How to Talk to your Boss about Analytics

1
How to Talk to your Boss about Analytics

• Presenter
• James Parry
• Sr. Systems Engineer
• SPSS Inc.

2
Are these your senior executives speaking?
There are many methods for predicting the
future. For example, you can read horoscopes, tea
leaves, tarot cards, or crystal balls.
Collectively, these methods are known as "nutty
methods." Or you can put well-researched facts
into sophisticated computer models, more commonly
referred to as "a complete waste of time."
Scott Adams, The Dilbert Future
3
Why predictive analytics is not used in many
organizations?
The entry barrier is no longer technology, but
whether you have executives who understand
this Thomas Davenport, Competing on Analytics
4
Agenda

• Why data mine Demystifying and myth busting
• Four steps to planning and presenting your data
  mining project plan
• Reporting
• Conveying the strength of a data mining model
• What is lift?
• Considerations for efficient reporting
• Tips for when talking to your boss about data
  mining
• Q A
• Close

5
Demystifying and myth busting
6
Myth 1 Its not for me

• Predictive analytics is rocket science its way
  above and beyond what I need to do.

7
Analytics is now a hit in the Top 50
Best-Selling Business Books
8
And is catching on in institutional fundraising
as well
9
Predictive analytics becomes mainstream
10
Myth 2 I dont understand it.

• The idea of predictive analytics sounds good,
  but I really dont understand what it does, and I
  couldnt possibly explain it to anyone else to
  get their buy-in.

11
Predictive Analytics Defined

• Data driven approach to problem solving
• Focused on business objectives
• Leverages organizational data
• Uncovers patterns using predictive and
  descriptive techniques
• Uses results to help improve organizational
  performance

12
What Does Predictive Analytics Do?

• Predictive Analytics uses existing data to
• Predict
• Group
• Associate
• Find outliers

13
Predictive Analytics What it isnt

• A product, a particular piece of software, or a
  given algorithm
• It is a business process that is enabled by
  technology
• A model, segmentation scheme or business rules
• Those are some outputs from the Predictive
  Analytics process
• It is a method of discovery that yields
  information and insight leading to some action
• An end product in and of itself
• It is a means of harnessing the insight often
  trapped in large masses of data
• It is an iterative, ever improving, feedback
  cycle
• A SQL query, an OLAP hub, or a BI Dashboard
• Statistics per se

14
Predictive Analytics is Part of CRISP-DM,
the Industry Standard

• Phases
• Business Understanding
• Data Understanding
• Data Preparation
• Modeling
• Evaluation
• Deployment

15
Myth 3 Ive got one already!

• We already do analytics through our business
  intelligence tools and corporate dashboards.

16
Key Differences between BI and Predictive
Analytics (PA)

• BI supplies the core facts of an organization
• Core business metrics
• KPIs
• Factual reporting
• PA helps you to interpret these facts as
  actionable information
• Predictive associations
• Optimized models
• Causal reporting
• Key Performance Predictors

17
Strategic Viewpoint Differences between BI and PA

• Typical BI applications provide a great picture
  of what has happened
• a rear view perspective
• Dashboards in real time show current conditions
  and metrics
• a clear windshield view
• Predictive analytics enables future views and
  forecasting
• a peek around the approaching corner
• and can create new metrics for closing the
  feedback loop into the BI system

18
Myth 4 It wont pay off

• Our organization is under constant pressure to
  lower the amount spent to raise a dollar.
  Predictive analytics will never pay back in time
  to make a real impact on our campaigns.

19
Predictive analytics is important because it
delivers value

• The median ROI for the projects that
  incorporated predictive technologies was 145,
  compared with a median ROI of 89 for those
  projects that did not.
• Source IDC, Predictive Analytics and ROI
  Lessons from IDCs Financial Impact Study

20
Nucleus Research . . .

• Nucleus Research The Real ROI from SPSS Inc.
• 94 of customers achieved a positive ROI, with an
  average payback period of 10.7 months
• Key benefits achieved include reduced costs,
  increased productivity, improved customer
  employee satisfaction, and greater visibility
  into operations
• 81 of projects deployed on time, 75 on or under
  budget

This is one of the highest ROI scores Nucleus
has ever seen in its Real ROI series of research
reports. Rebecca Wettemann, Vice President of
Research, Nucleus Research
21
Why is Predictive Analytics so critical to
business decisions?
Beforeanalytics
Afteranalytics 21 18 10 60
Banner ad click through rates 0.3 Mail
response rates 0.5 Conversion rates
(post-response) 0.9 Buyer repeat rates 2.0

• Performance of analytics targeted to certain
  consumers cross-industry and channel, research
  from Forrester, Jupiter, Amazon.com and Ovum (DM
  Review, Feb 11, 2003)

22
Four steps to planning and presenting your data
mining project plan
23
Step 1 Determine Business Objectives

• Thoroughly understand what you want to accomplish
• Describe the criteria for a successful or useful
  outcome to the project from a business point of
  view
• EG Increase the number of transfers from low to
  medium donation groups.

24
Step 2 Assess Your Situation

• Create an inventory of your available resources,
  including

Computing Resources
Software
Personnel
Data
25
Step 3 Determine Data Mining Goals

• Describe the intended project outputs and how you
  will arrive at them
• Business goals vs. Data Mining Goals
• Example business goal Increase the average gift
  amount among annual fund donors by X.
• Corresponding data mining goal Predict the
  propensity of annual fund donors to give more
  than they gave last year, using their giving
  history, demographic information, and stated
  level of satisfaction with your advancement
  program.

26
Step 4 Prepare and Present Your Project Plan

• List and describe each project stage, including
• Whos involved?
• What other resources are required?
• What is the outcome or objective?
• When will it be completed?
• Remember to include in your plan specific points
  in time to regroup and review progress and make
  updates as necessary

27
Create and follow a strategic plan to secure
executive buy-in- Recap

• Determine Business Objectives
• Assess your Situation
• Determine Data Mining Goals
• Present your Project Plan

28
Data Mining and Reporting
29
Generated Models
The gold nuggets.
29
30
Reporting Considerations

• Visually Explaining Competing Models
• Model lift
• Eliminating Tedious, Repetitive, Time-Consuming
  Edits (3 Ds . . .)
• Design reusable graphs and graph templates
• Getting the Right Information into the Right
  Hands, Securely
• Socializing/Publishing results - quickly
• Self-Service Reporting Portal
• Create secure, online reporting environment
• Place the onus on the end-user, not the analyst

Automate!!
31
Data Mining Whos Involved?

• The Power User
• More hands-on
• Understand how to connect to the data
• Understands data preparation
• Creates Report Templates
• Ad-Hoc Reporter/Analyst
• Runs graphs and tables upon request (many, many)
• Socializes/Publishes Results
• Consumer
• Usually stake-holder or C-level
• Does not license desktop application
• Relies on thin client

32
After you run some models . . . then what?
33
Measuring Lift
34
The Perfect Model Doesnt Exist, But
The perfect model
34
35
Further Comparison Business Rules
Business rules
35
36
Picking Our Model
Compare the C5.1 decision tree model to the
others at the 40th percentile engagement point.
36
37
Presenting the Results
PASW Statistics Base
PASW Collaboration Deployment
Services (Predictive Enterprise Browser)
PASW Modeler
38
Design a Template (Analyst/IT)
39
Pre-Template Chart
40
Post-Template Chart
41
Post-Template Chart
42
SPSS User Publishes to Web
43
Consumer Log-in
44
Predictive Enterprise Browser
45
Predictive Enterprise Browser
46
Results Rendered in Browser
47
Reporting Recap

• Model Lift conveys in why using a predictive
  algorithm makes sense.
• Graph Templates decrease busy work, save in
  efficiency
• Publishing to the Web
• Self-Service Reporting Platform takes the
  burden off the IR office thus making it more
  efficient

48
Additional tips for talking to your boss about
data mining
49
Laying the communication groundwork

• There is a communication gap between the analyst
  (the maker) and the executive (user)
• Consumer of analytics is usually non-technical
  prefers simple answers to complex explanations
• Analyst methods are treated like a black box of
  information or voodoo but now more than ever,
  analysts are being called upon to explain how
  they arrived at an answer

50
Important first steps

• Set proper expectation levels as soon as possible
• Bosses can have expectations which are too high
  Its magic and will work perfectly
• They need to be brought down to earth before they
  get disappointed and it reflects negatively on
  you
• Bosses can also have mistakenly low expectations
• They dont realize the potential of powerful
  analytics and set their sights to low to
  demonstrate significant impact

51
Remember the audience at all times

• Make all output relevant to the consumer
• Use business terms, not math, tech, stat verbiage
• Use graphs not words
• Turn everything into prospects or dollars
• Place everything into a problem-solving context
• Consider the price of inaction or not knowing

52
Words to avoid at all costs
R-squared

• Logistical regression

• Algorithm

Neural networks
Hierarchical clustering
Coefficient
53
Words to use frequently
Growth

• Efficiency

Stewardship

• ROI

Affinity

• Prospects

• YIELD

Cost reduction
Capacity Ranking
54
You are not alone in the struggle

• Look beyond your own domain
• Other departments within your institution may
  already be employing predictive analytics and/or
  using SPSS solutions.
• List-servs and professional groups such as
  Prospect DMM, APRA, and CASE, AACRAO, AIR.
• Befriend the IT organization
• Bridge the gap between data expertise and domain
  expertise
• Involve IT to align goals and communicate needs

55
Over-arching principles

• Demystify
• Others are doing it
• It has been proven
• You can do it in small bites
• Have a strong plan in place before you start!
• Seek help

56
Questions?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
57
Key Take-aways

• Remove the jargon and rocket science
• Stay focused on the goal or business objective
• Use external sources as support
• Automate insight
• Identify internal allies

58
Contact Information
James Parry Sr. Systems Engineer SPSS Inc. P.
800.543.2185 extension 2092 e-mail
jparry_at_spss.com website www.spss.com