Brian Gardner - Director, Product Management
Ilya Izrailevsky - Engineering Manager
Implementing analytic
and optimization tools
on your website or app
WiFi: QBConnect Password: not required#QBConnect
2
Today’s Speakers
Brian Gardner
Director, Product Management
@brianwgardner
Ilya Izrailevsky
Engineering Manager
@iizrailevsky
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Introduction & Overview
Understand the landscape analytic and optimization tools
How to implement an A/B test
Gain the ability to make informed decisions about what in your
product is working (or not)
Agenda
Introduction & Overview
Understand the landscape analytic and optimization tools
How to implement an A/B test
Gain the ability to make informed decisions about what in your
product is working (or not)
Agenda
5
“The sooner you run an
experiment, the sooner you
are likely to find a surprise,
and the surprise is the market
speaking to you, telling you
something you didn't know.
Several of our businesses
here came out of surprises.”
– Scott Cook, Intuit founder
& chairman
Intuit’s tradition of experimentation
Deep
Customer
Empathy
Go Broad to
Narrow
Rapid
Experiments
with
Customers
Customer
Delight
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
6
What is an A/B test?
BA
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Why A/B Test?
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Discover which
experiences
delight your
customer
Accelerate your
learning loop
Make faster,
better decisions
with data
8
Who We Serve
9
Introducing:
WASABI
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
10
First time use test: QuickBooks Self Employed
Test: Complete tax profile first Control: Connect account first
In the test cell, we see:
• Significant increase in tax profile completion
• Slight increase in categorizing a transaction from a card
• No change to connecting a bank
• No change to categorizing on the transactions page
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
11
First time use test: QuickBooks Self Employed
Test: Complete tax profile first Control: Connect account first
In the test cell, we see:
• Significant increase in tax profile completion: +20%
• Slight increase in categorizing a transaction from a card: +6%
• No change to connecting a bank: Flat
• No change to categorizing on the transactions page: Flat
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Introduction & Overview
Understand the landscape analytic and optimization tools
How to implement an A/B test
Gain the ability to make informed decisions about what in your
product is working (or not)
Agenda
13
• Marketing vs Product (changing content & UI vs features or back-end)
• How much engineering is involved?
• Do you have sensitive data?
• What platforms and devices do you need to test across?
• Cost
• Analytics provided / options
• Efficient use of traffic
• Performance
• Ease of use
How do you choose a tool for A/B testing?
What are the dimensions you should think about when choosing A solution?
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
14
Landscape of Popular Options
• WYSIWIG
• Minimal Engineering
• Marketing focused
• Developer focused
• Flexible engineering effort
• Own your own data
• Customized
• Ongoing maintenance
• Own your own data
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
COMMERCIAL OPEN SOURCE IN-HOUSE
WASABI
Introduction & Overview
Understand the landscape analytic and optimization tools
How to implement an A/B test
Gain the ability to make informed decisions about what in your
product is working (or not)
Agenda
16
The A/B Testing Lifecycle: Iterate to Awesomeness
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Step 1:
Plan an experiment
Step 2:
Develop a feature to test
Step 3:
Instrument your application with Wasabi
API Calls
Step 4:
Create and run the experiment
using the Wasabi UI
Step 5:
Deploy and track results
quickly
17
• Is this user already assigned for this experiment?
• Does the user qualify for the experiment? (Segmentation Rules)
• Is the user already in a conflicting experiment? (Mutual Exclusion)
• Randomization
• Check vs. Sampling Percentage
• Assign to an experience (Bucket)
• Return assignment
How does user assignment actually work?
What should I show this user when a WASABI test is active?
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Demo
19
Architecture: behind the scenes, here’s how it’s built
Web tier
Admin UI API API
Segmentation
Mutual Exclusion
Sampling
Bucket Dice Roll
Application tier
Cassandra MySQLData tier
Authentication
User Directory
Logging
Metrics
Pluggable
Components
Bucket
assignments
Events / analytics
Kafka
Stream
HDFS Spark
Vertica S3
Pluggable
Components
Wasabi A/B Testing Service
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Introduction & Overview
Understand the landscape analytic and optimization tools
How to implement an A/B test
Gain the ability to make informed decisions about what in your
product is working (or not)
Agenda
21
Best practices and considerations for a good A/B test
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Get started! (Don’t let the perfect be the enemy of the good)1
Think about what you’ll learn from a “failed” test in advance.
Be skeptical – don’t take your results at face value
• Track all your key metrics – check for unintended consequences
• Think about sources of bias, assumptions: timing, infrastructure, etc.
Don’t stop at one round of experiments!
2
3
4
5
Don’t forget qualitative research as well
22
What Should You Take Away?
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Everyone can A/B test!
(It’s not just for marketers)
A lot of A/B testing tools in the
market. Choose wisely!
Take time to come up with your
hypothesis and how you’ll
measure it. It can make or break
your experiment.
23
If you’re interested, here’s how you get Wasabi:
https://github.com/intuit/wasabi
Questions?
#QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
Contact us: data@intuit.com
Questions?
Thanks for coming!
We’d love your feedback.
Before you head out, take a minute to rate this session in the
QuickBooks Connect mobile app.
Keep the conversation going
developer.intuit.com/hub/quickbooks-connect
Access Session Slides
Engage with Speakers and Fellow Developers
Get Questions Answered
QuickBooks Connect 2016 - Implementing analytic and optimization tools on your website or

QuickBooks Connect 2016 - Implementing analytic and optimization tools on your website or

  • 1.
    Brian Gardner -Director, Product Management Ilya Izrailevsky - Engineering Manager Implementing analytic and optimization tools on your website or app WiFi: QBConnect Password: not required#QBConnect
  • 2.
    2 Today’s Speakers Brian Gardner Director,Product Management @brianwgardner Ilya Izrailevsky Engineering Manager @iizrailevsky #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 3.
    Introduction & Overview Understandthe landscape analytic and optimization tools How to implement an A/B test Gain the ability to make informed decisions about what in your product is working (or not) Agenda
  • 4.
    Introduction & Overview Understandthe landscape analytic and optimization tools How to implement an A/B test Gain the ability to make informed decisions about what in your product is working (or not) Agenda
  • 5.
    5 “The sooner yourun an experiment, the sooner you are likely to find a surprise, and the surprise is the market speaking to you, telling you something you didn't know. Several of our businesses here came out of surprises.” – Scott Cook, Intuit founder & chairman Intuit’s tradition of experimentation Deep Customer Empathy Go Broad to Narrow Rapid Experiments with Customers Customer Delight #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 6.
    6 What is anA/B test? BA #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 7.
    Why A/B Test? #QBConnect| @IntuitDev | @brianwgardner | @iizrailevsky Discover which experiences delight your customer Accelerate your learning loop Make faster, better decisions with data
  • 8.
  • 9.
    9 Introducing: WASABI #QBConnect | @IntuitDev| @brianwgardner | @iizrailevsky
  • 10.
    10 First time usetest: QuickBooks Self Employed Test: Complete tax profile first Control: Connect account first In the test cell, we see: • Significant increase in tax profile completion • Slight increase in categorizing a transaction from a card • No change to connecting a bank • No change to categorizing on the transactions page #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 11.
    11 First time usetest: QuickBooks Self Employed Test: Complete tax profile first Control: Connect account first In the test cell, we see: • Significant increase in tax profile completion: +20% • Slight increase in categorizing a transaction from a card: +6% • No change to connecting a bank: Flat • No change to categorizing on the transactions page: Flat #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 12.
    Introduction & Overview Understandthe landscape analytic and optimization tools How to implement an A/B test Gain the ability to make informed decisions about what in your product is working (or not) Agenda
  • 13.
    13 • Marketing vsProduct (changing content & UI vs features or back-end) • How much engineering is involved? • Do you have sensitive data? • What platforms and devices do you need to test across? • Cost • Analytics provided / options • Efficient use of traffic • Performance • Ease of use How do you choose a tool for A/B testing? What are the dimensions you should think about when choosing A solution? #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 14.
    14 Landscape of PopularOptions • WYSIWIG • Minimal Engineering • Marketing focused • Developer focused • Flexible engineering effort • Own your own data • Customized • Ongoing maintenance • Own your own data #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky COMMERCIAL OPEN SOURCE IN-HOUSE WASABI
  • 15.
    Introduction & Overview Understandthe landscape analytic and optimization tools How to implement an A/B test Gain the ability to make informed decisions about what in your product is working (or not) Agenda
  • 16.
    16 The A/B TestingLifecycle: Iterate to Awesomeness #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky Step 1: Plan an experiment Step 2: Develop a feature to test Step 3: Instrument your application with Wasabi API Calls Step 4: Create and run the experiment using the Wasabi UI Step 5: Deploy and track results quickly
  • 17.
    17 • Is thisuser already assigned for this experiment? • Does the user qualify for the experiment? (Segmentation Rules) • Is the user already in a conflicting experiment? (Mutual Exclusion) • Randomization • Check vs. Sampling Percentage • Assign to an experience (Bucket) • Return assignment How does user assignment actually work? What should I show this user when a WASABI test is active? #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 18.
  • 19.
    19 Architecture: behind thescenes, here’s how it’s built Web tier Admin UI API API Segmentation Mutual Exclusion Sampling Bucket Dice Roll Application tier Cassandra MySQLData tier Authentication User Directory Logging Metrics Pluggable Components Bucket assignments Events / analytics Kafka Stream HDFS Spark Vertica S3 Pluggable Components Wasabi A/B Testing Service #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky
  • 20.
    Introduction & Overview Understandthe landscape analytic and optimization tools How to implement an A/B test Gain the ability to make informed decisions about what in your product is working (or not) Agenda
  • 21.
    21 Best practices andconsiderations for a good A/B test #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky Get started! (Don’t let the perfect be the enemy of the good)1 Think about what you’ll learn from a “failed” test in advance. Be skeptical – don’t take your results at face value • Track all your key metrics – check for unintended consequences • Think about sources of bias, assumptions: timing, infrastructure, etc. Don’t stop at one round of experiments! 2 3 4 5 Don’t forget qualitative research as well
  • 22.
    22 What Should YouTake Away? #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky Everyone can A/B test! (It’s not just for marketers) A lot of A/B testing tools in the market. Choose wisely! Take time to come up with your hypothesis and how you’ll measure it. It can make or break your experiment.
  • 23.
    23 If you’re interested,here’s how you get Wasabi: https://github.com/intuit/wasabi Questions? #QBConnect | @IntuitDev | @brianwgardner | @iizrailevsky Contact us: data@intuit.com
  • 24.
  • 25.
    Thanks for coming! We’dlove your feedback. Before you head out, take a minute to rate this session in the QuickBooks Connect mobile app. Keep the conversation going developer.intuit.com/hub/quickbooks-connect Access Session Slides Engage with Speakers and Fellow Developers Get Questions Answered