Speakers
Successful Testing
with a Lean Team
Tim Kotchetov, Data Analysis & Business Intelligence, Seattle Times
Kate Nichol, Customer Success Manager, Optimizely
Today’s Speakers
Tim
Kotchetov
Kate
Nichol
Housekeeping
• We’re recording!
• Slides and recording will be
emailed to you after the
webinar is complete
• There will be time for
questions at the end
Agenda
• Staffing for Experimentation Success
• Experimentation at The Seattle Times
• Real-World Tests
• Overcoming Challenges
• Q&A
Staffing for
Experimentation
Success
‘Recommended’ Experimentation
Program
Successful Testing Skills
● HTML and CSS
● jQuery and JavaScript
● SEM and SEO
● Statistics
● Project Management
● UX Design
● Analytics
● Copy
● Communication
It’s still possible to achieve
success with a smaller team.
Experimentation at
The Seattle Times
The growing importance
of SeattleTimes.com
● Launched website in early 2000s with
unlimited access to all online content
● In 2013, SeattleTimes.com introduced a
paywall which allows 7 free articles/month
● 3 upsell products:
○ Digital Only
○ Digital Only + Sunday Print
○ Digital + Daily Print
Breaking Through the Paywall
Maximize revenue by
increasing number of
subscribers
State of the Experimentation Program
● A/B Testing since March 2016
● Goal to increase conversions, subscriber revenue, or
conversions that indirectly leads to subscriptions (such as
newsletters)
● Run approx. 1 a/b test per month
● Integrate with Google Analytics and Google BigQuery
● Approx. 1.5 persons dedicated to A/B testing program
● Testing ideas sourced from across circulation, marketing
team and across the org
Resources & Staff
● The Seattle Times Business Intelligence
team is made up of two analysts and a
manager
● Primary responsibility is to provide any
analytical needs to various departments
they cannot get on their own
● One analyst is responsible for
coordinating implementation of and
tracking of A/B tests and results
Business
Intelligence
Circulation Advertisement Newsroom
Real-World Tests
#1: Newsletter Subscribe Widget Test
● Test whether wording on widget email submit
button prevented people from subscribing to
newsletter
● Conducted 11-day A/B test of two versions:
○ Original: “Subscribe”
○ Variation: “Sign Up”
● Results: Variation generated +4.3%
more signups
Original
Variation
#2: Product Pricing Test
● We offer 3 products:
○ Digital Access only
○ Digital Access along with daily print delivery
○ Digital Access plus print delivery on Sundays (our
most profitable)
● Hypothesis: Would a decrease in price of our most
profitable product increase the number of purchasers
enough to offset this and increase overall revenue?
● Partnered with BI team member and Director of Circulation
to create test model accounting for ad revenue, audience
revenue and retention
● Used model to track test results over 9 weeks
Variation
Original
#2: Product Pricing Test - Results
We proved our hypothesis:
● Digital Access only purchases
dropped 10 percentage points
● Digital Access + daily print
purchases stayed consistent
● Digital Access + Sunday print
purchases increased 8%
● Overall revenue increased by
8%
Test whether newly designed landing page with same wording and messaging as previous
version would not drive down subscriptions and possibly increase them
Original Variation
#3: Paywall Landing Page Design Test
#3: Landing Page Design Test - Results
● The new landing page design was successful!
● Generated 8% more overall subscription starts
#4: Yearly Subscription Offer Test
● Hypothesis: Would overall revenue increase if we offered a yearly subscription for
‘Digital Only ‘at a 25% annual discount?
● For 8 weeks, tested this offer side-by-side with original monthly subscription
Original Variant
#4: Yearly Subscription Offer Test - Results
● Annual subscription offering is not profitable
○ Overall conversions for variant were 9% lower than original monthly offer
○ Digital conversions for variant were 10% lower
○ Print conversions for variant were 8% lower
#5 - Paywall Meter & Landing Page Copy Test
● Users can reach 7 articles per
month before hitting paywall
● Readers are shown periodic
pop-up messages giving them
article count
● Hypothesis: Will different
messaging encourage more
subscriptions?
Variant
#5 - Paywall & Landing Page Copy Test - Results
● Wording changes made a difference
○ Overall conversions generated 14% more subscriptions
○ Conversion for Digital Access increased 24%; virtually unchanged for Digital +
Print Sunday/Daily
○ Those clicking on meter pop-up messaging instead of waiting to hit paywall
25% more likely to subscribe
○ Those who hit paywall page 28% more likely to subscribe
What’s Next?
● A/B Test the # of monthly articles a user can access before hitting the paywall
● A/B Test placement of section blocks on the home page
● A/B Test position of newsletter signup widgets
Overcoming
Testing Challenges
Developer Resource Constraints
Challenge
● Many A/B tests required developer involvement/coding
● Developers are stretched thin
Solution
● Developers dedicate 2-3 hours to each A/B test
● Analyst can handle organizational, logistical, and measurement aspects
Statistical Significance
Challenge
● Industry online conversion rates are low
● Hard to generate statistically significant results
Solution
● Focus on data consistency
● Calculate weekly rate of conversions
● Throttle number of users to focus on test pages
Implementing Changes
Challenge
● Limited developer resources delay website changes
Solution
● Make change in Optimizely as stop-gap measure
Tips & Tricks for Testing with a Lean Team
● For your first test, try something simple to get you going
● Source ideas from stakeholders across your organization
● Important to establish relationship with IT/developers
○ Use developers’ time efficiently and regular meet to discuss priorities
● Integrate with Google Analytics to show impact to audience segments that are
important to the business
● Regular communication about performance to ensure test results are top of mind and
having an implementation plan for key findings
Questions?
Thank You!

Successful Testing with a Lean Team

  • 1.
    Speakers Successful Testing with aLean Team Tim Kotchetov, Data Analysis & Business Intelligence, Seattle Times Kate Nichol, Customer Success Manager, Optimizely
  • 2.
  • 3.
    Housekeeping • We’re recording! •Slides and recording will be emailed to you after the webinar is complete • There will be time for questions at the end
  • 4.
    Agenda • Staffing forExperimentation Success • Experimentation at The Seattle Times • Real-World Tests • Overcoming Challenges • Q&A
  • 5.
  • 6.
    ‘Recommended’ Experimentation Program Successful TestingSkills ● HTML and CSS ● jQuery and JavaScript ● SEM and SEO ● Statistics ● Project Management ● UX Design ● Analytics ● Copy ● Communication
  • 7.
    It’s still possibleto achieve success with a smaller team.
  • 8.
  • 9.
    The growing importance ofSeattleTimes.com ● Launched website in early 2000s with unlimited access to all online content ● In 2013, SeattleTimes.com introduced a paywall which allows 7 free articles/month ● 3 upsell products: ○ Digital Only ○ Digital Only + Sunday Print ○ Digital + Daily Print
  • 10.
    Breaking Through thePaywall Maximize revenue by increasing number of subscribers
  • 11.
    State of theExperimentation Program ● A/B Testing since March 2016 ● Goal to increase conversions, subscriber revenue, or conversions that indirectly leads to subscriptions (such as newsletters) ● Run approx. 1 a/b test per month ● Integrate with Google Analytics and Google BigQuery ● Approx. 1.5 persons dedicated to A/B testing program ● Testing ideas sourced from across circulation, marketing team and across the org
  • 12.
    Resources & Staff ●The Seattle Times Business Intelligence team is made up of two analysts and a manager ● Primary responsibility is to provide any analytical needs to various departments they cannot get on their own ● One analyst is responsible for coordinating implementation of and tracking of A/B tests and results Business Intelligence Circulation Advertisement Newsroom
  • 13.
  • 14.
    #1: Newsletter SubscribeWidget Test ● Test whether wording on widget email submit button prevented people from subscribing to newsletter ● Conducted 11-day A/B test of two versions: ○ Original: “Subscribe” ○ Variation: “Sign Up” ● Results: Variation generated +4.3% more signups Original Variation
  • 15.
    #2: Product PricingTest ● We offer 3 products: ○ Digital Access only ○ Digital Access along with daily print delivery ○ Digital Access plus print delivery on Sundays (our most profitable) ● Hypothesis: Would a decrease in price of our most profitable product increase the number of purchasers enough to offset this and increase overall revenue? ● Partnered with BI team member and Director of Circulation to create test model accounting for ad revenue, audience revenue and retention ● Used model to track test results over 9 weeks Variation Original
  • 16.
    #2: Product PricingTest - Results We proved our hypothesis: ● Digital Access only purchases dropped 10 percentage points ● Digital Access + daily print purchases stayed consistent ● Digital Access + Sunday print purchases increased 8% ● Overall revenue increased by 8%
  • 17.
    Test whether newlydesigned landing page with same wording and messaging as previous version would not drive down subscriptions and possibly increase them Original Variation #3: Paywall Landing Page Design Test
  • 18.
    #3: Landing PageDesign Test - Results ● The new landing page design was successful! ● Generated 8% more overall subscription starts
  • 19.
    #4: Yearly SubscriptionOffer Test ● Hypothesis: Would overall revenue increase if we offered a yearly subscription for ‘Digital Only ‘at a 25% annual discount? ● For 8 weeks, tested this offer side-by-side with original monthly subscription Original Variant
  • 20.
    #4: Yearly SubscriptionOffer Test - Results ● Annual subscription offering is not profitable ○ Overall conversions for variant were 9% lower than original monthly offer ○ Digital conversions for variant were 10% lower ○ Print conversions for variant were 8% lower
  • 21.
    #5 - PaywallMeter & Landing Page Copy Test ● Users can reach 7 articles per month before hitting paywall ● Readers are shown periodic pop-up messages giving them article count ● Hypothesis: Will different messaging encourage more subscriptions? Variant
  • 22.
    #5 - Paywall& Landing Page Copy Test - Results ● Wording changes made a difference ○ Overall conversions generated 14% more subscriptions ○ Conversion for Digital Access increased 24%; virtually unchanged for Digital + Print Sunday/Daily ○ Those clicking on meter pop-up messaging instead of waiting to hit paywall 25% more likely to subscribe ○ Those who hit paywall page 28% more likely to subscribe
  • 23.
    What’s Next? ● A/BTest the # of monthly articles a user can access before hitting the paywall ● A/B Test placement of section blocks on the home page ● A/B Test position of newsletter signup widgets
  • 24.
  • 25.
    Developer Resource Constraints Challenge ●Many A/B tests required developer involvement/coding ● Developers are stretched thin Solution ● Developers dedicate 2-3 hours to each A/B test ● Analyst can handle organizational, logistical, and measurement aspects
  • 26.
    Statistical Significance Challenge ● Industryonline conversion rates are low ● Hard to generate statistically significant results Solution ● Focus on data consistency ● Calculate weekly rate of conversions ● Throttle number of users to focus on test pages
  • 27.
    Implementing Changes Challenge ● Limiteddeveloper resources delay website changes Solution ● Make change in Optimizely as stop-gap measure
  • 28.
    Tips & Tricksfor Testing with a Lean Team ● For your first test, try something simple to get you going ● Source ideas from stakeholders across your organization ● Important to establish relationship with IT/developers ○ Use developers’ time efficiently and regular meet to discuss priorities ● Integrate with Google Analytics to show impact to audience segments that are important to the business ● Regular communication about performance to ensure test results are top of mind and having an implementation plan for key findings
  • 29.
  • 30.