HOW TO WIN AT
CONVERSION OPTIMIZATION
- SNEHAL SAMANT
WHO’S SNEHAL SAMANT?
Leads Product Marketing @ VWO
– World’s easiest A/B testing tool
Head of Digital @ TastyKhana – acquired by FoodPanda
Co-founder @ Black Bean Engagement
Creative Director @ Grey
OPINIONS ARE LIKE ASSHOLES,
EVERYONE HAS ONE
(BUT EACH THINKS THE OTHER ’S STINKS)
NO 1 REASON WHY COMPANIES FAIL AT GROWTH?
Not enough ideas.
We aren't tracking the right data.
Don't have a testing framework.
They don't run enough experiments
ONLY 1 OUT OF 7 TESTS WIN
6 OUT OF 7 LOSE
Then why do people keep experimenting?
WITH EACH EXPERIMENT LEARN
SOMETHING MORE ABOUT YOUR CUSTOMER
NOW YOU ARE CONTINOUSLY WINNING
MORE CUSTOMER KNOWLEDGE = GROWTH
Your customer knowledge
Revenue
1m ARR
10m ARR
100m ARR
BETTER FEATURES BETTER PRODUCTS BETTER SERVICE
WHERE WINS
LIVE?
MEASURE
SHIT
GREAT
INSIGHTS
PERSUASION
This deck isn't a
compilation of hacks.
LET'S LEARN TO FISH.
Chamath Palihapitiya - how we put Facebook on the path to 1 billion users
IDENTIFY YOUR MOST IMPORTANT METRICS
• AARRR framework
• Start tracking
• Establish baseline to
predict impact
Growth Hacking Like A Pirate,
A Beginner’s Guide To Pirate Metrics
Point-click to define events and
push to GA, MixPanel etc
COME UP WITH A STRONG HYPOTHESIS
Changing
[element tested - button, headline etc]
from ‘________________’
to ‘_____________’
will increase / decrease
[a conversion metric - clicks, downloads, signups etc].
by ‘______%’
BECAUSE ‘________’
A STRONG HYPOTHESIS TARGETS PAIN
Changing
THE CANCEL ACCOUNT SCREEN
from ‘Letting people cancel’
to ‘Letting people pause their account by downgrading to $9 plan’
will decrease
Account Churn Rate
by ‘20%’
BECAUSE ‘some customers do not want to lose their data
which we delete on cancellation.’
HOW TO FIND PAIN #1 - DATA ANALYSIS
Find aberration
High traffic + high bounce, Paid traffic + high bounce
Heatmaps, clickmaps, scrollmaps, visitor recordings
Look at internal search
Look at search landing pages from organic search
Become a GA ninja
HOW TO FIND PAIN #2 - HEURISTIC ANALYSIS
Structured approach to
critically evaluating UI /
UX. LIFT is one of several
frameworks.
An intro to the LIFT model
HOW TO FIND PAIN #3
TALK TO PEOPLE!
Ask users – Website / Email Surveys
Call them!
HOW TO PERSUADE
PEOPLE TO DO
WHAT YOU WANT
(ZE BLACK ARTS)
HOW #1 - USE USER VOICE
People skip marketing-speak
Use real-speak from surveys, conversation responses
Keep it simple
Learn to write super copy
How #2
VISUAL HIERARCHY
It works
Don’t fuck it up
HOW #3 - NUEROMARKETING 101
THE REPLTILIAN BRAIN TRIGGERS
Is There a Buy Button Inside the Brain?
HOW #4 - NUDGES - EXPLOIT BIASES
A nudge alters people’ s behaviour in a
predictable way without forbidding any
options or significantly changing their
economic incentives.
Putting fruit at eye level counts as a
nudge. Banning junk food does not.
COMMON BIASES Anchoring effect
40,000 6,000
COMMON BIASES Anchoring effect
900 200
Sarojini Market
Perception of cost is relative.
COMMON BIASES Current Moment Bias
People choose the pleasurable option available right now.
Constant feedback helps drive behavior.
COMMON BIASES Projection bias
Most people think like me, act like me and like what I like.
Most innovative, data-focused marketers
use VWO for driving growth
at products like Slack.
COMMON BIASES Observational Selection bias
Suddenly noticing things we didn’t notice that much before
and wrongly assume that the frequency has increased. A
perfect example is what happens after we buy a red car
and we inexplicably start to see red cars virtually
everywhere.
COMMON BIASES Confirmation bias
We love to agree with people who agree with us.
We tend to be put off by people and news sources that
make us feel uncomfortable or insecure about our views –
cognitive dissonance.
We focus on data which confirms our pre-conceived view.
5 Principles of Persuasive Web Design
20 cognitive biases
that skew decisions
RUNNING EXPERIMENTS
Simple sequential testing – establish baseline, make a change,
expose 100% of your audience
Control + Variation – expose 50% random of your audience to
each at the same time
Use a tool like VWO – visual editor + repeat visitors see same
control or variation + Bayesian stats engine (most use
Frequentist which takes too long and uses p value)
WHY EXPERIMENTS CAN SKEW
Calculate sample size in advance for
statistically significant results (not required with VWO)
Don’t make changes mid-way
Test in weeks for equal number of Mondays, weekends etc
Be wary of external forces
INTERPRETING YOUR RESULTS
Aim for validated learnings, not just wins
THINK BEYOND HACKS
Improve processes
Pilot new, different mediums
GROWTH IS OPTIONAL
The easiest choice is to choose to do nothing.
Get these slides on SNEHALSAMANT.COM

How to win at Conversion Optimization

  • 1.
    HOW TO WINAT CONVERSION OPTIMIZATION - SNEHAL SAMANT
  • 2.
    WHO’S SNEHAL SAMANT? LeadsProduct Marketing @ VWO – World’s easiest A/B testing tool Head of Digital @ TastyKhana – acquired by FoodPanda Co-founder @ Black Bean Engagement Creative Director @ Grey
  • 3.
    OPINIONS ARE LIKEASSHOLES, EVERYONE HAS ONE (BUT EACH THINKS THE OTHER ’S STINKS)
  • 4.
    NO 1 REASONWHY COMPANIES FAIL AT GROWTH? Not enough ideas. We aren't tracking the right data. Don't have a testing framework. They don't run enough experiments
  • 5.
    ONLY 1 OUTOF 7 TESTS WIN 6 OUT OF 7 LOSE Then why do people keep experimenting?
  • 6.
    WITH EACH EXPERIMENTLEARN SOMETHING MORE ABOUT YOUR CUSTOMER NOW YOU ARE CONTINOUSLY WINNING
  • 7.
    MORE CUSTOMER KNOWLEDGE= GROWTH Your customer knowledge Revenue 1m ARR 10m ARR 100m ARR BETTER FEATURES BETTER PRODUCTS BETTER SERVICE
  • 8.
  • 9.
    This deck isn'ta compilation of hacks. LET'S LEARN TO FISH.
  • 10.
    Chamath Palihapitiya -how we put Facebook on the path to 1 billion users
  • 11.
    IDENTIFY YOUR MOSTIMPORTANT METRICS • AARRR framework • Start tracking • Establish baseline to predict impact
  • 12.
    Growth Hacking LikeA Pirate, A Beginner’s Guide To Pirate Metrics Point-click to define events and push to GA, MixPanel etc
  • 13.
    COME UP WITHA STRONG HYPOTHESIS Changing [element tested - button, headline etc] from ‘________________’ to ‘_____________’ will increase / decrease [a conversion metric - clicks, downloads, signups etc]. by ‘______%’ BECAUSE ‘________’
  • 14.
    A STRONG HYPOTHESISTARGETS PAIN Changing THE CANCEL ACCOUNT SCREEN from ‘Letting people cancel’ to ‘Letting people pause their account by downgrading to $9 plan’ will decrease Account Churn Rate by ‘20%’ BECAUSE ‘some customers do not want to lose their data which we delete on cancellation.’
  • 15.
    HOW TO FINDPAIN #1 - DATA ANALYSIS Find aberration High traffic + high bounce, Paid traffic + high bounce Heatmaps, clickmaps, scrollmaps, visitor recordings Look at internal search Look at search landing pages from organic search
  • 16.
  • 17.
    HOW TO FINDPAIN #2 - HEURISTIC ANALYSIS Structured approach to critically evaluating UI / UX. LIFT is one of several frameworks.
  • 18.
    An intro tothe LIFT model
  • 19.
    HOW TO FINDPAIN #3 TALK TO PEOPLE! Ask users – Website / Email Surveys Call them!
  • 20.
    HOW TO PERSUADE PEOPLETO DO WHAT YOU WANT (ZE BLACK ARTS)
  • 21.
    HOW #1 -USE USER VOICE People skip marketing-speak Use real-speak from surveys, conversation responses Keep it simple
  • 22.
    Learn to writesuper copy
  • 23.
    How #2 VISUAL HIERARCHY Itworks Don’t fuck it up
  • 25.
    HOW #3 -NUEROMARKETING 101
  • 26.
  • 28.
    Is There aBuy Button Inside the Brain?
  • 29.
    HOW #4 -NUDGES - EXPLOIT BIASES A nudge alters people’ s behaviour in a predictable way without forbidding any options or significantly changing their economic incentives. Putting fruit at eye level counts as a nudge. Banning junk food does not.
  • 30.
    COMMON BIASES Anchoringeffect 40,000 6,000
  • 31.
    COMMON BIASES Anchoringeffect 900 200 Sarojini Market Perception of cost is relative.
  • 32.
    COMMON BIASES CurrentMoment Bias People choose the pleasurable option available right now. Constant feedback helps drive behavior.
  • 33.
    COMMON BIASES Projectionbias Most people think like me, act like me and like what I like. Most innovative, data-focused marketers use VWO for driving growth at products like Slack.
  • 34.
    COMMON BIASES ObservationalSelection bias Suddenly noticing things we didn’t notice that much before and wrongly assume that the frequency has increased. A perfect example is what happens after we buy a red car and we inexplicably start to see red cars virtually everywhere.
  • 35.
    COMMON BIASES Confirmationbias We love to agree with people who agree with us. We tend to be put off by people and news sources that make us feel uncomfortable or insecure about our views – cognitive dissonance. We focus on data which confirms our pre-conceived view.
  • 36.
    5 Principles ofPersuasive Web Design 20 cognitive biases that skew decisions
  • 37.
    RUNNING EXPERIMENTS Simple sequentialtesting – establish baseline, make a change, expose 100% of your audience Control + Variation – expose 50% random of your audience to each at the same time Use a tool like VWO – visual editor + repeat visitors see same control or variation + Bayesian stats engine (most use Frequentist which takes too long and uses p value)
  • 38.
    WHY EXPERIMENTS CANSKEW Calculate sample size in advance for statistically significant results (not required with VWO) Don’t make changes mid-way Test in weeks for equal number of Mondays, weekends etc Be wary of external forces
  • 39.
    INTERPRETING YOUR RESULTS Aimfor validated learnings, not just wins
  • 40.
    THINK BEYOND HACKS Improveprocesses Pilot new, different mediums
  • 41.
    GROWTH IS OPTIONAL Theeasiest choice is to choose to do nothing. Get these slides on SNEHALSAMANT.COM