Upload
sara-hoffman
View
107
Download
1
Embed Size (px)
Citation preview
Jen BolandAnalytics and Optimization Lead
Sara HoffmanWeb Development Manager
Brian RogelAssociate Director of SEO and SEM
About us
#Bridge16
Avon Breast Cancer Crusade
Beaconfire REDBeaconfire RED
CONFIDENTIAL 3
Agenda• Creating a plan• Test strategy• Test setup• Things to do while your are testing• How to evaluate results• What to do after testing
Define Website Pain Points1) Determine goals
2) List problem areas
3) Rank list based on costs and benefits
#Bridge16
Review Metrics of the Proposed Test
Key Metrics: • Unique Visits• Pageviews• Conversion Rate• Average Gift
#Bridge16
Determine Metrics that can be Tested
#Bridge16
TOP-FUNNEL (Landing Pages)
MID-FUNNEL(Transition Pages)
BOTTOM-FUNNEL(Conversion Forms)
Use
r’s Jo
urne
yDecision Tips:• Top-funnel tests run faster
than bottom-funnel tests• Testing conversions is not
always possible • Make sure the target
sample size can be reached in the proposed test duration
Determining Sample Size
Sample Size Tips:• Use sample size
calculators• Run tests until the
sample size is reached, not until statistical significance is reached
#Bridge16
A minimum detectable effect (MDE) is the smallest true effect that has a ‘good chance’ of being found to be statistically significant.
Test Duration: MinimumMinimum Test Duration Tips:• Run tests for full week cycles• One week minimum is possible for websites
with exceptionally high traffic• A two week minimum is preferred for most
websites
#Bridge16
Test Duration: MaximumMaximum Test Duration Tips:• Clearing cookies can cause sample pollution
o Lifehacker Poll: 72.35% of people clear cookies or use private browsing
o Microsoft Report: studies show 30% to 40% of people clear cookies monthly
• Avoid running tests longer than four weeks
#Bridge16
Test Duration and Skewed Results
#Bridge16
Avoid running tests during holidays or seasonal anomalies(Ex: End-of-year and non-end-of-year traffic acts very differently)
Test Duration and Sample Size Recap
#Bridge16
• Determine target sample size • Run tests for full week cycles• 2 to 4 week test durations are best• Run tests until the target sample size is reached,
not until statistical significance is reached• Be aware of holidays and seasonal anomalies
#Bridge16
Where are you going?Make a map
January February March April May June JulyHomepage User Testing
Homepage Hero Slider A/B Homepage Static Hero
Build Hero Ask String High-LowSurvey - Legal
Survey
Rebuild overlay
Pre Year-End Overlay POC
possible extension
Action TY Story TitleSuggested Amount
inahat.co/bridge-roadmap
CONFIDENTIAL 25
TestVariation Name
Test Length Traffic % # of Visitors VWO Type Test Location
ETA for Results
Hero Homepage Donation Ask No Test 7 60% 587 ABCDCMS Homepage 1/12/2015
Hero Homepage Donation Ask Control 7 13% 1401 ABCDCMS Homepage 1/12/2015
Hero Homepage Donation Ask Var1 7 13% 1458 ABCDCMS Homepage 1/12/2015
Hero Homepage Donation Ask Var2 7 13% 1464 ABCDCMS Homepage 1/12/2015
Grey Overlay Homepage Real Control 7 50% 883 Split testCMS Homepage 1/20/2015
Grey Overlay Homepage Grey Overlay 7 50% 857 Split testCMS Homepage 1/20/2015
Hamburger Control 14 50% 1828 ABC
Homepage and Directory 6/8/2015
Hamburger Button 14 25% 920 ABC
Homepage and Directory 6/8/2015
Hamburger Left Logo 14 25% 943 ABC
Homepage and Directory 6/8/2015
Hypothesis Creation
#Bridge16
Sections of a Hypothesis:1) Statement of Problem2) Proposed Solution3) Expected Results
Hypothesis Template
#Bridge16
Based on <data discovered> we believe <website issue> is causing <undesirable result>. We believe <change to be tested> will help resolve <website issue> and cause <desired result>.
Writing your Hypothesis
#Bridge16
Example Hypothesis:Based on a 31% below average conversion rate for paid search traffic we believe a lack of information on donation forms is causing a high drop-off rate. We believe sending people to an information rich splash page before the donation form will help resolve the high drop-off rate and cause an increase in conversion rate.
Test Setup • Preserve Query
Parameters• Link with Analytics
Tools
#Bridge16
https://testsite.org?Donation2?df_id=12362&12362.donation=form1&utm_source=newsletter&utm_medium=email&utm_content=link1&utm_campaign=fall
Beware of Peeping Toms
• Browser Tools like WAPPALYZER, dataSlayer, Pessimezely or WASP can expose your test and variant names!
#Bridge16
Micro Conversions
#Bridge16
Control
Variant
+9.7%Website sign-in clicks+4.9%Credit card sign-in clicks
+21%Registration clicks
Tools to Help with Testing – Handout!Your Good Friend – Excel!Percent Change (Current Value – Previous Value) / Previous Value
#Bridge16
“If you use p=0.05 to suggest that you have made a discovery, you will be wrong at least 30% of the time.” - David Colquhoun
Analyzing Data: What’s a Win?
#Bridge16
Checklist: Low P-Value High Power Test Reaches Target
Sample Size Test is Not an Industry
Outlier
Validate – Iterate!
inahat.co/bridge-validate
Hypothesis
Create / Iterate
Continuous Testing
Measure Results
#Bridge16
#Bridge16
Modify Your MapJanuary February March April May June JulyHomepage User Testing
Homepage Hero Slider A/B Homepage Static Hero
Build Hero Ask String High-Low Paid Search Ask
DRTV Ask
Survey - Legal
Survey
Rebuild overlay
Pre Year-End Overlay POC
possible extension
Paid Search Overlay
Action TY Story TitleSuggested Amount
inahat.co/bridge-roadmap
Sharing Lessons Across Channels
#Bridge16
Using Google AdWords to test how effective different copy variations are at improving click-through-rates.
Control Variants
+25.2%CTR Improvement
-6.9%CTR Reduction
Sharing Lessons Across Channels
#Bridge16
The difference in “Don’t miss” and “Don’t miss out on” made a 34.4% difference in click-through-rates. The test was significant at 99% with 99.93% statistical power.
Variant+34.4%CTR Improvement
Control
Cross-Channel Implementations
#Bridge16
Copy tests help determine the best messaging for your audience. Improvements discovered can be leveraged in many channels:
Meta Content (Organic)
Ad Copy (Paid)
• Email• Direct Mail• DRTV
• Social Media• Organic Meta Content• Display Ad Copy