Post on 13-Jul-2020
Lancet Data Sciences and Bluestem Brands Man vs. Analytics
Neil GunnBluestem Brands Inc.BI ManagerNeil.Gunn@bluestembrands.com
Jason ToddLancet Data SciencesPractice Leader & AEjtodd@lancetdatasciences.com
The Challenging Part of Analytics…
It can feel like team members are speaking a language that can be tricky to understand
via mathsyear7.wikispaces.comVia http://en.wikiquote.org/wiki/Talk:Computers
Agenda
• About Bluestem• About Lancet• A business centric approach
• Define• Organize• Match• Maintain
• Matchmaking• Advance analytical capabilities• Future
Consumable Analytics
Lancet Data Sciences
BI is in our DNA• Became a MSTR partner in
1997• Multiple Practices• Managed Services• 2013 MSTR Solutions
Provider of the Year• 2013 (#34)-Fastest growing
private companies in the Twin Cities
• 2013 Inc. 500|5000 list of the fastest-growing private companies in America
Lancet Data Sciences
• Headquartered in suburban Minneapolis
• Parent company to three fast-growing, dynamic, multichannel retail brands.
• We provide a unique mix of retail and payment options for a diverse range of customers with a wide range of financial needs.
• Three campuses and more than 1000 employees
Bluestem Brands
Lancet Data Sciences
• In 2010, Bluestem and Lancet collaborated on a BI Roadmap to transform our BI environment
• After implementation of a new Enterprise Data Warehouse, MicroStrategy was the final piece in the puzzle
Bluestem and Lancet – The Journey
• MicroStrategy Implemented in April 2012
100 Users less than a month after install
Utilized visual insight to gain momentum
Platform developed and administered by Lancet
Today there are now 300 users of MicroStrategyin several different departments across Bluestem
Consumable Analytics: A business focused style
Not just pushing buttons and configuring technical BI options
Define: Use case(s) & storyline(s)
Organize: Business questions into desired flows
Match: Business questions & Analytic methods
Maintain: The approach
Tunnel Vision A Data Science Approach
A business centric language
Consumable Analytics: Solving business problems
Type Question Examples
Historic What has happened?Sum, Avg, Rank, Min, Max, Counts
PredictiveWhat will happen if these trends continue?
Linear & Exponential Regression, Time Series
RelationAre these entitiesassociated?
Affinity
GroupingHow should I assemblethese items?
Clustering
ClassifyWhat’s the chance of this outcome?
Decision Tree, Logistic Regression
Putting analytics in a business focused language that engages users
Define: Use case(s) & storyline(s)
Lancet Data Sciences
Define: Use case(s) & storyline(s)
From Whiteboard to Dashboard!
What questions am I trying to answer?
What actions will be taken from this dashboard?
Who is the audience?
What is the purpose of this dashboard?
How do I navigate here?
Lancet Data Sciences
Define: Use case(s) & storyline(s)
From Whiteboard to Dashboard: Prototyping
Lancet Mock-up Dashboard
Lancet Data Sciences
Define: Use case(s) & storyline(s)
From Whiteboard to Dashboard: The Final Product
Confidential – for live presentation only
Final Dashboard Presentation
Co
mp
etit
ive
Ad
van
tage
What’s the best that can happen?
What will happen next?
What if these trends continue?
Why is this happening?
What actions are needed?
Where exactly is the problem?
How many, how often, where?
What happened?
Sophistication of Intelligence
Optimization
Predictive Modeling
Forecasting/extrapolation
Statistical analysis
Business Intelligence
Competing on Analytics: The New Science of Winning, by Thomas H. Davenport and Jeanne G. Harris (Harvard Business School Press, March 2007).
Alerts
Query/drill down
Ad hoc reports
Standard reports
Organize: Business questions into desired flows
Consumable Analytics: Most have seen…
Consumable Analytics: Organizing business questions
Historic Questions• Year to Date?• Quarter to Date?• Month To Date?• Today?
What are my current sales?
Predictive Questions• Forecast this month?• Forecast this quarter?• Forecast this year?
• Solution with match the desired information flow
• Builds an information app vs unsynchronized dashboard
• Reduces dashboard build time
• Engages the user community
Organize: Business questions into desired flows
Analysis/Method
Logistic Regression
Time Series
Association
Decision Tree
Clustering
Rank
Moving Average
Regression (Linear & Exp.)
Which stores have a similar product sales pattern?
What is my sales forecast?
What are my cross-sell recommendations?
What’s my optimal investment choice?
What is my average defect rate every 10 days by location?
How do I rank against my peer in the last 30 days?
AlgorithmK-means
Match: Business questions & Analytic methods
Consumable Analytics: Matching
Lancet Data Sciences
Match: Business questions & Analytic methods
Consumable Analytics: Create Personalized Expressions
• Expressions (rich features)– Formula– Level– Condition– Transformations– Properties
• Use Metrics Anywhere– Dashboards– Consolidations– Filters– Subtotals– Prompts– Custom Groups
• Deployable to – Web– Mobile– Office
• Common Metric Definitions!
Lancet Data Sciences
Match: Business questions & Analytic methods
Consumable Analytics: Association Use Case
What are my cross-sell recommendations?
Important Considerations!
• What is the time horizon?
• Are there any exclusions?
• What is the dimensionality?
• In a single shopping visit or over customer lifetime?
What products do customers commonly buy together?
Analysis/MethodAssociation
Lancet Data Sciences
Match: Business questions & Analytic methods
Consumable Analytics: Association Use Case
Lancet Mock-up Dashboard
• “Lots of fish in the sea” for matchmaking…• Can help to improve performance• Dynamic parameters• Lots of opportunity to utilize….
Basic
• Sum
• Count
• Average
• First
• Min
• Max
• Last
Comp. & Operators
• >
• <
• =
• Like
• Between
• +
• -
• ()
Date & Time
• Add Days
• Current Date
• Current Time
• Hour
• Minute
• Month
• Quarter
• Year
Financial
• IRR
• NPV
• PV
• PMT
• Rate
• FV
• MIRR
• Duration
Internal & String
• Apply Simple
• Apply Agg
• Banding
• Case
• Coalesce
• Concat
• Position
• Trim
Logical & Zero
• And
• If
• Not
• Or
• Is not null
• Null
• Null to zero
• Zero to null
Math & Stat.
• Int
• Round
• Slope
• Power
• Tan
• AvgDev
• Forecast
• Growth
• Intercept
OLAP & Rank
• Mov. Avg
• Mov. Count
• Run.Avg
• Run. Max
• Run. Min
• Ntile
• Rank
• Percentile
Match: Business questions & Analytic methods
Consumable Analytics: Comprehensive Library
• “Parameterized” models• Accepted algorithms• View model details• Lots of opportunity to utilize….
Analysis Details Uses
Linear Regression Ordinary Least Squares Forecasting
Exponential Regression Ordinary Least Squares Forecasting
Time Series Exponential Smoothing Forecasting
Logistic Regression Iteratively Re-weighted Least Squares
Predicts Categories
Decision Tree CART algorithm Predicts Categories
Clustering K-means algorithm Grouping
Association Rules Apriori algorithm Affinity
Match: Business questions & Analytic methods
Consumable Analytics: Solid data mining foundation
• Pass-thru functions• PMML• R Plug-In• GEO API’s• Lots of opportunity to utilize….
Match: Business questions & Analytic methods
Consumable Analytics: A business centric approach
Lancet Data Sciences
Match: Business questions & Analytic methods
Consumable Analytics: Integrating Tools
• Limited to 4GB currently• 64 bit will be coming!
Lancet Data Sciences
Match: Business questions & Analytic methods
Consumable Analytics: Building Models
• It can be an iterative process…
• Baseline knowledge of data mining algorithms
• Metadata is critical for success
• Many visualization options to display the output
Create Dataset Deploy MaintainBuild
Analytical Lifecycle
Analyze options Analyze variables
Validate ModelCalibrate Model
PresentationReview as-needed
Build
DeployMonitor & Maintain
• Building is only the 1st step…• Monitor & Maintain is
sometimes forgotten• Schedule frequent
business reviews• On predictive metrics,
review your accuracy• Adjust model parameters
as needed to fit the situation in the field
Analytical Lifecycle
A Lasting Relationship
Maintain: The approach
Consumable Analytics: Analytical lifecycle
Lancet Data SciencesConsumable Analytics: Monitor & Maintain
Enterprise Manager has a wealth of data waiting to be explored!
What are my most run reports?
Which reports or cubes run the
longest?
Are users experiencing
errors?
Who are your power users?
Should I look into my cube strategy?
Am I meeting my SLA’s?
Maintain: The approach
Lancet Data Sciences
Maintain: The approach
Consumable Analytics: A business centric approach
Language: Chat Session gone awry A Data Science ApproachA business centric language
Via http://en.wikiquote.org/wiki/Talk:Computers
Define: Use case(s) & storyline(s)
Organize: Business questions into desired flows
Match: Business questions & Analytic methods
Maintain: The approach
Lancet Data Sciences
Maintain: The approach
Consumable Analytics: Future
• Stats & Entertainment…?• A new kind of capital• Data Quality• Security, Risk, & Compliance• “Localization”• Cross Sell• Logistics