Mobile Apps
Analytics
Dmitry Klymenko
Hello
I am Dmitry Klymenko
Head of Enterprise Analytics / Digital and Productisation Team Leader
You can find me at @dmitry.klymenko.1
https://www.internetrix.com.au/blog/profile/dmitry-klymenko
.UA
About Me
.AU
Brief History
5,000 BC - Trypillian culture
2,500 BC - Egyptian pyramids build
~0 AD - A famous boy was born in Bethlehem
1985 AD - Dmitry Klymenko went to school
1999 AD - Dmitry Klymenko full-time employed
- Why Analytics? Personalisation - is that a thing?
- How mobile app analytics is different?
- App Analytics Platforms
- PII or “are we spying on customers?”
- GMP / GA / Firebase - a massive shift in the Google universe
Next 30 mins
Analytics
why Analytics?
Why Analytics is important
Insights & BI
Competitors do it :)
If you are not paying
for a product - you are
the product !
UX improvements
Growth “Killers”
Irrelevant content / functionality
Poor design / user experience
App speed
Poor marketing!
“You can’t improve what you
can’t measure”
9,526,124sessions last week
Whoa! That’s a big number. But is it good or bad?
Analytics Questions
Squirrel
How many users did we have yesterday?
123,456 customers have clicked this
button.
It seems users using the app for 43
seconds in the morning.
Ninjas
Has my user base increased or decreased?
Why app users engage more or less this
week?
When is the best time to use push
notifications targeting engagement?
“I will develop my own analytics”
“I can create a website like Facebook”
“I will re-write this library from scratch”
Collected Metrics
› Audience - who are your customers
› Acquisition - how we invited them
› Behaviour - what they have done on our digital
property
› Conversion - have they done what we wanted
them to do
Content Personalisation
Data is a currency
Personalisation and content actualisation is hard, but not impossible
It’s not just UI/UX
AI First concept replaces Mobile First
“Personalised Content”
User Accounts: so it can personalise for you!
› Different titles listed first at different times of the day.
› Different cover images for the same title.
› Email recommendations: “New on Netflix”.
› Constantly doing A/B testing.
Balance between value to customer and revenue
Recommendation Engine, Netflix
Web Analytics Platforms
› Google Analytics / Google Marketing Platform
› Adobe Experience Cloud
› IBM Coremetrics
› WebTrends, SnowPlow
› Log Analysers: AWStat
› Screen Recorders: ClickTale
How Mobile App Analytics is Different
when compared to web analytics
Website is first, Apps supplement data
› Mobile has two classes of analytics: for BI and for Devs
› Web has 99% of data completeness while mobile offline data might get lost
› Injecting “pixel” is much simpler on web
› The need to publish version with new tagging solution (weeks/months)
› A/B tests on mobile are much less mature web AB tests tools
› Marketing and Business Intelligence have the same expectation of data quality and
flexibility on mobile as they previously did on web
Mobile Analytics and Web
For Developers
- Firebase (Fabric)
- Crashlytics
- devtodev
- Performance-based, debug tools
- App Stores
Mobile Analytics For Devs
For Business Intelligence
- App Stores (iOS App Analytics)
- Google Analytics/Firebase
- Adobe Experience Cloud
- Flurry Analytics
- Localytics
- Mixpanel
Mobile Analytics - For BI
- Appsee
- Countly
- App Annie
- AppFigures
- Bango
- UXCam
- Heap Analytics
- ClickTale
- SnowPlow
- Filters by Platform, Date, Audience, User Properties (custom filter). Funnels*.
- Active Users, Conversions, Engagement, Revenue.
- Adoption and Acquisition.
- Retention and Audience - Location, Devices, Demographics
- Goals
- Events
- Automatically collected: first time, in-app purchase, ads.
- Custom: what’s important for your business.
Mobile Analytics - Common
- Cohorts, Funnels, User Paths
- Segmentation
- Acquisition channel, Usage, Age,
Gender, Language, Geo, etc
- Crash Analytics
- Real-Time & Revenue Analytics
- Audience targeting users via Flurry
Push
Mobile Analytics - Flurry Analytics
Flurry Analytics - implementation details
https://irx.page.link/9qXE
- Engagement
- Retention
- Funnels
- Audience
- A/B testing
- Push notifications
Mobile Analytics - Mixpanel
- Discover, Engage, Optimise
- Predictive Analytics
- Attribution
- Remarketing / Audience, Funnels
- “Mobile CRM”, Push, In-App, APIs
- True Impact (tm), A/B Testing,
Uninstall Tracking
Mobile Analytics - Localytics
- Games Oriented
- Retention
- Segmentation
- SDK Integrations to different
game platforms
- SQL-like report builder
- Push notifications
Mobile Analytics - devtodev
- User Experience Oriented
- User Session Recording (events)
- Touch Heatmaps
- Crash Reporting (how to reproduce)
- Real-Time Analytics and Alerts
- Integration with Big Analytics
platforms
Mobile Analytics - appsee
- User Experience Oriented
- User Session Recording (events)
- Touch Heatmaps
- User Analysis
- Screen UX Analysis
Mobile Analytics - uxcam
- Marketing Performance Oriented
- App Store insights, rank and app reviews
monitor
- Sales Prospecting
- Not just for app publishers and marketing,
but for SDK makers as well
- Screen UX Analysis
Mobile Analytics - appfigures
- App Market Data
- Top Charts, Rank History, Rating, Reviews
and Feature Tracker
- Market Size, Audience Estimate
Mobile Analytics - AppClix
- User Profiles, Cohorts
- In App Usage Analytics: Retention,
Custom Events, Conversion Funnels
- Engage with your customers
- Crash Analytics
- Ratings and Feedback
- Extensible via plugins (~100)
Mobile Analytics - Countly
- PII Masking / GDPR compliance
- Session Replays (filter by KPIs / Events / Actions)
- Visualize User Experience for proactive & strategic UX
optimizations
- Uncover most pressing UX & funnel issues
- Automated surfacing of influential segments per app
screen
- Crash Trends / Application Not Responding, tied back to
session replay to understand experience
- Powerful integrations including analytics, VoC,
Application performance monitoring
- Minimal impact on performance (data sync over WiFI)
- Visualizations / Heatmaps
Mobile Analytics - Clicktale
Firebase [Analytics]
- App Analytics
- Database messages
- Crash Reporting
- Performance Monitoring
- In App Messaging
- AB Testing
- Predictive modelling
- Cloud messaging
- Remote configuration
- App Indexing
- Dynamic Links
Mobile Analytics Comparison
Clicktale
- App Analytics & Heatmaps
- Visualize User Experience for proactive &
strategic UX optimizations via session replays
- Crash Reporting & session replay to
understand experience
- Performance Monitoring & Automated
surfacing of influential segments per app
screen
- Powerful integrations including analytics, VoC,
Application performance monitoring
- Event tracking
WHAT HAPPENED vs WHY DID IT HAPPEN
Analytics Types
Based on events
Based on Fact or
Event.
App generates events
and sends it with
metadata to backend.
Recording
Event-based
recording of taps,
mouse movements,
screenshots.
Access logs
App pings it’s
backend creating a
“digital footprint”
which is being
analysed later by
analytics software.
Big Data
› No longer a trend/hype, but rather a must-have
› AI & ML are the main consumers for Big Data.
Storage and processing is cheap - cloud is cost-effective
› GA & Firebase have BigQuery.
Others use Amazon Redshift
› Cheaper to store data indefinitely then delete selectively
Data Warehouse vs Data Lake
Data Warehouse
› Traditional approach
› Structured data
› Schema on write
› CRM, financial transactions, ERP
Data Lake
› Latest “invention”
› Structured or unstructured
› Schema on read
› Social media, web server logs,
sensor data, documents, media
files, etc
https://irx.page.link/d1Zt
What is Framework?
Structure and collection of magic tools to solve the high level problem
› View (screen/page)
› Session
› User
› Unique (over period)
› Active
› Time on Page/Screen
› Campaigns, UTMs
› Real-Time
› Organic
Very Short Analytics Glossary
› Landing
› Segment
› Experiment (A/B, Multi-variant)
› Funnel, Goal, Conversion
› CPC, CPM
› Bounce
› Audience
› Tracking, pixel, snippet,
conversion
› *Framework - favourite one
Testing is not QA for Analytics People!
› Challenge is to define a Control Group and an Experiment Group correctly. Remembering we are
talking about Customers!
› Particular user should not be exposed to multiple variants.
› It’s not just more or less. A/B tool must prove that results are statistically significant.
› We are poor at assessing the value of our ideas.
Data Privacy and Analytics
did someone said GDPR?
Are we spying on our customers?
Not at an individual level
- PII is not recorded.
- Collection of anonymous users.
- Aggregated data matters. Looking at
individual is waste of time.
- Personalisation as a goal.
- Web Server logs been there for ages!
- Screen recording, shadow browsing -
invaluable for support.
- Remarketing is annoying - how Google
knows that I want to buy a car?
- AI / ML to detect when you are hungry to
offer you a pizza.
- Offline beacons / apps on your phone
Personalised Content or Advertising
A GIRL HAS NO NAME
especially in Google Analytics
User Explorer Report in Google Analytics
Do not save PII in GA
Individual, non-personalised
interactions per each user
GA1.2.667464943.
140000782834
PII in Google Analytics, explained
https://goo.gl/gzXAaI
New dimension exposed
Not a PII
Client ID vs User ID
Client ID
- Represents an anonymous device
- A random number generated by GA
- Always required
- Disappears when cookies are gone
User ID
- Represents a single user
- Generated by your website backend
and passed to GA
- Optional
- Affects data collection, requires
filtered View
“Say GDPR one more time...”
Google Analytics GDPR compliance
› “Do not track” unless consent given
› Individual row deletion
› Privacy policy explaining GA and DoubleClick cookies
› Anonymise IP
› Server logs
Intelligent Tracking Prevention - ITP 2.1
› Google Analytics cookie lifespan is couple days.
LocalStorage isn’t really a solution
› Safari share is 18% and growing https://irx.page.link/TkFr
› This is unavoidable
› Server-side tracking
Tag Management
Google Tag Manager, Tealium, etc
Tag Manager
› “External Config” for apps or web because marketing and BI
needs updates more often than functionality
› Very effective way to hack your own website. Bypass QA
› Honours Data Layer - context in machine readable format
› Tag Manager updates are very fast
› Third-party library in your codebase
Tag Manager
› Tag Manager business logic executes locally within
browser or app
› Common case - developers emit all possible types of
events to data layer and Tag Manager Business Logic
makes decision what is being sent and where
GMP / GA / Firebase
current state of apps analytics in Google universe
Measurement Protocol
Reference: https://goo.gl/2KApXu
Required parameters for all hit types
v - MP protocol version - Always equal to 1.
tid - Tracking ID.
cid - Client (device, browser) ID - Required if User ID is not specified.
uid - User ID - Required if cid is not specified
t - HitType - Pageview, event, etc.
z - (optional, but recommended) - Cache buster
POST (or GET) request to https://www.google-analytics.com/collect?PARAMS
GA SDK
GTM SDK
Legacy
Legacy GA for Apps
UA-XXXXXX-XX
Tag Manager +
Firebase SDK
iOS & Android - v5
Recommended
“External Config”
Google Tag Manager
Will be
depreciated
Google Analytics
for Firebase
Depreciated and in the
future data deletion
Actively developed
Firebase SDK
Firebase Analytics Firebase Cloud /
Console Big Query
Inside an app
Hit-level data in DB
Internetrix, Dmitry Klymenko, www.internetrix.com.au
iOS - v3, Android - v4
Tracking Recommendation
› Use Tag Manager + Firebase SDK
› Bunch of automatic events
› Everything sent to Tag Manager by default is routed to
Firebase
› Tag Manager controls Tracking Business Logic in your app
via it’s own website
THANKS!
Any questions? Free pizza
You can find me at:
@dmitry.klymenko.1
dmitry.klymenko@internetrix.com.au
https://www.linkedin.com/in/dmitry-klymenko-b9469810a/

Mobile App Analytics. Why, How, What's new - Mar 2019

  • 1.
  • 2.
    Hello I am DmitryKlymenko Head of Enterprise Analytics / Digital and Productisation Team Leader You can find me at @dmitry.klymenko.1 https://www.internetrix.com.au/blog/profile/dmitry-klymenko
  • 3.
  • 4.
    Brief History 5,000 BC- Trypillian culture 2,500 BC - Egyptian pyramids build ~0 AD - A famous boy was born in Bethlehem 1985 AD - Dmitry Klymenko went to school 1999 AD - Dmitry Klymenko full-time employed
  • 5.
    - Why Analytics?Personalisation - is that a thing? - How mobile app analytics is different? - App Analytics Platforms - PII or “are we spying on customers?” - GMP / GA / Firebase - a massive shift in the Google universe Next 30 mins
  • 6.
  • 7.
    Why Analytics isimportant Insights & BI Competitors do it :) If you are not paying for a product - you are the product ! UX improvements
  • 8.
    Growth “Killers” Irrelevant content/ functionality Poor design / user experience App speed Poor marketing!
  • 9.
    “You can’t improvewhat you can’t measure”
  • 10.
    9,526,124sessions last week Whoa!That’s a big number. But is it good or bad?
  • 11.
    Analytics Questions Squirrel How manyusers did we have yesterday? 123,456 customers have clicked this button. It seems users using the app for 43 seconds in the morning. Ninjas Has my user base increased or decreased? Why app users engage more or less this week? When is the best time to use push notifications targeting engagement?
  • 12.
    “I will developmy own analytics” “I can create a website like Facebook” “I will re-write this library from scratch”
  • 13.
    Collected Metrics › Audience- who are your customers › Acquisition - how we invited them › Behaviour - what they have done on our digital property › Conversion - have they done what we wanted them to do
  • 14.
    Content Personalisation Data isa currency Personalisation and content actualisation is hard, but not impossible It’s not just UI/UX AI First concept replaces Mobile First
  • 15.
  • 16.
    User Accounts: soit can personalise for you! › Different titles listed first at different times of the day. › Different cover images for the same title. › Email recommendations: “New on Netflix”. › Constantly doing A/B testing. Balance between value to customer and revenue Recommendation Engine, Netflix
  • 17.
    Web Analytics Platforms ›Google Analytics / Google Marketing Platform › Adobe Experience Cloud › IBM Coremetrics › WebTrends, SnowPlow › Log Analysers: AWStat › Screen Recorders: ClickTale
  • 18.
    How Mobile AppAnalytics is Different when compared to web analytics
  • 19.
    Website is first,Apps supplement data › Mobile has two classes of analytics: for BI and for Devs › Web has 99% of data completeness while mobile offline data might get lost › Injecting “pixel” is much simpler on web › The need to publish version with new tagging solution (weeks/months) › A/B tests on mobile are much less mature web AB tests tools › Marketing and Business Intelligence have the same expectation of data quality and flexibility on mobile as they previously did on web Mobile Analytics and Web
  • 20.
    For Developers - Firebase(Fabric) - Crashlytics - devtodev - Performance-based, debug tools - App Stores Mobile Analytics For Devs
  • 21.
    For Business Intelligence -App Stores (iOS App Analytics) - Google Analytics/Firebase - Adobe Experience Cloud - Flurry Analytics - Localytics - Mixpanel Mobile Analytics - For BI - Appsee - Countly - App Annie - AppFigures - Bango - UXCam - Heap Analytics - ClickTale - SnowPlow
  • 22.
    - Filters byPlatform, Date, Audience, User Properties (custom filter). Funnels*. - Active Users, Conversions, Engagement, Revenue. - Adoption and Acquisition. - Retention and Audience - Location, Devices, Demographics - Goals - Events - Automatically collected: first time, in-app purchase, ads. - Custom: what’s important for your business. Mobile Analytics - Common
  • 23.
    - Cohorts, Funnels,User Paths - Segmentation - Acquisition channel, Usage, Age, Gender, Language, Geo, etc - Crash Analytics - Real-Time & Revenue Analytics - Audience targeting users via Flurry Push Mobile Analytics - Flurry Analytics
  • 24.
    Flurry Analytics -implementation details https://irx.page.link/9qXE
  • 25.
    - Engagement - Retention -Funnels - Audience - A/B testing - Push notifications Mobile Analytics - Mixpanel
  • 26.
    - Discover, Engage,Optimise - Predictive Analytics - Attribution - Remarketing / Audience, Funnels - “Mobile CRM”, Push, In-App, APIs - True Impact (tm), A/B Testing, Uninstall Tracking Mobile Analytics - Localytics
  • 27.
    - Games Oriented -Retention - Segmentation - SDK Integrations to different game platforms - SQL-like report builder - Push notifications Mobile Analytics - devtodev
  • 28.
    - User ExperienceOriented - User Session Recording (events) - Touch Heatmaps - Crash Reporting (how to reproduce) - Real-Time Analytics and Alerts - Integration with Big Analytics platforms Mobile Analytics - appsee
  • 29.
    - User ExperienceOriented - User Session Recording (events) - Touch Heatmaps - User Analysis - Screen UX Analysis Mobile Analytics - uxcam
  • 30.
    - Marketing PerformanceOriented - App Store insights, rank and app reviews monitor - Sales Prospecting - Not just for app publishers and marketing, but for SDK makers as well - Screen UX Analysis Mobile Analytics - appfigures
  • 31.
    - App MarketData - Top Charts, Rank History, Rating, Reviews and Feature Tracker - Market Size, Audience Estimate Mobile Analytics - AppClix
  • 32.
    - User Profiles,Cohorts - In App Usage Analytics: Retention, Custom Events, Conversion Funnels - Engage with your customers - Crash Analytics - Ratings and Feedback - Extensible via plugins (~100) Mobile Analytics - Countly
  • 33.
    - PII Masking/ GDPR compliance - Session Replays (filter by KPIs / Events / Actions) - Visualize User Experience for proactive & strategic UX optimizations - Uncover most pressing UX & funnel issues - Automated surfacing of influential segments per app screen - Crash Trends / Application Not Responding, tied back to session replay to understand experience - Powerful integrations including analytics, VoC, Application performance monitoring - Minimal impact on performance (data sync over WiFI) - Visualizations / Heatmaps Mobile Analytics - Clicktale
  • 34.
    Firebase [Analytics] - AppAnalytics - Database messages - Crash Reporting - Performance Monitoring - In App Messaging - AB Testing - Predictive modelling - Cloud messaging - Remote configuration - App Indexing - Dynamic Links Mobile Analytics Comparison Clicktale - App Analytics & Heatmaps - Visualize User Experience for proactive & strategic UX optimizations via session replays - Crash Reporting & session replay to understand experience - Performance Monitoring & Automated surfacing of influential segments per app screen - Powerful integrations including analytics, VoC, Application performance monitoring - Event tracking WHAT HAPPENED vs WHY DID IT HAPPEN
  • 35.
    Analytics Types Based onevents Based on Fact or Event. App generates events and sends it with metadata to backend. Recording Event-based recording of taps, mouse movements, screenshots. Access logs App pings it’s backend creating a “digital footprint” which is being analysed later by analytics software.
  • 36.
    Big Data › Nolonger a trend/hype, but rather a must-have › AI & ML are the main consumers for Big Data. Storage and processing is cheap - cloud is cost-effective › GA & Firebase have BigQuery. Others use Amazon Redshift › Cheaper to store data indefinitely then delete selectively
  • 37.
    Data Warehouse vsData Lake Data Warehouse › Traditional approach › Structured data › Schema on write › CRM, financial transactions, ERP Data Lake › Latest “invention” › Structured or unstructured › Schema on read › Social media, web server logs, sensor data, documents, media files, etc https://irx.page.link/d1Zt
  • 38.
    What is Framework? Structureand collection of magic tools to solve the high level problem
  • 39.
    › View (screen/page) ›Session › User › Unique (over period) › Active › Time on Page/Screen › Campaigns, UTMs › Real-Time › Organic Very Short Analytics Glossary › Landing › Segment › Experiment (A/B, Multi-variant) › Funnel, Goal, Conversion › CPC, CPM › Bounce › Audience › Tracking, pixel, snippet, conversion › *Framework - favourite one
  • 40.
    Testing is notQA for Analytics People! › Challenge is to define a Control Group and an Experiment Group correctly. Remembering we are talking about Customers! › Particular user should not be exposed to multiple variants. › It’s not just more or less. A/B tool must prove that results are statistically significant. › We are poor at assessing the value of our ideas.
  • 41.
    Data Privacy andAnalytics did someone said GDPR?
  • 42.
    Are we spyingon our customers? Not at an individual level - PII is not recorded. - Collection of anonymous users. - Aggregated data matters. Looking at individual is waste of time. - Personalisation as a goal. - Web Server logs been there for ages! - Screen recording, shadow browsing - invaluable for support.
  • 43.
    - Remarketing isannoying - how Google knows that I want to buy a car? - AI / ML to detect when you are hungry to offer you a pizza. - Offline beacons / apps on your phone Personalised Content or Advertising
  • 44.
    A GIRL HASNO NAME especially in Google Analytics
  • 45.
    User Explorer Reportin Google Analytics Do not save PII in GA Individual, non-personalised interactions per each user GA1.2.667464943. 140000782834 PII in Google Analytics, explained https://goo.gl/gzXAaI
  • 46.
  • 47.
    Client ID vsUser ID Client ID - Represents an anonymous device - A random number generated by GA - Always required - Disappears when cookies are gone User ID - Represents a single user - Generated by your website backend and passed to GA - Optional - Affects data collection, requires filtered View
  • 48.
    “Say GDPR onemore time...”
  • 49.
    Google Analytics GDPRcompliance › “Do not track” unless consent given › Individual row deletion › Privacy policy explaining GA and DoubleClick cookies › Anonymise IP › Server logs
  • 50.
    Intelligent Tracking Prevention- ITP 2.1 › Google Analytics cookie lifespan is couple days. LocalStorage isn’t really a solution › Safari share is 18% and growing https://irx.page.link/TkFr › This is unavoidable › Server-side tracking
  • 51.
    Tag Management Google TagManager, Tealium, etc
  • 52.
    Tag Manager › “ExternalConfig” for apps or web because marketing and BI needs updates more often than functionality › Very effective way to hack your own website. Bypass QA › Honours Data Layer - context in machine readable format › Tag Manager updates are very fast › Third-party library in your codebase
  • 53.
    Tag Manager › TagManager business logic executes locally within browser or app › Common case - developers emit all possible types of events to data layer and Tag Manager Business Logic makes decision what is being sent and where
  • 54.
    GMP / GA/ Firebase current state of apps analytics in Google universe
  • 55.
    Measurement Protocol Reference: https://goo.gl/2KApXu Requiredparameters for all hit types v - MP protocol version - Always equal to 1. tid - Tracking ID. cid - Client (device, browser) ID - Required if User ID is not specified. uid - User ID - Required if cid is not specified t - HitType - Pageview, event, etc. z - (optional, but recommended) - Cache buster POST (or GET) request to https://www.google-analytics.com/collect?PARAMS
  • 56.
    GA SDK GTM SDK Legacy LegacyGA for Apps UA-XXXXXX-XX Tag Manager + Firebase SDK iOS & Android - v5 Recommended “External Config” Google Tag Manager Will be depreciated Google Analytics for Firebase Depreciated and in the future data deletion Actively developed Firebase SDK Firebase Analytics Firebase Cloud / Console Big Query Inside an app Hit-level data in DB Internetrix, Dmitry Klymenko, www.internetrix.com.au iOS - v3, Android - v4
  • 57.
    Tracking Recommendation › UseTag Manager + Firebase SDK › Bunch of automatic events › Everything sent to Tag Manager by default is routed to Firebase › Tag Manager controls Tracking Business Logic in your app via it’s own website
  • 58.
    THANKS! Any questions? Freepizza You can find me at: @dmitry.klymenko.1 dmitry.klymenko@internetrix.com.au https://www.linkedin.com/in/dmitry-klymenko-b9469810a/

Editor's Notes

  • #8 Wolf Messing, secret service
  • #13 You can develop your own analytics, but devil is in details
  • #15 Wolf Messing, secret service
  • #16 This is what BI people look at and urge developers to implement personalisation :)
  • #49 GDPR is good