Free guide to creating high-impact #analytics for any organization, large or small. Includes links to over 200 free resources, including books and software. Useful for data mining, data science, machine learning and predictive analytics.
2. Scott.Clendaniel@MktgSciences.com
The steps you need to
follow for analytics
success
2- Step-by-Step
Some special methods for
maximizing results
3. Tips & Tricks
The ROI of Analytics.
1- ROI
Time for questions from the
audience.
5. Questions
A collection of free resources for
High-Impact Analytics for
Mortals!.
4. Free Resources
AGENDA
3. Scott.Clendaniel@MktgSciences.com
WHERE HAVE THESE TIPS WORKED?
Past clients and employers
IMPORTANT: All views expressed are solely my own, and should not be taken
as being those of current or past employers, clients or others.
6. Scott.Clendaniel@MktgSciences.com
TIPS SOURCES
Where do the recommendations originate?
197 Kaggle Winner
Interviews
How did they win?
50 In-depth Case
Studies
Which factors mattered
25,000 Head-to-Head
Tests
What made the difference?
9. Scott.Clendaniel@MktgSciences.com
TWO PURPOSES FOR ANALYTICS
If your analytics don’t meet one or both, you’re doing analytics wrong.
ANSWERS, “SO WHAT?”
Why are your analytics important?
BOOSTS ROI…
…or whatever your key object might be.
10. Scott.Clendaniel@MktgSciences.com
SOURCE: Nucleus Research, 2014
http://nucleusresearch.com/research/single/analytics-pays-back-13-01-for-every-dollar-spent/
HIGH-IMPACT ANALYTICS: The Hard ROI Numbers
Does this sound familiar to anyone?
12. Scott.Clendaniel@MktgSciences.com
HIGH-IMPACT ANALYTICS: Stage 1 ROI
Stage 1: Automating Reports (ROI of 188%)
188%
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
200%
Stage 1- Automating Reports
Analytics ROI by Stage
SOURCE: Nucleus Research, 2012
http://nucleusresearch.com/press/roi-of-business-analytics-increases-significantly-as-solution-matures/
13. Scott.Clendaniel@MktgSciences.com
HIGH-IMPACT ANALYTICS: Stage 2 ROI
Stage 2: Improve Decision-Making (ROI of 389%)
188%
389%
0%
50%
100%
150%
200%
250%
300%
350%
400%
450%
Stage 1- Automating Reports Stage 2- Improve Decision Making
Analytics ROI by Stage
SOURCE: Nucleus Research, 2012
http://nucleusresearch.com/press/roi-of-business-analytics-increases-significantly-as-solution-matures/
14. Scott.Clendaniel@MktgSciences.com
HIGH-IMPACT ANALYTICS: Stage 3 ROI
Stage 3: Strategic Analytics (ROI of 968%)
188%
389%
968%
0%
200%
400%
600%
800%
1000%
1200%
Stage 1- Automating
Reports
Stage 2- Improve Decision
Making
Stage 3- Strategic Analytics
Analytics ROI by Stage
SOURCE: Nucleus Research, 2012
http://nucleusresearch.com/press/roi-of-business-analytics-increases-significantly-as-solution-matures/
15. Scott.Clendaniel@MktgSciences.com
HIGH-IMPACT ANALYTICS: Stage 4 ROI
Stage 4: Extended to Partners and Social Media (ROI of 1209%)
188%
389%
968%
1209%
0%
200%
400%
600%
800%
1000%
1200%
1400%
Stage 1- Automating
Reports
Stage 2- Improve
Decision Making
Stage 3- Strategic
Analytics
Stage 4- Partners &
Social Media
Analytics ROI by Stage
SOURCE: Nucleus Research, 2012
http://nucleusresearch.com/press/roi-of-business-analytics-increases-significantly-as-solution-matures/
16. Scott.Clendaniel@MktgSciences.com
SOURCE: Barn Raisers, 2015
http://barnraisersllc.com/2015/11/16-case-studies-companies-proving-roi-of-big-data/
HIGH-IMPACT ANALYTICS: The Bad News
A full one-third of companies are losing money on Big Data initiatives
17. Scott.Clendaniel@MktgSciences.com
WHAT DRIVES THE ROI DIFFERENCES?
• Senior executive project champion
• Low-hanging fruit
• Clear, defined outcomes
• “Phased-Gate” approach to investments:
focus on “Do what you can, with what
you have, where you are right now.”
(Quote: Theodore Roosevelt)
POSITIVE ROI
Success Factors
• Large initial capital investments
• No clear financial targets
• “Swing for the fences”
• Complex initial projects
NEGATIVE ROI
Failure Factors
19. Scott.Clendaniel@MktgSciences.com
Create a list of potential projects and
targets. Brainstorm with decision-
makers to get initial buy-in.
STEP 1. Pick targets.
Check to see which targets have:
• Either/ or outcomes
• Reliable, detailed data
• Significant potential organizational
impact
STEP 2. Evaluate targets.
Use data preparation process to
identify both a clear outcome and
potential inputs. This is often the key to
success or failure!
STEP 3. Wrangle the data.
Be ruthless in determining what makes
sense and what will have impact.
STEP 5. Evaluate results.
Here’s where the power of analytics
can have the most impact- use pivot
tables and/ or predictive models.
STEP 4. Pivot and/ or model.
STEP-BY-STEP: The “No-Fail” Checklist
Features
21. Scott.Clendaniel@MktgSciences.com
• Rationale:
• Expands list of
available algorithms
• Reduces complexity
• Reduces error
• Increases
interpretability
• Methodology:
• Presence of condition
being studied (fraud,
response, attrition)
caries value of 1,
others carry 0
• Convert continuous
dependent variables
into thresholds/ bins/
buckets
• If dependent variable
is missing, remove the
record
TIP: Convert Problem to Either/ Or Outcome
22. Scott.Clendaniel@MktgSciences.com
• Rationale:
• Algorithms are “stupid
and greedy”
• They need to focus on
what you don’t
already know, or you
don’t need an
algorithm
• If you don’t account
for this step, you will
never find the really
“interesting” patterns.
• Methodology:
• ALWAYS run EDA
(Exploratory Data
Analysis first.
• Remove cases where
you “know” the
outcomes with a
reasonable certainty.
TIP: Follow The Painfully Obvious Theorem
23. Scott.Clendaniel@MktgSciences.com
• Rationale:
• Models work best
identifying coherent
patterns leading to a
single outcome.
• If widely different
patterns lead to the
same outcome, you
can “confuse” the
model.
• Separating the use
cases solve this issue.
• Methodology:
• Run a tree-based
classifier when
beginning a project.
• If your tree uses
completely different
variable classes after
the first splitter,
consider creating
multiple models.
TIP: Confirm The Number of Problems You’re Solving
24. Scott.Clendaniel@MktgSciences.com
• Rationale:
• Models are useless if
you can’t apply them.
• Don’t engage in a
modeling strategy that
uses data, processes
or target variables
that will prevent you
from using what you
learned.
• It is much easier to fix
this on the front end.
• Methodology:
• Get an understanding
of how you want to
use your model first.
• Work backward from
those constraints to
create your modeling
strategy.
• Hold planning
meetings with
stakeholders BEFORE
modeling.
TIP: Begin With The End In Mind
25. Scott.Clendaniel@MktgSciences.com
• Rationale:
• Complex systems have
many more mail fail
points than simple
ones.
• If you don’t
understand it when it
works, how will you
fix it when it breaks?
• Methodology:
• Get an understanding
of how you want to
use your model first.
• Work backward from
those constraints to
create your modeling
strategy.
• Hold planning
meetings with
stakeholders BEFORE
modeling.
TIP: Aim for Simplicity
29. Scott.Clendaniel@MktgSciences.com
FREE RESOURCES: 88 Data Science Resources PDF
PDF of other free resources
SOURCE: Elite Data Science
https://elitedatascience.com/wp-content/uploads/2016/11/Supercharge-Your-Data-Science-Career-1.pdf