FROM ANALYTICS TO DATA SCIENCE How to build your data maturity on - PowerPoint PPT Presentation

ANTTI ERONEN | @AEVARR | APRIL 17, 2018 FROM ANALYTICS TO DATA SCIENCE How to build your data maturity on Google Cloud Platform

AGENDA 1 2 3 DATA MATURITY MODEL BUSINESS CASE EXAMPLES ARCHITECTURE Characteristics of each What did we do? How was the example case level Why? implemented? @QVIK

IT’S ALL ABOUT DEVELOPING BUSINESS CAPABILITIES AT THE SAME TIME WITH TECHNOLOGICAL CAPABILITIES @QVIK

LEVEL 0 NO ANALYTICS ‣ Maybe tracking some vanity metrics like total downloads (always improving!) ‣ Standard in 2010… @QVIK

LEVEL 0 CASE: MARKETING APP ‣ Only purpose of the app is to get visibility to the brand BUSINESS QUESTIONS ‣ What it our position in App Store top downloaded list? THIS WAS #1 IN APP STORE! @QVIK

LEVEL 0 ARCHITECTURE iOS @QVIK

LEVEL 1 USAGE ANALYTICS ‣ “Analytics as a user story” ‣ Concerned with questions like… ‣ How many users? ‣ How do they navigate? ‣ What content do they consume? ‣ How much do they buy? @QVIK

LEVEL 1 CASE: MEDIA IN 2013 ‣ Analytics frameworks were used to track reader numbers ‣ Management was interested in the total number of users ‣ Development team did not use analytics to improve product/service @QVIK

LEVEL 1 ARCHITECTURE Google Analytics iOS / Android / Web @QVIK

LEVEL 2 CASE: BANKING APP ‣ Development team wanted a dashboard to see impact of their actions BUSINESS QUESTIONS ‣ Can we get more active users? ‣ What is the de fi nition of an active user? @QVIK

LEVEL 2 ARCHITECTURE Team dashboard Data Studio Report & Share 
 Tableau Calculated daily Analytics events Analytics results BigQuery Firebase BigQuery iOS / Android Daily queries with cron Compute Engine @QVIK

AND NOW FOR THE PAINFUL PART… LEVELING UP: 2 3 @QVIK

LEVEL 3 BUILD–MEASURE–LEARN ‣ Testing hypotheses to uncover behavior. ‣ If we want to ask questions afterwards, 
 we need to have raw data available. ‣ Concerned with questions like… ‣ Why should a change in fl uence this KPI? ‣ If users do A, do they also do B? ‣ Why are we losing customers? ‣ Are our design assumptions correct? @QVIK

LEVEL 3 CASE: RETAIL LOYALTY ‣ Analytics data from multiple sites and apps ‣ Client has many analysts and the hypotheses in development process are tested with real data BUSINESS QUESTIONS ‣ How to understand loyal customer behaviour when users are not logged in? @QVIK

LEVEL 3 ARCHITECTURE Team dashboard Data Studio Report & Share 
 Tableau Calculated daily Analytics events Analytics results BigQuery Firebase BigQuery iOS / Android Daily queries with cron Compute Engine @QVIK

LEVEL 3 ARCHITECTURE 3rd party ad pixel.gif framework Cloud Storage Notify new logs Pub/Sub Access logs Google Analytics Cloud Storage Parse logs Web Dataflow Analytics events User ID’s Report & Share 
 BigQuery BigQuery 3rd party tools Calculated daily results BigQuery Analytics Firebase iOS / Android Daily queries with cron Compute Engine @QVIK

LEVEL 4 DATA SCIENCE ‣ Using mathematical analysis or machine learning to also see what data can tell us without seeking answers to pre-de fi ned questions ‣ Needs a lot of (consistent) data ‣ The technical basis required: ‣ Data has to be combinable from di ff erent channels (mobile apps, web users, visits to physical stores, CRM, ERP, etc.) ‣ Customer / user IDs must be uniform or joinable to enable interesting observations on a single user level ‣ Concerned with questions like… ‣ Are there patterns in usage data that we did not assume? ‣ Can we build prediction models based on past behaviors? @QVIK

LEVEL 4 CASE: PREDICTING CUSTOMER FLOW ‣ Using historical data of customer fl ows to predict near future ‣ Combining external Google Trends data to improve prediction BUSINESS QUESTIONS ‣ How many employees we need working next week? @QVIK

LEVEL 4 ARCHITECTURE 3rd party ad pixel.gif framework Cloud Storage Notify new logs Pub/Sub Access logs Google Analytics Cloud Storage Parse logs Web Dataflow Analytics events User ID’s Report & Share 
 BigQuery BigQuery 3rd party tools Calculated daily results BigQuery Analytics Firebase iOS / Android Daily queries with cron Compute Engine @QVIK

LEVEL 4 ARCHITECTURE fy new logs /Sub rse logs Analysis Dataflow Datalab Calculated daily Prediction model results Cloud ML BigQuery @QVIK

FIREBASE PREDICTIONS @QVIK

QVIKIES HERE TODAY JERRY JALAVA JARI LINDHOLM ANTTI ERONEN PIA HÖGLUND Senior System Architect, 
 Head of Cloud Business Designer Digitalisation Google Developer Expert Business Consultant @QVIK