The document discusses A/B testing and multivariate testing. It explains how to efficiently test multiple variations of website elements through techniques like orthogonal arrays. This allows testing all pairwise combinations of variables while reducing the total number of tests needed. The document provides an example of reducing tests from 54 combinations to just 9 combinations through an L9 orthogonal array design. It also discusses analyzing test results through visualizing data plots and statistical methods like analysis of variance.
1 of 47
Download to read offline
More Related Content
Intro ab-taguchi
1. Prepared for:
NYC UX + DATA Meetup
March 12, 2014
Pivotal Labs, NewYork
A/B and PairwiseTesting
How I Learned to Stop Worrying and Love
Data-Driven Decisions
Wednesday, March 12, 14
2. About Me
Founded Splitforce in 2013 - Data is
power, and it should be easy to leverage
Marketing for Chinese media company
in Shanghai
Designed experiments and predictive
analytics for ILABS in Montreal
Studied in economics and statistics at
McGill University in Montreal
Wednesday, March 12, 14
5. User Base
Publish two different versions
of your app...
50% sees version B50% sees version A
Wednesday, March 12, 14
6. User Base
...and see which one is driving
desirable user behavior.
Publish two different versions
of your app...
50% sees version B50% sees version A
Wednesday, March 12, 14
13. And the Winner is...
+40%
increase in conversion
rate
2.9 million
additional donators
$60 million
value of additional
donations
Wednesday, March 12, 14
16. Obamalytics
Original Conversion Rate: 8.3%
New Conversion Rate: 11.6%
10 million signups from NewVersion would have been
7.12 million signups with the OriginalVersion
Wednesday, March 12, 14
17. Obamalytics
Original Conversion Rate: 8.3%
New Conversion Rate: 11.6%
10 million signups from NewVersion would have been
7.12 million signups with the OriginalVersion
+2.88 million additional signups
Wednesday, March 12, 14
18. Obamalytics
Original Conversion Rate: 8.3%
New Conversion Rate: 11.6%
10 million signups from NewVersion would have been
7.12 million signups with the OriginalVersion
+2.88 million additional signups
$21 average donation per signup
Wednesday, March 12, 14
19. Obamalytics
Original Conversion Rate: 8.3%
New Conversion Rate: 11.6%
10 million signups from NewVersion would have been
7.12 million signups with the OriginalVersion
+2.88 million additional signups
$21 average donation per signup
Approximately $60 million in additional donations
Wednesday, March 12, 14
21. MultivariateTesting
Every screen has X components (ex: Marilyns hair)
For each, we can test Y variations (ex.: Green)
In total, we have [Y1 x Y2 x Y3] combinations
Wednesday, March 12, 14
22. Costs ofTesting
Risk of false positives (Type I error, saying something
is there when its not)
Need for adequate sample size
Testing presents an opportunity cost
Wednesday, March 12, 14
23. Design of Experiments
Lets say we have four variables:
Header Banner (A, B, C)
Main Copy (1, 2, 3)
Button Color (Cyan, Magenta,Yellow)
Call to Action (Buy!, Check Out)
Wednesday, March 12, 14
24. Design of Experiments
Option 1: Full factorial design -
multiply out for all different
combinations
Wednesday, March 12, 14
25. Design of Experiments
Option 1: Full factorial design -
multiply out for all different
combinations
Example: (3 header banners) x (3 main
copy) x (3 button colors) x (2 CTAs) =
54 combinations
Wednesday, March 12, 14
26. Design of Experiments
Option 1: Full factorial design -
multiply out for all different
combinations
Example: (3 header banners) x (3 main
copy) x (3 button colors) x (2 CTAs) =
54 combinations
Can we get similar information
with fewer tests?
Wednesday, March 12, 14
27. Design of Experiments
Option 2: Orthogonal arrays tests pairs
of combinations instead of all combinations
Wednesday, March 12, 14
28. Design of Experiments
Option 2: Orthogonal arrays tests pairs
of combinations instead of all combinations
Risk: pairing will hide some combinations,
and the effects that paired variables have on
each other
Wednesday, March 12, 14
29. Design of Experiments
Option 2: Orthogonal arrays tests pairs
of combinations instead of all combinations
Risk: pairing will hide some combinations,
and the effects that paired variables have on
each other
Mitigation: pair variables that are unlikely
to in鍖uence each other
Wednesday, March 12, 14
30. L9 Array
Compare any pair of variables across all combinations
and youll see that theyre all represented!
Wednesday, March 12, 14
31. Design of Experiments
Lets say we have four variables:
Header Banner (A, B, C)
Main Copy (1, 2, 3)
Button Color (Cyan, Magenta,Yellow)
Call to Action (Buy!, Check Out)
Wednesday, March 12, 14
32. Design of Experiments
Four variables:
Header Banner (A, B, C)
Main Copy (1, 2, 3)
Button Color (Cyan, Magenta,Yellow)
Call to Action
(Buy, Purchase)
Combo # HB MC BC CTA
1 A 1 Cyan Buy
2 A 2 Magenta Purchase
3 A 3 Yellow
4 B 1 Magenta
5 B 2 Yellow Buy
6 B 3 Cyan Purchase
7 C 1 Yellow Purchase
8 C 2 Cyan
9 C 3 Magenta Buy
Wednesday, March 12, 14
33. Design of Experiments
Four variables:
Header Banner (A, B, C)
Main Copy (1, 2, 3)
Button Color (Cyan, Magenta,Yellow)
Call to Action
(Buy, Purchase)
Combo # HB MC BC CTA
1 A 1 Cyan Buy
2 A 2 Magenta Purchase
3 A 3 Yellow Buy
4 B 1 Magenta Purchase
5 B 2 Yellow Buy
6 B 3 Cyan Purchase
7 C 1 Yellow Purchase
8 C 2 Cyan Buy
9 C 3 Magenta Buy
Wednesday, March 12, 14
34. Design of Experiments
Four variables:
Header Banner (A, B, C)
Main Copy (1, 2, 3)
Button Color (Cyan, Magenta,Yellow)
Call to Action
(Buy, Purchase)
Combo # HB MC BC CTA
1 A 1 Cyan Buy
2 A 2 Magenta Purchase
3 A 3 Yellow Buy
4 B 1 Magenta Purchase
5 B 2 Yellow Buy
6 B 3 Cyan Purchase
7 C 1 Yellow Purchase
8 C 2 Cyan Buy
9 C 3 Magenta Buy
Weve reduced need to
collect data on 54
combinations to just 9
(6x ef鍖ciency increase)
Wednesday, March 12, 14
35. FROM 54 COMBINATIONS
A1CyanBuy,
A1CyanPurchase,
A1MagentaBuy,
A1MagentaPurchase,
A1YellowBuy,
A1YellowPurchase,
A2CyanBuy,
A2CyanPurchase,
A 2 M a g e n t a B uy,
A 2 M a g e n t a P u rc h a s e ,
A 2 Ye l l ow B uy,
A2YellowPurchase,
A3CyanBuy,
A3CyanPurchase,
A3MagentaBuy,
A3MagentaPurchase,
A3YellowBuy,
A3YellowPurchase,
B1CyanBuy,
B1CyanPurchase,
B1MagentaBuy,
B1MagentaPurchase,
B1YellowBuy,
B1YellowPurchase,
B2CyanBuy,
B2CyanPurchase,
B 2 M a g e n t a B uy,
B 2 M a g e n t a P u rc h a s e ,
B 2 Ye l l ow B uy,
B2YellowPurchase,
B3CyanBuy,
B3CyanPurchase,
B3MagentaBuy,
B3MagentaPurchase,
B3YellowBuy,
B3YellowPurchase,
C1CyanBuy,
C1CyanPurchase,
C1MagentaBuy,
C1MagentaPurchase,
C1YellowBuy,
C1YellowPurchase,
C2CyanBuy,
C2CyanPurchase,
C 2 M a g e n t a B uy,
C 2 M a g e n t a P u rc h a s e ,
C 2 Ye l l ow B uy,
C2YellowPurchase,
C3CyanBuy,
C3CyanPurchase,
C3MagentaBuy,
C3MagentaPurchase,
C3YellowBuy,
C3YellowPurchase
Wednesday, March 12, 14
36. TO JUST 9 (+6X EFFICIENCY)
A1CyanBuy,
A1CyanPurchase,
A1MagentaBuy,
A1MagentaPurchase,
A1YellowBuy,
A1YellowPurchase,
A2CyanBuy,
A2CyanPurchase,
A 2 M a g e n t a B uy,
A 2 M a g e n t a P u rc h a s e ,
A 2 Ye l l ow B uy,
A2YellowPurchase,
A3CyanBuy,
A3CyanPurchase,
A3MagentaBuy,
A3MagentaPurchase,
A3YellowBuy,
A3YellowPurchase,
B1CyanBuy,
B1CyanPurchase,
B1MagentaBuy,
B1MagentaPurchase,
B1YellowBuy,
B1YellowPurchase,
B2CyanBuy,
B2CyanPurchase,
B 2 M a g e n t a B uy,
B 2 M a g e n t a P u rc h a s e ,
B 2 Ye l l ow B uy,
B2YellowPurchase,
B3CyanBuy,
B3CyanPurchase,
B3MagentaBuy,
B3MagentaPurchase,
B3YellowBuy,
B3YellowPurchase,
C1CyanBuy,
C1CyanPurchase,
C1MagentaBuy,
C1MagentaPurchase,
C1YellowBuy,
C1YellowPurchase,
C2CyanBuy,
C2CyanPurchase,
C 2 M a g e n t a B uy,
C 2 M a g e n t a P u rc h a s e ,
C 2 Ye l l ow B uy,
C2YellowPurchase,
C3CyanBuy,
C3CyanPurchase,
C3MagentaBuy,
C3MagentaPurchase,
C3YellowBuy,
C3YellowPurchase
Wednesday, March 12, 14
37. Design of Experiments
Where do orthogonal arrays come from?
Derived by hand (like playing Sudoku!)
Look them up (U Michigan, UYork, Hexawise.com)
Wednesday, March 12, 14
38. Design of Experiments
Where do orthogonal arrays come from?
Derived by hand (like playing Sudoku!)
Look them up (U Michigan, UYork, Hexawise.com)
How to choose a design?
Number of variables
Number of states for each variable
Wednesday, March 12, 14
39. Design of Experiments
Where do orthogonal arrays come from?
Derived by hand (like playing Sudoku!)
Look them up (U Michigan, UYork, Hexawise.com)
How to choose a design?
Number of variables
Number of states for each variable
How to analyze results?
Plot data,Analysis ofVariance (ANOVA), binning
Wednesday, March 12, 14
40. Analyzing Results
Plot data and look at it
Some things you dont need statistics to tell you, its just there
Your eye is a pretty good analysis tool
Wednesday, March 12, 14
41. Analyzing Results
Plot data and look at it
Some things you dont need statistics to tell you, its just there
Your eye is a pretty good analysis tool
Analysis ofVariance (ANOVA)
One-way ANOVAs to 鍖nd in鍖uence of a one variable on the
result (assume that other variables have minimal in鍖uence)
Two-way ANOVAs to 鍖nd in鍖uence of two variables on
result at once
Wednesday, March 12, 14
42. Analyzing Results
Plot data and look at it
Some things you dont need statistics to tell you, its just there
Your eye is a pretty good analysis tool
Analysis ofVariance (ANOVA)
One-way ANOVAs to 鍖nd in鍖uence of a one variable on the
result (assume that other variables have minimal in鍖uence)
Two-way ANOVAs to 鍖nd in鍖uence of two variables on
result at once
Binning
Group combinations based on results (high vs. low)
How many Header Banner As have high result? low result?
Wednesday, March 12, 14
43. Analyzing Results
Plot data and look at it
Some things you dont need statistics to tell you, its just there
Your eye is a pretty good analysis tool
Analysis ofVariance (ANOVA)
One-way ANOVAs to 鍖nd in鍖uence of a one variable on the
result (assume that other variables have minimal in鍖uence)
Two-way ANOVAs to 鍖nd in鍖uence of two variables on
result at once
Binning
Group combinations based on results (high vs. low)
How many Header Banner As have high result? low result?
Takeaway: You can extrapolate data from a subset of combinations
to make a conclusion about a full factorial set
Wednesday, March 12, 14
44. Design of Experiments
Can get pretty complex, but super ef鍖cient!
L36 array - reducing ~94 million combinations to 36
Wednesday, March 12, 14
45. Comparison of
A/BTesting Platforms
Google Analytics Optimizely Splitforce
Platform
Web / mWeb X X
Platform
Native Mobile X
A/BTesting X X
Experiment
Design
Multivariate X X
Automation X X
Other
In-Browser Editor X X
Other
Consulting X X
Wednesday, March 12, 14
46. In-House vs.Agency
In-House Agency
Pros
Lower initial costs
More control over testing process
Better understanding of business
objectives
No need for internal resources
Faster results as agency provides specialized
expertise
Learn best practices and accelerate internal
competency
Cons
Long time to build expertise from
scratch
Longer time to start achieving great test
results
Higher initial costs
Less understanding of complexities /
nuances of your business
Less control over testing
Wednesday, March 12, 14
47. ThankYou!
For more information:
Zac Aghion, CEO & Co-Founder
zac@splitforce.com
China: (+86)1592-1631-924
USA: (+1)617-750-6684
www.splitforce.com
Wednesday, March 12, 14