3. Agenda
Introductions
Part One: Marketing Analytics Framework
4 Problem Statement
4 Solution Framework
Break
Part Two: SQL Server & Analysis Services Deep Dive
4 Marketing Data Mart Schema
4 Marketing Cubes
4 Interactive Analysis
Summary/Q&A
3
4. Speaker & Company Bio
Sandeep Giri
• Senior Director, Responsys, Inc..
• 15 years in software product development
• Focus:
4 Marketing databases
4 Business intelligence
Responsys, Inc. recently acquired Loyalty Matrix, Inc.
• Recently acquired San Francisco based; founded in 2001
• On-demand marketing analytics to optimize direct marketing
• Clients include Apple, 24 Hour Fitness, Chicago Sun-Times
• Sponsors openi.org – : open source BI app
• http://www.responsys.com
Copyright (c) 2007, Responsys, Inc. 4
6. Direct Marketing: An industry that does cartwheels on
“achieving” 99% failure
Common questions NOT being answered effectively
“How do I know which customers and
prospects are most likely to respond to my
marketing programs?“
“Who are my most valuable customers?”
“How much should I spend on marketing in
each segment?”
“I’d like to know in advance exactly which
customers I’m at risk of losing so I can at
least do something about it in time”
6
Copyright (c) 2007, Responsys, Inc.
7. Black Box of Marketing:
Identify Optimal Segments & Tactics
Leverage data Identify most relevant segments…
Segment A Segment B Segment C
Customer/ Behavioral
Prospect (Usage, Black Box
List Txn’s)
of Program
1
Campaign 3rd Party
Marketing Program Program
& Response Appends Analytics 2 3
Program Program
4 5
… and the most optimal
tactics for each segment
Copyright (c) 2007, Responsys, Inc. 7
8. What Happens Inside the Black Box?
Build
Phase I. Define
Marketing
WHAT is going on? KPI’s
Data Mart
Response Segments
Find
Phase II. LTV Predictive
Define
WHY is this occurring? Lists
Triggers
Test Learn
Retention
Tactical
Segments
Execute
Phase III. Contact Analyze Past
HOW do you improve? Strategy Campaigns
Test
Plans Tactics
8
9. Segments: What Separates the Best from Worst?
Before you segment
Customers
• What you are segmenting for? & Prospects
• Identify events you want to control: response, attrition
Define Key Performance Indicators (KPI’s)
• Response: response rate, volume Segment A Segment B Segment C
• Value: annual sales, LTV, profit
• Propensity:
4 To attrite
4 To purchase (up-sell/cross-sell)
Find predictive triggers for each KPI
• Consolidate all data attributes related to the event/customer
• Find strongest predictors of the event (predictive analytics)
• Define segments based on the strongest predictors
Copyright (c) 2007, Responsys, Inc. 9
11. Example: Model Identifying Strongest Predictors
Among the Candidates
Copyright (c) 2007, Responsys, Inc. 11
12. Tactics: What Works the Best for Each Segment?
Marketing
Right offer to the right person at the Programs
List Frequency
right time with the right media
Media Creative
• Segmentation answers “who” (person) Offer
• Next up is “how” (offer, time, media, etc.)
“History repeats itself, it is a good thing”
Program
• Consolidate past campaign data 1
Program Program
• Overlay segment definitions 2 3
• Apply predictive analytics to Program
4
Program
5
identify optimal mixes of tactics
• Create tactical segments
Contact Strategy & Test Plan based on tactical segments
Copyright (c) 2007, Responsys, Inc. 12
13. Example: Response Rates on Multiple Contacts for
Different Tactical Segments …
Copyright (c) 2007, Responsys, Inc. 13
14. … Results into Executable Direct Marketing Plan
Segment A Segment B Segment C Segment D
Former Weekend Former Weekday Former Weekend Former Weekday Former Weekend Former Weekday Former Weekend Former Weekday
(Segment A3) (Segment A7) (Segment B3) (Segment B7) (Segment C3) (Segment C7) (Segment D3) (Segment D7)
Month 1 Weekly 3D WB1 Weekly 7D WB1 Weekly 3D WB1 Weekly 7D WB1 Weekly 3D WB1 Weekly 7D WB1 Weekly 3D WB1 Weekly 7D WB1
Month 2
Month 3 Weekly 3D WB2 Weekly 3D WB2 Weekly 3D WB2 Weekly 3D WB2 Weekly 3D WB2 Weekly 3D WB2
Month 4
Month 5 Weekly 3D Weekly 7D Weekly 3D Weekly 7D Weekly 3D Weekly 7D
Month 6 Annual 3D Annual 7D Annual 3D Annual 7D Annual 3D Annual 7D
Month 7 Weekly 3D Weekly 3D Weekly 3D Weekly 3D Weekly 3D Weekly 3D
Month 8 Annual 3D Annual 3D Annual 3D Annual 3D Annual 3D Annual 3D
Month 9 Annual + 3D Annual + 7D Annual + 3D Annual + 7D Annual + 3D Annual + 7D
Month 10 Weekly 3D Weekly 7D Weekly 3D Weekly 7D
Month 11 Annual 3D Annual 7D Annual 3D Annual 7D
Month 12 Annual + 3D Annual + 3D Annual + 3D Annual + 3D
Ongoing 3 Month
Ongoing Ongoing 6 Month cycle To segment 5 after To segment 5 after To Segment 5 After To Segment 5 After To segment 5 After To segment 5 After
Cycle as in Weeks 7-
Rule as in Weeks 7-12 month 12 month 12 Month 9 Month 9 month 1 month 1
9 and 10-12
Copyright (c) 2007, Responsys, Inc. 14
15. Test-Learn-Execute-Repeat
How do you determine the effectiveness of your marketing
analytics? Test Learn
Test Execute
• Create test plans for each recommendation
• Models available to calculate sizes for test and control groups
Learn
• Actual lift versus predicted lift
• Champion models
Execute, and test again!
Copyright (c) 2007, Responsys, Inc. 15
19. Design and Build
Marketing Data Mart
• ETL: Extract, Transform, and Load
• Star Schema (versus Snowflake): Denormalization
• Facts = events
• Dimensions = slicer attributes
OLAP
• Cubes – Facts, Dimensions, and Measures
• Careful of the joins
Visualization: Reporting Services, open source options such as
OpenI
Walkthrough
Copyright (c) 2007, Responsys, Inc. 19
20. Summary
Marketing Analytics = Segmentation & tactical
optimization based on marketing data
(customer/prospects, commerce, campaign)
Microsoft SQL Server and Analysis Services provide:
• ETL: DTS, SSIS
• Marketing data mart: star schema in SQL database
• OLAP: Analysis Services cubes organized by marketing
facts and dimensions (Campaign, Sales, etc.)
As always, setting analytics objectives as important (if
not more) as designing/building the solution
Copyright (c) 2007, Responsys, Inc. 20
21. Questions/Comments?
Now would be the time
Feel free to email me: sgiri@responsys.com
This presentation will be downloadable at baadd.org
This conversation will continue at
http://sandeep-giri.blogspot.com
Copyright (c) 2007, Responsys, Inc. 21