The Real World View Of Analytics-How To Use Data To Drive Business Impact
The Real World View of Analytics
How To Use Data To Drive Business Impact
What will be covered:
• The 6 steps to a successful analytic
• The 4 key elements that drive strong,
Wikipedia: Business analytics make extensive use of data,
statistical and quantitative analysis, explanatory and predictive
modeling, and fact-based management to drive decision
Investigation of past business
performance to quickly gain
insight and drive business
Drive Fast Business Decisions
Successful Analytics Begin with Understanding How and Where the
Results will be Used
Many people begin an analytic with the data…
… looking for gold in all the noise
Step 1: Define the business problem
your analytics are to solve
• Who are my best customers?
• What offers should I make and
when should I make them?
• Where are my next best markets?
• Why are my customers leaving?
• Who is influencing my customers?
• Who’s a fraudster, etc.?
Once the business problem is defined,
the next step is to consider the touch
points where those analytics will have
the greatest impact.
Step 2: Understand your touch points
External Touch Points
• On-line advertising
• Sales teams
• Customer Service call centers
• Print and broadcast media
• Point of sale
Internal Touch Points
• Annual budget
• Capital investment
• Strategic decisions
• Vendor selection
All touch points have constraints
and challenges; understand the
constraints before starting the
Step 3: Within the targeted touch
points, understand the constraints
• System Limitations
• Access to data
Design your analytic understanding these
constraints and knowing which are worth
Avoid developing analytics that will require
months or years of effort to implement.
Step 1: Define the business problem your analytics are to solve
• Who are my best customers?
• What is the next best offer I should make to them?
• Solution: Analyze historical and daily sales activities to identify best customers and to
develop next best offer
Step 2: What touch points can be employed to influence the customer?
• Field sales staff making home delivery/sales rounds and
Step 3: Understand the constraints. Sale staff need the promotional recommendations provided
to the them in the field daily based on their customer routes
• This information has to be easily accessed and available on their mobile computers
Example: A home food delivery company with a sales staff of more than 3,000 in the field
wants to increase its revenue.
Example: Bust-Out Fraud Losses are the Ultimate Disappearing Act
• 80% of Bust-Out Fraud Cases had more than 6 Months of tenure; furthermore, 30% of Bust-Outs had
more than 5 years of tenure.
• Bust-Out trade lines ramp balances quickly and charge off at well over the credit limit.
• Here the customer uses credit with intent to max out the Line of Credit before disappearing.
Customer spends normally
and pays regularly (6-12mo)
Line increase requested
Customer spends rapidly
towards new credit limit
Customer makes payment with
Open to Buy
Increased spending on fencible
checks or balance transfers
No payment or
Bust -Out Triggers
Step 4: Select data for the analytic that allows you to solve the business problems but that
also allows for fast and efficient implementation
Considerations when deciding what data to use:
• Timeliness and accuracy of the data
• Completeness and access of the data
• Relevancy to the business problem
• Regulatory or reputational concerns
• Ability to get the data to the appropriate touch points
Data that can be exploited for an analytic moves progressively from structured to
semi-structured to unstructured data, and with that shift comes increased complexity.
Database data includes:
• Historical sales
• Employment and wealth
• Address and identity
Example: The first card program in
• Developed a new behavior score
• A New model incorporated legal
data element Tenure with Ward
• Used a new behavior score to
reduce lines on high risk accounts
• Many older, long-tenured Wards customers
where receiving Adverse Action letters stating,
“Your credit line has been reduced…” and the top
reason was “…have only been with Wards for 20
• A TV reporter interviewed a grandmother in front
of a Ward store. Having received such a letter,
she complained about bad treatment by Ward
and the bank and said she
uses the card only to buy
• Informing a women you know
she is pregnant can be a sensitive
• An angry man entered a store
outside of Minneapolis demand-
ing to speak with the store
“My daughter got this catalog in the mail.
She’s in high school, and you’re sending her
coupons for baby cloths and cribs. Are you
trying to encourage her to get pregnant?”
• Must consider how the solution will be
viewed by your customers.
Example: Target’s Pregnancy Score
• Birth records are usually public;
therefore, couples with new babies
are instantaneously barraged with
offers and incentives.
• Target notices certain buying habits of
women who are pregnant including:
Increased intake of supplements
such as calcium, magnesium and
Buying scent-free soaps, hand
sanitizer, cotton balls wash cloths
• The score predicts Probability of
Pregnancy and estimated due date.
Example: Give to Get
• On a plane seated next to a Google Executive.
• He was talking about a new Beta app.
• Having dinner with friends but he needed to pick
up his daughter from LaGuardia Airport.
• While he was seated at dinner, the app sent him a
notification that his daughter’s plane was late. It
also told him, based on his current location and
current traffic, when he should leave the restaurant
to meet his daughter and the best route to take.
• In order for the app to work, it needed access to
much of his personal information
Step 5: Design the analytic
Make sure your analytic design:
• Meets the business needs that have
• Takes into account the touch points
• Takes into account the data being
used to solve the business needs
• Uses analytical techniques that are
appropriate for the industry and the
Don’t over engineer the design, use the
Examples: Many different analytical techniques
• Regression – logistic, linear, Poisson, gamma
• Artificial Intelligence such as neural networks
• Linear and quadratic programming
• Machine learning
• Hybrid solutions such as combinations of
regression and neural networks
• Multiple fusion techniques
• Decision trees
Choose the analytical technique most
comfortable for your team so that the analytics
can be completed quickly and are easy to
Step 6: Analytics will often answer some questions, but good analytics raise new
• After completing your analytics, you may have
the answers you were seeking. But it’s not
unusual to derive results requiring further
• Incorporate these results to refine your
Six steps to Successful Analytics
1. Define the business problem.
2. Determine the touch points where the analytics will have the most impact.
3. Understand the constraints and challenges for taking actions based on the
results of those analytics at each touch point.
4. Understand the data that is available at these touch points given the
5. Design the analytics based on the data and implementation constraints
you are not willing to change or do not have time to change.
6. Conduct the analytics, and based on the results, either refine the analytic
or implement the recommendations.
Analytics That Drives Business Impacts Boils Down To Four Basic Elements
Easy and efficient access to relevant data.
Industry knowledge to understand the customer problems to
be solved and the data and analytical techniques to be used.
Otherwise, there is a steep and costly learning curve.
The ability to bring solutions to market and to make these
solutions reliable and easy to use and to deliver them to the
right place at the right time.
The ability to attract and retain talent with deep analytical
expertise in math, statistics and operations research.
A recent study by McKinsey indicated analytic talent will be in short supply.
Therefore, companies must focus on attracting and retaining talent with the
right skills across several
types of analytics.
Having a pool of experienced
analytical talent will be a
very important and valuable