Advanced Analytics Primer
- 1. An Advanced Analytics Primer
For Everyone
Gordon MacMaster
© 2016 Gordon MacMaster All Rights Reserved
- 2. What is Advanced Analytics?
Simply put, Advanced Analytics is answering questions with data by using a
variety of mathematical, statistical and modelling techniques.
Advanced Analytics is not…
Magic - You do not pour data in and out comes answers.
Objective - Asking the question introduces bias. The data that exists may be
skewed or incomplete. While it is not purely objective the results are
supported.
Big Data - Data of all sorts is used. Big Data is a source of raw material.
© 2016 Gordon MacMaster All Rights Reserved
- 3. What are the
results?
Insight
Insight will not create change.
Actions taken based on the
results is required.
The primary result of any analytics
activity is insight. Insight runs
along a spectrum.
Learn something completely
new. This includes invalidating
well held cherished beliefs.
Confirming or adding more
depth to understanding
something that was known.
Fail to find any useful insight at
all.
© 2016 Gordon MacMaster All Rights Reserved
- 4. 1 + 1 = 5
Analytic insight creates nonlinear results. The effort to find the insight and take
action is disproportionate to the significant gains that can be achieved.
Statements along the lines of this one change brought about $1M in savings, think
about what more analytics could bring, is an indicator of linear thinking and a
misunderstanding of analytics.
© 2016 Gordon MacMaster All Rights Reserved
- 5. So how does it happen?
The main steps are:
1. What is the question to be
answered?
2. Get, prepare and understand
the data.
3. Apply analytic techniques.
4. Validate the results.
5. Analyze some more.
6. Draw a conclusion.
% of Effort
© 2016 Gordon MacMaster All Rights Reserved
- 6. The bulk of the
effort in Advanced
Analytics is before
the analysis.
© 2016 Gordon MacMaster All Rights Reserved
- 7. The first two steps are the most difficult
1. What is the question to be answered?
The original question is usually not the real
question. It is often very difficult to get an
understanding of what a Client needs
versus what is thought to be needed.
A generalist rather than a specialist is
better suited to conceptualize the
question.
● Specialists only see differences.
● Generalists only see commonalities.
This is the key success factor in any
Advanced Analytics activity.
© 2016 Gordon MacMaster All Rights Reserved
2. Get, prepare and understand the data.
Data may exist, but it may be insufficient
to answer the question. 3rd party data can
supplement data sources.
Business Rules are key in understanding
the data. This is how the data is used.
Even within the same organization, the
same term may have different meanings.
The differences are often nuanced but
essential to understand.
- 8. Finally, the
analysis starts.
Once the question and data are
set, what happens next?
Here are three ways to describe
analytic discoveries with real Client
examples:
Sherlock Holmes
Goldilocks
Dark Matter
© 2016 Gordon MacMaster All Rights Reserved
- 9. Sherlock Holmes
The fictional detective was famous for his powers of observation and deduction.
By looking at a man’s shoes he could reasonably deduce the man’s life story.
Advanced Analytics looks at patterns and correlations to deduce insight.
Remember - correlation does not equal causation!
The detection of patterns is what most people see as Advanced Analytics.
Bringing together data that is usually never connected is often a great source of
insight.
© 2016 Gordon MacMaster All Rights Reserved
- 10. Goldilocks
Extending on the nonlinear aspect of analytics, the results may recommend a
range of actions rather than a specific item.
Much like Goldilocks, the insight may be to do things that are neither too hot or
too cold.
Most people recognize that doing little of something is usually bad. But the same
can be true of doing too much.
Identifying that balance can lead to great gains.
© 2016 Gordon MacMaster All Rights Reserved
- 11. Dark Matter
In astrophysics, the movement of some stars and planets appear to be influenced
by something that cannot be seen.
This is referred to as dark matter.
Analytics can detect patterns where there is no obvious correlation or cause.
Figuring out the dark matter in these situations require more than just analytic
skills. It requires creativity.
Insights gained here can be tremendous breakthroughs.
© 2016 Gordon MacMaster All Rights Reserved
- 12. Sherlock Holmes Example: Claim Processing
The Question: Can claim data provide any ideas for improvement?
The real question: Is the claim payment process creating unnecessary corrections?
The Data: The data was policy information and associated claims. No data was available regarding
transaction activity of the claims processors.
The Analysis: Looking at lag times between policy events, claim payments and claim corrections
indicated a hole in their process. Information was not being provided in a timely manner to the right
processors.
The Results: Correcting the hole in their process with a simple report resulted in over $1M in cost
avoidance each year.
© 2016 Gordon MacMaster All Rights Reserved
- 13. Goldilocks Example: A Call Center
The Question: How can we measure the effectiveness of our call agents?
The real question: What is the best path for a call agent to follow to get the desired result?
The Data: The call agents had three, interdependent categories of activity. Each category had over a
dozen possible outcomes.
The Analysis: There was not a single path that would lead to success. Depending on the outcomes in
each category, there was a range of tasks that were most effective.
The Results: By measuring success using the range of tasks, call agent success quadrupled.
© 2016 Gordon MacMaster All Rights Reserved
- 14. Dark Matter Example: Health Insurance Company
The Question: Where can we save money?
The real question: We already have Payment Integrity Programs in place to ensure claims are paid
correctly. Is it enough? Can we recover more incorrectly paid claims?
The Data: Data was siloed and fragmented amongst systems. There was no unifying aspect to tie it all
together. The data was fused and a comprehensive event history constructed.
The Analysis: Life events of a person had the greatest impact on determining if a claim was recoverable.
The traditional analysis of the claims was not required.
The Results: An increase from $5M in recoveries a year to $36M.
© 2016 Gordon MacMaster All Rights Reserved
- 16. What about Data Science?
© 2016 Gordon MacMaster All Rights Reserved
Advanced Analytics and Data Science are often used interchangeably. Data Science is a broader umbrella term
that encompassess many activities including Advanced Analytics.
Big Data
Data Fusion
Data Modelling
Meta Data
KPIs
Benchmarking
Reports
Dashboards
Data Mining
Retrodictive
Analytics
Predictive
Analytics
Prescriptive
AnalyticsMachine
Learning
Data
Management Business
Intelligence Advanced Analytics