This document provides an overview of analytics and its maturity levels from descriptive to predictive to prescriptive and even decisive analytics. It discusses what each level involves in terms of typical experiences, necessary data and questions, and key aspects. Descriptive analytics involves understanding what has happened through dashboards and KPIs. Predictive analytics predicts what may happen through alerts and notifications. Prescriptive analytics anticipates outcomes and recommends actions. The document also briefly introduces machine learning concepts and discusses ensuring analytics solutions have a strong user experience.
2. Background
• Head of Products – Cisco’s Cloud Analytics Products
• Product and Engineering sponsor for Cisco’s acquisition of AppDynamics
• 10+ Years at Microsoft with primary focus on Azure and Analytics
• Building SaaS and Cloud Analytics products since 2000
• Founder and leader of the Machine Learning and Analytics Engineering
workgroup within Microsoft
• Author
• Coder
• Technical/Product Advisor to several start-ups
• Entrepreneur – Founded and lead 3 startups
• LinkedIn profile
7. Me (Implicit): “How long is it
going to take me to reach
home?”
Google Maps: “It will take you
74 mins because the traffic is
unusually heavy today”
19. Dang it! I should have shorted
Oil between July and Sept 2014
20.
21. Could we have identified the
Ebola breakout earlier and
contained it?
22. The Key to going Descriptive:
1) Questions: You need the right questions
2) Data: You need the data to answer these
questions
3) Visualization: Build the right visualization to
answer the questions
29. Based on the global oil production
and storage-levels, I predict at
least a 30% decrease in Oil price
in the next few months.
30. Based on the hospital data in the
field in Africa, human health
history in the region, the
likelihood of an Ebola outbreak is
high
31. The Key to going Predictive:
1) Descriptive Analytics: You need to know
what’s happening before predicting.
2) Questions: You need the right questions
3) Data: Lots of data to answer these questions
4) Models: Predictive models on the key features
5) Operationalization: Models to production
6) Visualization/Experience
35. Experiment to Operationalization
Score /
Evaluate
Model
Train
Insights
Visualizations
Enterprise LOB
Other databases
LOB App
ML Process
Data Pipeline
Data Sources
Data
Experimentation
Operationalization Operationalized Models
Model
Train
Score/Evaluate
36. Main Classes of Learning Problems
• Classification (Supervised): Assign a category to each
item (Good | Bad | Neutral).
• Regression (Supervised): Predict a continuous value
for each item (price, currency, temperature).
• Clustering (Unsupervised): Partition items into
homogeneous groups (clustering twitter posts by
topic).
36
41. Helps you in better decision
making and recommends
actions
42. I am anticipating a “C” grade and
therefore:
1) Study these key definitions
2) Common questions from this text on
the internet
3) List of questions asked by the teacher
in the past 2 years
43. Oil Prices are going to fall
1) Buy Puts on the Oil stock
2) Once the stock falls 50%, buy
calls
44. Ebola outbreak is anticipated:
1) Increase the vaccine production by
20%
2) Alert WHO key communication
channels
3) Alert Health volunteer channels
45. The Key to going Prescriptive:
1) Descriptive & Predictive Analytics
2) Questions: You need the right questions
3) Outcomes: Desired and Possible Outcomes
4) Data: Lots of data to answer these questions
5) Models: Predictive and Prescriptive models
6) Operationalization: Models to production
7) Visualization/Recommendations
51. I want an A+ in Social Studies,
could you please appear for
my exam?
52. I want to invest $100,000 with
an expected ROI of at least
20%. Invest it for me.
53. WHO Robot: “I am predicting
an Ebola outbreak and I am
requesting an increase in
vaccine production and
alerting all the health
communication channels”
54. The Key to going Decisive:
1) Descriptive, Predictive & Prescriptive
2) Questions: You need the right questions
3) Outcomes: Desired and Possible Outcomes
4) Data: Lots of data to answer these questions
5) Models: Prescriptive models on key features
6) Operationalization: Models to production
7) Experience: Automation, Bots, etc.
60. Presentation
and action
-
Search and query
Data analytics (Excel)
Web/thick
client dashboards
Storage
(Long and Short)
Relational
Cloud
Azure Storage
AWS S3
NO SQL
Data pipeline
Transformation
Real-time stream
analytics
Batch Analytics
Ingestor
(broker)
Scalable
event Broker
Field Gateways
Collection
Cloud Gateways
(Cloud Collectors)
Applications
Devices
61. Do you have Big Data?
Megabytes
Gigabytes
Terabytes
Petabytes
Purchase detail
Purchase record
Payment record
ERP
CRM
WEB
BIG DATA
Offer details
Support Contacts
Customer Touches
Segmentation
Web logs
Offer history
A/B testing
Dynamic Pricing
Affiliate Networks
Search Marketing
Behavioral Targeting
Dynamic Funnels
User Generated Content
Mobile Web
SMS/MMSSentiment
External Demographics
HD Video, Audio, Images
Speech to Text
Product/Service Logs
Social Interactions & Feeds
Business Data Feeds
User Click Stream
Sensors / RFID / Devices
Spatial & GPS Coordinates
Increasing Data Variety and Complexity
Transactions +
Interactions +
Observations
= BIG DATA
62. Patterns Across Verticals and Business Cases
Vertical Refine Explore Enrich
Retail & Web • Log Analysis/Site Optimization
• Loyalty Program Optimization
• Brand and Sentiment Analysis
• Market basket analysis
• Dynamic Pricing
• Session & Content Optimization
• Product recommendation
Telco • Customer profiling • Equipment failure prediction • Location based advertising
Government • Threat Identification • Person of Interest Discovery • Cross Jurisdiction Queries
Finance • Risk Modeling & Fraud Identification
• Trade Performance Analytics
• Surveillance and Fraud Detection
• Customer Risk Analysis
• Real-time upsell, cross sales marketing offers
Energy • Smart Grid: Production Optimization
• Grid Failure Prevention
• Smart Meters
• Individual Power Grid
Manufacturing • Supply Chain Optimization • Customer Churn Analysis
• Dynamic Delivery
• Replacement parts
Healthcare • Electronic Medical Records (EMPI)
• HL7
• Clinical decision support
• Clinical Trials Analysis
• Insurance Premium Determination
• Targeted Subscriber Communication