Upload
craig-sullivan
View
2.489
Download
1
Tags:
Embed Size (px)
Citation preview
The Myths, Lies and Illusions of Cross Device Testing
Craig Sullivan, Optimiser of Everything, @OptimiseOrDie
@OptimiseOrDie• Split Testing, Analytics, UX, Agile, Lean, Growth• 50M+ visitors tested, 19 languages, over 200 sites• 70+ mistakes I’ve made personally during testing• Like riding a bike, really really badly…
• Optimise or Kickstart your programme? • Get in touch!
Cross Device Testing Myths
1. Responsive solves everything
2. All our customers are on iPhones, right?
3. The customer journey is in your head
4. You don’t integrate with analytics
5. You think you’re tracking people
6. You only imagine the context
7. We think we have a hypothesis thingy
8. You think best practice is other tests
9. You just start testing, right?
10.What you see is what you get
11.95% confidence is enough for me!
12.You always get the promised lift
13.Segmentation is too hard
14.Who cares if it’s a phone?
15.Testing makes you a data scientist
@OptimiseOrDie
1. Motorola Hardware Menu Button2. MS Word Bullet Button3. Android Holo Composition Icon4. Android Context Action Bar Overflow (top right on Android
devices)
@OptimiseOrDie
Increase in revenue of > $200,000 per annum!
bit.ly/hamburgertest
@OptimiseOrDie
2: All our Customers use iPhones, right?
• Most common answer is “iPhones and iPads”
• Do you really know your mix?• Most people undercount
Android
• Use Google Analytics to find out
• Replace guesswork with truth!
• 2-3 hours work only• Get your top testing mix right
on:Desktop, Tablet, Mobile
• I’m writing an article – why?@OptimiseOrDi
e
2: Browser reports
@OptimiseOrDie
IE11 is our top desktop browser?
Chrome is DOUBLE this user volume
2: Browser reports
@OptimiseOrDie
• Be very careful when doing numbers on desktop browsers, tablet or mobile devices
• Chrome and Safari are on auto upgrades• All browser ‘upgrade’ at different speeds• Chrome and Firefox are fragmented as a
reason• Same for mobile – there are thousands of
Android handsets makers but the models are similar
• When looking at apple, you need to split by model (you can use the resolution to figure this out).
• iPads you can’t distinguish in GA*• Is the analysis you did really true?• Cluster your devices or browsers where
needed!• Article is coming!
3: The Customer Journey is in your Head
• The routes people take are not what we expect.
• Analytics data and Usability research are big pointers!
• Most common problem is the team not owning, experiencing and being immersed in the problems with your key journeys
• One charity wasted nearly 0.5M on a poor pathway
• Can you imagine someone from Mcdonalds never visiting their stores?
“So that’s how it looks on mobile!”
@OptimiseOrDie
3: Customer Journey - Solutions
"Great user experiences happen beyond the screen and in the gaps.“ Paul Boag• Test ALL key campaigns• Use Real Devices• Get your own emails• Order your own products• Call the phone numbers• Send an email• Send stuff back• Be difficult• Break things• Experience the end-end• Team are ALL mystery shoppers• Wear the magical slippers of the actual customer experience!• Be careful about dogfood though!
@OptimiseOrDie
• Investigating problems with tests• Tests that fail, flip or move around• Tests that don’t make sense• Broken test setups• Segmenting by Mobile, Tablet,
Desktop• Other customer segments• What drives the averages you see?
4: Our AB testing tool tells us all we need…
5: You think you’re tracking people
Basket
Shipping
Details
Pay
Product
Basket
ShippingDetail
sPay
Mobile Desktop
@OptimiseOrDie
5: You think you’re tracking people
• Keep people logged in• Use Social Logins• Identify unique customers• Feed this data to Universal Analytics• Follow users, not just device experiences
• It’s an attribution problem!
@OptimiseOrDie
6: You only imagine the context
• Tasks• Goals• Device• Location• Data rate• Viewport• Urgency• Motivation• Data costs• Call costs• Weather!
@OptimiseOrDie
6: You only imagine the context - solutions
bit.ly/multichannels@OptimiseOrDi
e
7: Other people’s tests are Best Practice
“STOP copying your competitors
They may not know what the f*** they are doing either” Peep Laja, ConversionXL
@OptimiseOrDie
Best Practice Testing?• Your customers are not the same• Your site is not the same• Your advertising and traffic is not the same• Your UX is not the same• Your X-Device Mix is not the same
• Use them to inform or suggest approaches• They’re like the picture on meal packets• Serving Suggestion Only• There are obvious BEST PRACTICES but
these are usually in the category of ‘bugs’ or ‘UX problems’
Just fix that now!
We have no clue
@OptimiseOrDie
Best Practice Testing?
“The Endless Suck of Best Practice and Optimisation Experts”bit.ly/socalledexperts
@OptimiseOrDie
Insight - Inputs
#FAIL
Competitor copying
GuessingDice rolling
An article the CEO
read
Competitor change
Panic
Ego
OpinionCherished
notions Marketing whims Cosmic rays
Not ‘on brand’ enough
IT inflexibility
Internal company
needs
Some dumbass
consultant
Shiny feature
blindnessKnee jerk reactons
8: You think you have a Hypothesis!
@OptimiseOrDie
Insight - Inputs
Insight
Segmentation
SurveysSales and
Call Centre
Session Replay
Social analytics
Customer contact
Eye tracking
Usability testing
Forms analytics Search
analytics Voice of Customer
Market research
A/B and MVT testing
Big & unstructured
data
Web analytics
Competitor evalsCustomer
services
8: These are inputs you need…
@OptimiseOrDie
Because we observed data [A] and feedback [B],
We believe that doing [C] for People [D] will make outcome [E] happen.
We’ll know this when we observe data [F] and obtain feedback [G].
(Reverse this)
@OptimiseOrDie
Because our CEO had an idea, that nobody else agreed with:
We believe that putting Orange Buttons on our Homepage will make people feel ‘Funkier’
We’ll know this when…
@OptimiseOrDie
Baseline Checks
Analytics health check
Developer onboarding
Goals & Metrics
Tool Setup
Analytics & Modelling
Discover ideas
Prioritise
Test cycles
9: You just start testing, right?
@OptimiseOrDie
TEST CYCLE
Hypothesis Design
Sketch
WireFrame
Mockup /
Prototype
Signoff Process
Build
QA
Soft and Hard
Launch
Analyse &
Publish
Learn
@OptimiseOrDie
Desktop www.crossbrowsertesting.com
www.browserstack.comwww.spoon.netwww.saucelabs.com
www.multibrowserviewer.com
Mobile & Tabletwww.appthwack.com
www.deviceanywhere.comwww.opendevicelab.com
Read this articlebit.ly/devicetesting
10: What you see is what you get…
@OptimiseOrDie
The 95% Stopping Problem
• Many people use 95, 99% ‘confidence’ to stop
• This value is unreliable and moves around• Nearly all my tests reach significance
before they are actually ready• You can hit 95% early in a test (18
minutes!)• If you stop, it could be a false result• Read this Nature article : bit.ly/1dwk0if• Optimizely have changed their stats
engine• This 95% thingy – is the cherry on the
cake!• Let me explain
@OptimiseOrDie
The 95% Stopping Problem
Scenario 1 Scenario 2 Scenario 3 Scenario 4
After 200 observations Insignificant Insignificant Significant! Significant!
After 500 observations Insignificant Significant! Insignificant Significant!
End of experiment Insignificant Significant! Insignificant Significant!
“You should know that stopping a test once it’s significant is deadly sin number 1 in A/B testing land. 77% of A/A tests (testing the same thing as A and B) will reach significance at a certain point.”Ton Wesseling, Online Dialogue
@OptimiseOrDie
The 95% Stopping Problem
“Statistical Significance does not equal Validity”http://bit.ly/1wMfmY2
“Why every Internet Marketer should be a Statistician”http://bit.ly/1wMfs1G
“Understanding the Cycles in your site”http://mklnd.com/1pGSOUP
@OptimiseOrDie
Business & Purchase Cycles
• Customers change• Your traffic mix changes• Markets, competitors• Be aware of all the
waves• Always test whole cycles• Don’t exclude slower
buyers• When you stop, let test
subjects still complete!
Start Test Finish Avg Cycle
@OptimiseOrDie
• TWO BUSINESS CYCLES minimum (week/mo)• 1 PURCHASE CYCLE minimum• 250 CONVERSIONS minimum per
creative • 350 & MORE! it depends on
response• FULL WEEKS/CYCLES never part of
one
• KNOW what marketing, competitors and cycles are doing
• RUN a test length calculator - bit.ly/XqCxuu• SET your test run time , RUN IT, STOP IT,
ANALYSE IT• ONLY RUN LONGER if you need more data• DON’T RUN LONGER just because the test isn’t
giving the result you want!
How Long to Run My Test and When to Stop
@OptimiseOrDie
The result is a range
• Version A is 3% conversion• Version B is 4% conversion• Yay! That’s a 25% lift• Let’s tell everyone
• When it goes live, you do NOT get 25%
• That’s because it was A RANGE
• 3% +/- 0.5 (could be 2.5-3.5)• 4% +/- 0.4 ( could be 3.6-4.4)
• Actual result was 3.5% for A• Actual result was 3.7% for B@OptimiseOrDi
e
Always Segment Experiences
• If you segment by devices, the sample gets smaller.
• A = 350 conversions• B = 300 conversions
• Desktop A 200, Tablet A 100, Mobile A 50• Desktop B 180, Tablet B 80, Mobile B 40
• It’s vital to segment by device class • You may also segment by breakpoint,
viewport or model• Make sure you know the proportion of
devices!• If you want to analyse, plan ahead!
@OptimiseOrDie
13: Wow – the browser is a phone too!
• Add call tracking• Buy a solution or• Make your own!• Measure calls• ROI on phone mix• Vital for PPC• Explain the costs or• Make it free
• Responsive solves everything No, it’s just an attribute
• All our customers are on iPhones, right? Make sure you know what they use!
• The customer journey is in your head Customer Insight, Research, Data
• You don’t integrate with analytics Use analytics, not the AB test data
• You think you’re tracking people Have an authentication strategy
• You only imagine the context Customer Insight, Diary Studies
• We think we have a hypothesis thingy Challenge all work with my outline
• You think best practice is other tests Leverage your customers, not theirs
14: SUMMARY
@OptimiseOrDie
• You just start testing, right? Preparation, Methodology, Prioritisation
• What you see is what you get QA testing with your Customer Shizzle
• 95% confidence is enough for me! Don’t stop tests when they hit 95%
• You always get the promised lift Quote ranges, not predictions
• Segmentation is too hard Segmentation – watch sample sizes
• Who cares if it’s a phone? Add call tracking or add ‘Tap 2 Call’
• Testing makes you a data scientist No, it doesn’t – it makes you humble
14: SUMMARY
@OptimiseOrDie
Rumsfeldian Space
• What if we changed our prices?• What if we gave away less for free?• What if we took this away?• What about 3 packages, not 5?• What are these potential futures I can
take?• How can I know before I spend money?
• UPS left hand turning -10 Million Gallons saved
• http://compass.ups.com/UPS-driver-avoid-left-turns/
• McDonalds Hipster Test Store• bit.ly/1TiURi7
#1 Culture & Team#2 Toolkit & Analytics investment#3 UX, CX, Service Design, Insight#4 Persuasive Copywriting#5 Experimentation (testing) tools
The 5 Legged Optimisation Barstool