T. Scott Clendaniel provides a complete guide to creating and implementing your analytics strategy. Includes the Accenture model, a wide variety of Tips and Tricks, and several bonuses in the appendix.
5. SOURCE: Nucleus Research, 2014
http://nucleusresearch.com/research/single/analytics-pays-back-13-01-for-every-dollar-spent/
ANALYTICS: THE PROMISE OF PROFITABILITY
Few technologies offer the ROI potential of Artificial Intelligence
6. ANALYTICS: THE PROMISE OF PROFITABILITY
https://mms.businesswire.com/media/20200310005668/en/778810/5/IRTNTR30994.jpg
7. ANALYTICS: THE PROMISE OF PROFITABILITY
https://www.amclaboratories.com/wp-content/uploads/2019/11/blog_04_top.jpg
8. ANALYTICS: THE PROMISE OF PROFITABILITY
https://www.accenture.com/us-en/insight-artificial-intelligence-future-growth
13. TOP ANALYTICS TIPS AND TRICKS
These tips will help ensure implementation success.
14. TIP: HOW MOST PEOPLE VIEW DATA AND
ANALYTICSThis is a good example of how pushing “data” can make people feel.
15. TIP: THE ONE QUESTION ANALYTICS PROJECTS
MUST ANSWER1. Almost all A.I. projects attempt to answer basically the same question
Credit:
16. • Analyze organization:
o Background and history
o Primary objectives
o Project sponsors, beneficiaries and chain of
approval
o Understand prior efforts
o Define constraints
• Determine and prioritize challenges
o Revenue/ expense/ insight
o Estimate resources and feasibility
o Identify s
TIP: IDENTIFY THE PROBLEM
17. TIP:
BEGIN WITH THE END IN MIND
These tips will help ensure implementation success.
https://www.behance.net/gallery/26182691/Summary-of-
Stephen-Covey-bestseller-7-habits
18. TIP: IDENTIFY YOUR STAKEHOLDERS’ NEEDS
AND GOALS
Hidden Secret: AI and models are about people, not technologies
19. TIP: STOP CALLING IT “DATA-DRIVEN”
Consider “insights-enabled,” “data-assisted,” etc.
https://www.amaboston.org/blog/three-tips-for-driving-better-insights/
21. TIP: ENABLE PEOPLE, NOT DATA
No one wants to lose control of their work.
https://authorbeckyjohnen.files.wordpress.com/2015/05/control-illusion.jpg
22. TIP: LESS “PUSH,” MORE “PULL”
People support what they create- so help them create positive results with
data!
https://www.amaboston.org/blog/three-tips-for-driving-better-insights/
23. TIP. IDENTIFY GOALS AND GUARD RAILS FIRST
People don’t want a ¼” drill bit- they want a ¼” hole.
GoalGuard Rail Guard Rail
24. TIP: PHASE GATE APPROVALS
Avoid the “Big Reveal” syndrome- obtain approvals at each step.
https://www.iamip.com/news/blog/successful-development-process
25. TIP. PRESENT RESULTS USING THE FIRE ALARM
RULE
Hint: Make “executive summaries” your friends.
27. • The Value Chain for Machine Learning depends
on driving better actions through insights.
• Following a standard process:
o saves time
o reduces error
o allows repeatability
o builds trust with stakeholders
• The process outlined here reflects best practices
from past industry projects such as CRISP-DM,
SEMMA and Microsoft's Team Data Science
Process (TDSP).
• Core stages of the process are:
o Identify/ Formulate Problem
o Data Preparation
o Data Exploration
o Transform and Select
o Build Models
o Ensemble/ Validate Models
o Deploy Models
o Evaluate/ Monitor Results
OVERVIEW
28. R5. UNDERSTAND “MODEL” VS. “MAGIC”
They’re both 5-letter words beginning with the letter “m,” but they’re not the same
29. R7. SAVE $2,500.00 PER EMPLOYEE ON
TRAINING
https://www.kdnuggets.com/2018/11/10-free-must-see-courses-machine-learning-data-science.html
30. R8. SAVE $400.00 PER EMPLOYEE ON DATA
SCIENCE BOOKS
https://www.learndatasci.com/free-data-science-books/
31. DATA “EXPERTS” DON’T ALWAYS MAKE THINGS
EASIER“Rules of Evil Programmers” would be an example of making things
“incomprehensible.”
RULE 1- “’Real’ programmers don’t
document their code.”
RULE 2- “If it was really hard to write,
it should be really hard to read.”
RULE 3- “If it was supposed to be
easy, we wouldn’t be calling it code.”
Credit: https://images-na.ssl-images-amazon.com/images/I/61jq5MuWT9L._SX679_.jpg
32. CULTURE EATS STRATEGY FOR BREAKFAST,
PART 1
https://theironicmanager.com/blog/culture-eats-strategy-for-breakfast-doesn-t-it
33. 33
TIP 6: STOP MAKING THINGS SO COMPLICATED!
Simplification is important.
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.
34. Moving Away From… Moving Toward…
TIP 12: THINK “WATCHMAKER,” NOT “ASSEMBLY
LINE”
35. GARTNER’S ARTIFICIAL INTELLIGENCE HYPE
CYCLE, PART 1
The starting point for Artificial Intelligence was “Innovation Trigger” & “Peak of Inflated Expectations…
Source: https://blogs.forbes.com/louiscolumbus/files/2019/09/Gartner-Hype-Cycle-For-Artificial-Intelligence-2019.jpg
36. EXAMPLE TASK LIST, DIVIDED BY SAS
LIFECYCLE, PART 2
Unfortunately, those two steps are followed by the “Trough of Disillusionment”
Source: https://blogs.forbes.com/louiscolumbus/files/2019/09/Gartner-Hype-Cycle-For-Artificial-Intelligence-2019.jpg
38. TIP 5: PLANNING FOR OBSOLESCENCE
Understand and accept the models “drift,” or decay, over time and plan for it up-front.
39. TIP 6: PLAN AROUND THE ANALYTICS MATURITY
MODEL
You need to match your implementation complexity to the organization’s maturity level.
https://www.gartner.com/smarterwithgartner/the-cios-guide-to-artificial-intelligence/
40. TIP 7: VALIDATION OVERKILL
Hope for the best, prepare for the worst, and measure everywhere.
1. Hold-Out
Sample
Validation
2. Cross-
Validation
5. Recruit
Validation
4. “Silent”
Validation
6. Roll-Out
Validation
7. Non-
Stoptimization
3. Hold-Out
Timeframe
Validation
41. TIP 8: SIMPLICITY AND FLEXIBILITY ARE YOUR
ALLIES
Complexity sows the seeds of its own destruction.
42. TIP 9: CREATE A CHAMPION/ CHALLENGER
TESTING APPROACHhttps://powerdigitalmarketing.com/blog/multivariate-vs-b-testing/#gref
43. TIP 10: WHAT IS YOUR “UNDO” PLAN?
Always, always, ALWAYS test your “undo” plan BEFORE you implement.