A/B testing helps organisations identify and evaluate potential product improvements using data and statistics, which eliminates the guesswork and helps them understand what makes users tick. In this talk, I’ll explain what A/B testing is and give examples of how it can be used. Then I’ll share lessons learned from 10 years of AB tests on millions of users and some of the world's most popular mobile games: what are prerequisites for a successful A/B test, A/B test process that ensure actionable results, how A/B test infrastructure speeds up implementation and analysis, and how to avoid the common pitfalls.
7. What is A/B Testing
Variant A
Variant B
data
Users Randomly
assigned into
2 groups
Each group uses
one variant
Observe, measure,
and record data
Use statistics to
compare variants
14. Data driven decisions based on
statistical methodology to:
● Improve product
● Improve monetisation
● Learn
Why?
revenue
game difficulty
or
rewards generosity
27. Start with users’ or business problem, not just a test idea!
Who:
● Controlling
● Product
● Users
● Analytics
Define metric(s) to be improved
1
.Identify problems
28. Select ideas by estimating:
● impact
● test power
● cost
● needed time
2. A/B test ideas
29. 2. Select A/B test: estimate impact
$
$/user
# users
% viewers $/ad
ads/user
banner
$/user
installs/day
retention
by DSI
users
by DSI
interstitial
$/user
video
$/user
IAP
$/user
39. 3. BE configurable A/B test
ID Price Test ID black list Test ID white list
default 90 prices_B
prices_C
cheap_prices 50 prices_B
expensive_prices 200 prices_C