In the online world, sell-through is known as conversion rate optimization. For almost every online business, conversion rate is the single highest leverage point for increasing sales and return on marketing spend.
In this presentation, Zac Aghion goes through the basics of conversion rate optimization. From A/B to multivariate testing and beyond, the presentation reviews the theory, case studies and best practices in this critical field of online marketing.
The presentation begins with an overview of the basic web analytics terms and concepts needed to grasp the value of conversion rate optimization. It continues with an example of
2. Some Basic Web Analytics
Visitors – Number of people that visit your
website
Conversions – Number of visitors that complete a
goal (purchase, sign up, download, click, etc.)
Conversion Rate = Conversions ÷ Visitors
Example:
200 Visitors and 5 Purchases
Conversion Rate = 5 ÷ 200 = 2.5%
12. Multivariate Testing Overview
Every landing page has X elements (MM’s hair)
For each element, we test Y variations (Neon)
In total, we have (Y1× Y2 × Yx) LP combinations
18. +40.6% CR = +$60 MM
Original CR 8.26%
New CR 11.6% (+40.6%)
10 million signups from New (would be closer to
7,120,000 signups with Original)
+2,880,000 additional signups
$21 average donation / signup
+$60,000,000 in additional donations
20. Confidence Intervals
Confidence intervals indicate the statistical
significance of an observed conversion rate
95% CI = Observed CR +/- (1.96 × Standard Error)
Standard Error = √ (p × (1-p) ÷ n)
Common Pitfalls
Low # trials = non-significant results
High # trials = significant results
General balance between speed and reliability
My website has less traffic than Obama
But I still want +40% CR
How??
21. Enter Genichi Taguchi
Japanese statistician
(1924 – 2012)
Taguchi Methods – fractional
factorial experimental designs
Basic Concept – data collected
for a specific subset of
combinations can be
extrapolated to determine the
best performer of a full factorial
set
22. Orthogonal Arrays
Taguchi arrays are used to determine what
subset of combinations need to be tested to
extrapolate results
Most Basic Taguchi Array, L4:
23. L8 Array: Testing Efficiency +32x
7 page elements, 2 variations for each
2 × 2 × 2 × 2 × 2 × 2 × 2 = 256 LP combinations
Data extrapolated from specific subset of 8
combinations determines best of 256
24. In-House or Agency?
In-House Agency
Pros
Lower initial costs
More control over the
testing process
More relevant
advanced goal tests
No need for internal resources
Faster results as agency
provides specialized expertise
Learn best practices and gain
experience
Cons
Long time to build
expertise from scratch
Longer time to start
achieving great test
results
Higher initial costs
Less understanding of
complex details of your
business goals and industry
Less control over testing