Data Foundation for
Analytics Excellence
Cathy Tanimura
Director of Analytics & Big Data @ Okta
ctanimura@okta.com
Agenda
• Intro
• Data Foundation
• Finding the Problem(s)
• Getting Started: Proof of Concept
• Picking the Technology
• Building Out: What to Expect
• People Foundation
• Building the Team
• Partners and Champions
• Bringing it All Together
Intro
Background
Okta?
“In meteorology, an okta is a unit
of measurement used to describe
the amount of cloud cover at any
given location such as a weather
station.
Sky conditions are estimated in
terms of how many eighths of the
sky are covered in cloud, ranging
from 0 oktas (completely clear
sky) through to 8 oktas
(completely overcast).”
- Wikipedia
4 Million+
People
10 Million+
Devices
The Enterprise Identity Network
3,000+
Applications
OnPremCloudMobile
1,600+ Organizations
Problems Okta Solves
• User Password Fatigue
• Failure-Prone Manual Provisioning & De-Provisioning
Process
• Compliance Visibility: Who Has Access to What?
• Siloed User Directories for Each Application
• Managing Access across an Explosion of Browsers and
Devices
• Keeping Application Integrations Up to Date
• Different Administration Models for Different
Applications
• Sub-Optimal Utilization, and Lack of Insight into Best
Practices
Focus on the end-user
Data Foundation
Data Foundation
•Finding the Problem(s)
•Getting Started: Proof of Concept
•Picking the Technology
•Building Out: What to Expect
Finding the Problem
•First thing you want to tackle
•Prove value
•Research for long-term infrastructure
What Makes a Good Problem
•Big business impact: $$’s, time
•Data available
•Someone has tried to tackle
•Engaged business partner
•Clear vision of what will change
Common “Problems”
•Marketing optimization
•Multi-channel attribution
•User behavior
•Fraud detection
•Recommendations
•Viral / market penetration
•Retention / churn
•Resource allocation
Finding the Problem
Finding the Problem
• “Virals” were major
growth and retention
tool
• How many new users
did we attract?
• How many came back?
• How effective was this
feature at driving
traffic?
• How does play spread
from friend to friend?
Finding the Problem
Activities:
• Add directory
• Import users
• Add apps
• Assign users
• Rollout plan
Adoption
Finding the Problem
Why do we care about adoption?
• Happy customers  renewals,
references, upsell opportunities
Sub-Problems:
• How many customers?
• Does it really affect churn?
• Can we influence?
Proof of Concept
•Find the data
•Simple, low cost tools
•Build something
•Get feedback
POC: Find the Data
Social
Cloud Apps
In-house Apps
On-Prem Databases
3rd party
Finding the Data Example
POC: Simple, low-cost tools
•What do you already have
•Open-source
•Trials / community editions
POC: Example Data Infrastructure
Building
•Define the metrics
• Understandable
• Measurable
• Actionable
•Visualize
Building the Metric: Example
• At a high level, Adoption = Usage / Entitlement
• What is the best “usage” measure?
Showing the Metric Matters
• Some outliers, but adoption correlated with renewal
Get Feedback
•Share
•Listen
•Pay attention to where the data doesn’t
fit the “smell test”. At first your clients
will have a better sense than you do
Feedback: Prototype Example
Pick the Technology
•The fun part (sort of)
•Start with requirements discovered
during POC
•Be aware of the market, but not
distracted
Data Store Decisions
Vs.
Vs.
ETL Decisions
Front-End Options
Tips on Selling the Technology
•Educate: what does each piece do (in
layman’s terms)
•Present S,M,L cost options
Data Mining,
Modeling, Stats
BI ToolsSource Systems
Operational
Systems (“Prod”)
Cloud
Services
Web Data
External
Data
Data StorageETL /
Data Integration
Streaming, Event
Processing
“End to End”
Analysis, Viz
Data Warehouse
(SQL)
Hadoop Platforms
Point Solutions
Example: Tech & Vendor Landscape
Example: S,M,L Options
Small
• $0k
• 0 extra FTE
• Rely on forums,
learn as we go
• Timeline: 12+
Months
Medium
• $100k
• 1 FTE
• Access to
expertise
• Timeline: 6-9
Months
Large
• $200k
• 2 FTE
• Access to
expertise
• Timeline: 3 – 6
Months
Building Out: What to Expect
•It will never go “as expected”
•Time will be more than expected
•$ will be more than expected
 Develop the vision up-front, fill in
details as you go
 Consider Agile development
Building Out: What to Expect
•Stuff that happens:
•People change
•New data source
•Holidays & vacations
•Integrations break
•Data quality
What to Expect
You never “finish” analytics…
Known
Knowns
Easy stuff
Unknown
Knowns
Duh
Unknown
Unknowns
Uh-oh
Known
Unknowns
Aha!
People Foundation
Building the Team
•Who
•When
Building a Team
Who
Data Analyst
Focus:
• Analysis,
reports,
dashboards
Aligned to:
• Business
Languages:
• SQL, R, Excel
Data Scientist
Focus:
• Data products
• Modeling
Aligned to:
• Product
Languages:
• R, Python, SQL
Data Engineer
Focus:
• Data
infrastructure
• Scalability
Aligned to:
• Engineering
Languages:
• Java, Python,
MapReduce
When to Build the Team
Delphi Analytics, April 1, 2013
When to Build the Team
•Scale with business
•Infrastructure in place
•Generate demand from clients
Partners & Champions
•Easily overlooked but key to success
•Partners are your clients
• Typically Marketing, Finance, Product,
BizDev
•And the teams you rely on
• IT, Engineering, Product
Partners & Champions
•Champions are execs and people on the
ground who can spread the word
• Execs want clear and simple messages:
what are the benefits, how much will it
cost
• You never know who your other
champions are going to be. Don’t miss
opportunities to help people out
Putting It All Together
Tech Stack - Vision
What are the Effects?
• Time savings
• Time spent collecting & processing data by Customer
Success, Renewals, Product
• Time spent telling anecdotes
• Revenue:
• Save at-risk renewals: early awareness tells us where to
intervene
• Upsells: Visibility into usage lets sales people have more
timely & informed discussions about upsells
• Focus
• On the features that matter (not ones that don’t)
• Take the guesswork out of meetings
Questions?

Data Foundation for Analytics Excellence by Tanimura, cathy from Okta

  • 1.
    Data Foundation for AnalyticsExcellence Cathy Tanimura Director of Analytics & Big Data @ Okta ctanimura@okta.com
  • 2.
    Agenda • Intro • DataFoundation • Finding the Problem(s) • Getting Started: Proof of Concept • Picking the Technology • Building Out: What to Expect • People Foundation • Building the Team • Partners and Champions • Bringing it All Together
  • 3.
  • 4.
  • 5.
    Okta? “In meteorology, anokta is a unit of measurement used to describe the amount of cloud cover at any given location such as a weather station. Sky conditions are estimated in terms of how many eighths of the sky are covered in cloud, ranging from 0 oktas (completely clear sky) through to 8 oktas (completely overcast).” - Wikipedia
  • 6.
    4 Million+ People 10 Million+ Devices TheEnterprise Identity Network 3,000+ Applications OnPremCloudMobile 1,600+ Organizations
  • 7.
    Problems Okta Solves •User Password Fatigue • Failure-Prone Manual Provisioning & De-Provisioning Process • Compliance Visibility: Who Has Access to What? • Siloed User Directories for Each Application • Managing Access across an Explosion of Browsers and Devices • Keeping Application Integrations Up to Date • Different Administration Models for Different Applications • Sub-Optimal Utilization, and Lack of Insight into Best Practices
  • 8.
    Focus on theend-user
  • 9.
  • 10.
    Data Foundation •Finding theProblem(s) •Getting Started: Proof of Concept •Picking the Technology •Building Out: What to Expect
  • 11.
    Finding the Problem •Firstthing you want to tackle •Prove value •Research for long-term infrastructure
  • 12.
    What Makes aGood Problem •Big business impact: $$’s, time •Data available •Someone has tried to tackle •Engaged business partner •Clear vision of what will change
  • 13.
    Common “Problems” •Marketing optimization •Multi-channelattribution •User behavior •Fraud detection •Recommendations •Viral / market penetration •Retention / churn •Resource allocation
  • 14.
  • 15.
    Finding the Problem •“Virals” were major growth and retention tool • How many new users did we attract? • How many came back? • How effective was this feature at driving traffic? • How does play spread from friend to friend?
  • 16.
    Finding the Problem Activities: •Add directory • Import users • Add apps • Assign users • Rollout plan Adoption
  • 17.
    Finding the Problem Whydo we care about adoption? • Happy customers  renewals, references, upsell opportunities Sub-Problems: • How many customers? • Does it really affect churn? • Can we influence?
  • 18.
    Proof of Concept •Findthe data •Simple, low cost tools •Build something •Get feedback
  • 19.
    POC: Find theData Social Cloud Apps In-house Apps On-Prem Databases 3rd party
  • 20.
  • 21.
    POC: Simple, low-costtools •What do you already have •Open-source •Trials / community editions
  • 22.
    POC: Example DataInfrastructure
  • 23.
    Building •Define the metrics •Understandable • Measurable • Actionable •Visualize
  • 24.
    Building the Metric:Example • At a high level, Adoption = Usage / Entitlement • What is the best “usage” measure?
  • 25.
    Showing the MetricMatters • Some outliers, but adoption correlated with renewal
  • 26.
    Get Feedback •Share •Listen •Pay attentionto where the data doesn’t fit the “smell test”. At first your clients will have a better sense than you do
  • 27.
  • 28.
    Pick the Technology •Thefun part (sort of) •Start with requirements discovered during POC •Be aware of the market, but not distracted
  • 29.
  • 30.
  • 31.
  • 32.
    Tips on Sellingthe Technology •Educate: what does each piece do (in layman’s terms) •Present S,M,L cost options
  • 33.
    Data Mining, Modeling, Stats BIToolsSource Systems Operational Systems (“Prod”) Cloud Services Web Data External Data Data StorageETL / Data Integration Streaming, Event Processing “End to End” Analysis, Viz Data Warehouse (SQL) Hadoop Platforms Point Solutions Example: Tech & Vendor Landscape
  • 34.
    Example: S,M,L Options Small •$0k • 0 extra FTE • Rely on forums, learn as we go • Timeline: 12+ Months Medium • $100k • 1 FTE • Access to expertise • Timeline: 6-9 Months Large • $200k • 2 FTE • Access to expertise • Timeline: 3 – 6 Months
  • 35.
    Building Out: Whatto Expect •It will never go “as expected” •Time will be more than expected •$ will be more than expected  Develop the vision up-front, fill in details as you go  Consider Agile development
  • 36.
    Building Out: Whatto Expect •Stuff that happens: •People change •New data source •Holidays & vacations •Integrations break •Data quality
  • 37.
    What to Expect Younever “finish” analytics… Known Knowns Easy stuff Unknown Knowns Duh Unknown Unknowns Uh-oh Known Unknowns Aha!
  • 38.
  • 39.
  • 40.
  • 41.
    Who Data Analyst Focus: • Analysis, reports, dashboards Alignedto: • Business Languages: • SQL, R, Excel Data Scientist Focus: • Data products • Modeling Aligned to: • Product Languages: • R, Python, SQL Data Engineer Focus: • Data infrastructure • Scalability Aligned to: • Engineering Languages: • Java, Python, MapReduce
  • 42.
    When to Buildthe Team Delphi Analytics, April 1, 2013
  • 43.
    When to Buildthe Team •Scale with business •Infrastructure in place •Generate demand from clients
  • 44.
    Partners & Champions •Easilyoverlooked but key to success •Partners are your clients • Typically Marketing, Finance, Product, BizDev •And the teams you rely on • IT, Engineering, Product
  • 45.
    Partners & Champions •Championsare execs and people on the ground who can spread the word • Execs want clear and simple messages: what are the benefits, how much will it cost • You never know who your other champions are going to be. Don’t miss opportunities to help people out
  • 46.
  • 47.
  • 48.
    What are theEffects? • Time savings • Time spent collecting & processing data by Customer Success, Renewals, Product • Time spent telling anecdotes • Revenue: • Save at-risk renewals: early awareness tells us where to intervene • Upsells: Visibility into usage lets sales people have more timely & informed discussions about upsells • Focus • On the features that matter (not ones that don’t) • Take the guesswork out of meetings
  • 49.