
and larger targets.
UPA Boston Conference 27
Control Test 1
Significant increase in
CR of 1.5% was seen.
Law of diminishing
returns?
Test 2
Test 2: No impact.
9/12/2013
29. What the experts say…
• Offer ―sufficient information about your products to answer users‘ pre-
purchase questions.‖ [1]
• ―Helping customers see what they want, when they want it, and in the
way they want to see it, is substantially preferable to forcing them
through a hub-and-spoke, back-and-forth path.‖ [2]
• ―The number of clicks isn't what is important to users, but whether or not
they're successful at finding what they're seeking.‖ [3]
And internally we‘ve learned…
• When a customer becomes invested in one of our designs, the more
likely they‘ll convert.
• Multiple screens/clicks are OK if customers know they‘re making
progress.
UPA Boston Conference 29
[1] NN/g, E-Commerce User Experience
[2] Bryan Eisenberg et. al., Call to Action: Secret Formulas to Improve Online Results, 2005
[3] Jared Spool, Testing the Three-Click Rule, 2003
9/12/2013
30. Design / Flow
UPA Boston Conference 30
Get started Choose Design Template
9/12/2013
31. Design / Flow
UPA Boston Conference 31
Preview
Template
Personalize! Checkout Process
9/12/2013
32. That seems to work well. But what if…?
UPA Boston Conference 32
Significant increases to Website CR for
new customers.
• DTI CR increased 9.58%
• Acquisition campaigns increased 5.18% Less friction?
9/12/2013
34. Good ol‘ Gary Larson…
UPA Boston Conference 349/12/2013
35. What the experts say…
• ―We've known since our first studies of how users read on the
Web that they typically don't read very much.‖ [1]
• ―Satisficing describes the situation where people settle with a
solution to a problem that is "good enough".‖ [2]
• “Users don‘t make optimal choices…Neither do they scan
web-page in a linear fashion, going sequentially from one site
section to another one.‖ [3]
UPA Boston Conference 35
[1] Jakob Nielsen, How Little Do Users Read?, 2008
[2] Interaction-Design.org, Satisficing, 2004
[3] Smashing Magazine, 10 Principles of Effective Web Design, 2008
9/12/2013
36. Product Page
UPA Boston Conference 36
Control
Test 2
Significant increase in
paper stock upgrades
9/12/2013
Relevant copy is
helpful for decision
making.
37. More Product Page Testing
UPA Boston Conference 37
Control Test
Marginal increase in GM/order
Significant shift in take rate of
100 lb. premium and recycled paper
and a decrease in Matte and Glossy.
So copy is making a difference…
9/12/2013
38. Sign In Button
UPA Boston Conference 38
Significantly hurt
overall CR by 1.25%
Sign in page exit rate
increased significantly.
Shut off after 7 days of
testing
Control
Test
Even changing 2
letters can have an
impact.
9/12/2013
39. Summary
• Best practices are a great starting point but they‘re
general.
• A/B testing can help you zero-in on specific changes
that can help you make changes that benefit your
site and your customers.
• Context and relevancy matters.
• Even small changes can make a significant
difference in site performance.
• Nothing is straightforward and simple.
UPA Boston Conference 399/12/2013
40. Summary
• There‘s a healthy tension between Standardization
and Innovation.
• This isn‘t the end state. We‘ll continue to optimize
and learn from our customers.
• It‘s all about balance… and learning more about
your customers through various avenues…
• What the experts say: It depends… [1]
UPA Boston Conference 40
[1] Jared Spool, User Interface Engineering
9/12/2013
41. Summary
• Consistency is one of the most powerful usability
principles…
• A foolish consistency is the hobgoblin of little
minds…
- Ralph Waldo Emerson
• A/B testing helps us learn when context is more
important than consistency.
UPA Boston Conference 419/12/2013
43. Thanks!
Special thanks to Dave Yuknat for his fantastic editing
expertise… and for gathering supporting examples for the
presentation.
Susan Rice leads global UX design and
Kirk Doggett leads global UX research at Vistaprint.
And just a plug… we’re hiring!
UPA Boston Conference 439/12/2013
Notas del editor
The pop-up re-presents the same selections on the pageBrowse designs byBusiness or IndustryStylesThe pop-up is automatic – it is unsolicited by the user
Illustrates the principle that people will keep clicking as long as they feel they are making progress toward their goal, and each step is quick & easy.
The progress bar shows where you are in the process, and what steps come next.This progress indicator was tested on our Japanese website, and the translation into English is not quite right.This result supports what we learned from the removing the progress bar.
We tried a new visual design to give more prominence to the progress bar, and to indicate forward motion with the chevrons.
We created an alternate Home page for our returning customers.The content we show is based on clickstream data from the First Time Home page
Larger target and visual hierarchy worked… verb w/ security messaging didn’t.Winner for all locales except Japan. In Japan, this test actually hurt conversion rate with the hypothesis that it was due to the loss of the Clear Cart feature, so we rolled it out to all locales except Japan. More work to understand the impact in that locale.
Testing helps you determine how far to take UX best practices and how to balance your desire to improve KPI’s with a possibly more elegant design.
Design coolness comes at a priceHover effects, but not a clear CTACTA, but no hover effect
Going from 5 steps to 2 steps
We created a “Post Order Configuration” flow that was available immediately after purchase.
But wait… you’re probably going to say that this research is based on sites with a lot of content like news sites, etc… and you’re right. So what does this mean to an ecommerce site with little text overall…? That’s what we’re trying to figure out.
Instead, the right question is, “Will the user’s current knowledge help them understand how to use what I’m designing?” Current knowledge is the knowledge the user has when they approach the design. It’s the sum of all their previous experiences with relevant products and designs.
Instead, the right question is, “Will the user’s current knowledge help them understand how to use what I’m designing?” Current knowledge is the knowledge the user has when they approach the design. It’s the sum of all their previous experiences with relevant products and designs.