The document discusses A/B testing and optimization. It covers why companies do A/B testing, common mistakes in testing like poor data quality and not following statistics properly, and how to properly conduct tests. Key recommendations include testing everything, prioritizing high impact tests, understanding customer problems by analyzing metrics and flows, removing friction from the user experience, and building a data-driven culture.
A/B testing, optimization and results analysis by Mariia Bocheva, ATD'18
1. By Mariia Bocheva, OWOX BI
A / B testing,
Optimization and Results Analysis
2. שלוםלכולם
Mariia Bocheva
Head of Sales and Marketing
5 years @ OWOX
Ecommerce, Analytics, BI, CRO
m.bocheva@owox.com
@BochevaMariia
@mariia.bocheva
27. 1. Is our GTM/GA set up properly?
2. Do we filter out unnecessary traffic?
3. What do we capture?
4. Where do we store it?
5. How do we track refunds and ROPO-effect?
BS in – BS out
29. A/B Testing Ad Campaigns
Campaign A
Campaign B
150 Clicks
100 Clicks
30. A/B Testing Ad Campaigns
Campaign A
Campaign B
150 Clicks
100 Clicks
$1,000
$2,000
31. A/B Testing Ad Campaigns
Campaign A
Campaign B
150 Clicks
100 Clicks
$1,000
$2,000
-$100 in Returns -$500
32. A/B Testing Ad Campaigns
Campaign A
Campaign B
150 Clicks
100 Clicks
$1,000
$2,000
-$100 in Returns -$500
33. 1. Is our GTM/GA set up properly?
2. Do we filter out unnecessary traffic?
3. What do we capture?
4. Where do we store it?
5. How do we track refunds and ROPO-effect?
6. Do we trust the data we have?
BS in – BS out
40. HIPPO Opinion
“If we have data, let’s look at
data. If all we have are opinions,
let’s go with mine.”
– Jim Barksdale, former Netscape CEO
41. 1. Quality of data
2. Not doing your statistics right
3. Big Shot A/B testing
4. HIPPO opinion
5. Optimizing for a micro conversion
What Could Go Wrong?
42. What Could Go Wrong?
1. Quality of data
2. Not doing your statistics right
3. Big Shot A/B testing
4. HIPPO opinion
5. Optimizing for a micro conversion
6. Assumed reproducibility
43. 1. Quality of data
2. Not doing your statistics right
3. Big Shot A/B testing
4. HIPPO opinion
5. Optimizing for a micro conversion
6. Assumed reproducibility
7. Copying your competitors
What Could Go Wrong?
44. “Stop Copying Your Competitors:
They Don’t Know What They’re
Doing Either”
– by Peep Laja
45. 1. Quality of data
2. Not doing your statistics right
3. Big Shot A/B testing
4. HIPPO opinion
5. Optimizing for a micro conversion
6. Assumed reproducibility
7. Copying your competitors
8. Unclear hypothesis and expected results
What Could Go Wrong?
46. Hypothesis Kit
1. Because we saw (data/feedback)
2. We expect that (change) will cause (impact)
3. We’ll measure this using (data metric)
53. 1. Test everything
2. Test atomically
3. Prioritize testing
4. Focus on the right thing
5. Question things you don’t understand
Doing it Right
6. Build data-driven culture
54. “You can have data without
information, but you cannot have
information without data.”
– Daniel Keys Moran
55. 1. Test everything
2. Test atomically
3. Prioritize testing
4. Focus on the right thing
5. Question things you don’t understand
6. Build data-driven culture
Doing it Right
7. Automate testing
8. Understand cross device
67. 1. Who is this page for (age, gender, geo)?
2. Where do customers come from?
3. What do we want our customers to do next and what they actually
do?
Understand the Problem
69. 1. Who is this page for (age, gender, geo)?
2. Where do customers come from?
3. What do we want our customers to do next and what they actually
do?
4. What are the key metrics (conversion rate, bounce rate, time on
page, etc.)?
5. What’s the mix of devices and resolutions?
Understand the Problem
72. 1. Who is this page for (age, gender, geo)?
2. Where do customers come from?
3. What do we want our customers to do next and what they actually
do?
4. What are the key metrics (conversion rate, bounce rate, time on
page, etc.)?
5. What’s the mix of devices and resolutions?
Understand the Problem
6. How does the page look like on those devices?
7. Is the page load time okay?
8. What’s going on with JS errors
82. 1. Eliminate friction
2. Understand the problem
3. Don’t go with HIPPO
4. Make sure you have enough data
5. Build the culture
6. Prove that your/others ideas were wrong
7. Learn from the failure
TL;DR