2. Macy’s, Inc. Background
Macy’s, Inc. is one of the nation’s premier
omnichannel retailers
Fiscal 2015 sales of $27.1 billion
Operates 870 stores in 45 states
Brands: Macy’s, Macy’s Backstage,
Bloomingdale’s, Bloomingdale’s Outlet,
Bluemercury as well as macys.com,
bloomingdales.com, bluemercury.com
Ships products to over 100 countries
Workforce includes over 157,000 employees
3. Digital Growth
World went digital
Macys generated its first billion-dollar month
of sales from digital platforms in December
2015
Filled nearly 17 million online orders at
macys.com in November/December 2015 an
increase of about 25% over previous year
Based on significant new fulfillment capacity,
site functionality, and aggressive digital
marketing
4. Why Hadoop at Macy’s?
Traditional data architecture is
inflexible and not nimble
Inability to tap into historical data
Severe compute capacity limitations
Significant cost implications to
scaling
Unstructured data sources
5. Why BI on Hadoop?... Why Not?!
Single data architecture can cater to
a comprehensive list of use-cases
Integrated eco-system of data,
process, and tools
Analytics, Experimentation, and
Production can be collocated
Low total cost of ownership
6. What does it mean to be
Operational?
Ability to move quickly from
testing/experimentation cycle to
production
Reliable data quality, governance,
and security
Acceptable levels of stability and
robustness to meet SLAs
Automation to the nth degree
9. Need a Robust Experimentation Framework
Problem Statement
• What issue are we trying to
solve for?
• Why is it important to the
business?
Size of Problem
• What’s the $ impact?
• % customers affected?
• What can/can’t we influence?
Hypotheses
• What’s the root cause?
• What change will have the best
ROI?
• Are there alternatives?
Supporting Data
• Validate (or adjust) our
hypotheses
• Rule out false positives
Tests
• What’s the safest way to test
our riskiest assumptions?
• Who/when/how?
Predictors
• What variables are most highly
correlated with our problem
and/or solution?
KPIs / Success
• What outcome would we define
as “success”
• What’s our response to
success/failure?
Key
“Who provides?”
Team 1
Team 2
Team 3
10. Data
Domains
Orders
Customers
Products
Clicks
Marketing
External
Big Data Repository
In-memory Data
De-dupe
AggregationTransformation
Blending
Tools
Campaign Management/
Optimization
Statistical Analysis
Consumers
Merchandizing
Marketing
Product
Management
Analytics
Data Scientists
Advanced Analysis/
Modeling
Data Visualization/
Data Mining
Other Business
groups
Storage and Enrichment
Data Management
Data Security
11. Growing pains
Challenge
Significant time spent on data
engineering
Long analytic iteration times
Inability for analysts to collaborate
Solve
Need to establish a virtual semantic layer
Seamlessly integrate with existing tools
Deploying in-memory Big-Data OLAP tool
12. How to drive adoption?
Quality
Release
SocializeTrain
Measure
13. Adoption Checklist Center of Operations
+
Center of Evangelism
Confidence in data quality
Data governance, and security
Standardized release process
Socialize and Train
Monitor adoption (Qualitative and
Quantitative)
14. Keys to Success
Laser focus in delivering business
value
Keep process overheads at check
Continuous operational improvement
Tolerance to a maturing solution for
the greater good
Flexible resource model