BSI team recommends technologies from Teradata Aster and Aprimo, a Teradata company, for better marketing via event-based Marketing, GoldenPath Analytics, and Attribution/Digital Marketing Optimization.
1. HOW WE DID THE The Sad Case of
INVESTIGATIONS StagnoBank – Part 1
2. Prelude – Part 1
This deck accompanies the
Sad Case of StagnoBank - Part 1 Video
at http://youtu.be/MScwTqhM3TI
You can find this with a search for “BSI”,
“Teradata”, “Case”, “StagnoBank”.
It is designed to answer questions about the
technology shown in the story
2
3. Note from the Investigators
Hi Everybody,
We’re the brains behind the scenes
and wanted to answer your questions
about “how we did the StagnoBank
brainstorming so fast.”
This write-up will give you an idea of
our clients’ architecture and some
details of the BI screens.
Take a look, and if you still have
questions, send them to us! We’re
both on Facebook.
Yours truly,
Max Ridge and Jodice Blinco
3
4. Scene Synopsis
• Jodice’s Office at BSI HQ – Simon explains the situation,
shows Jodice KPIs and reports, and commissions the work
• Jodice kicks off the Project with Max, Mercedes, and Mathieu
4 Weeks Later
• BSI Conference Room – readout of ideas for Better
Marketing (Max), Better Customer Service (Mercedes), and
Mobile Apps (Matt)
• This deck show’s Max’s work – Part 1; see also Part 2 for
Mercedes’ ideas, and Part 3 for Matt’s
4
5. Summary of the Ideas from the BSI Team
Max Ideas Event Based GoldenPath Attribution
Campaigns to Analysis to Analytics and
increase increase Digital
relevance channel Marketing
effectiveness Optimization
Mercedes Ideas “One and Customized Same agent
Done” button call routing
screens for pushes on the
contact Interactive
center agents Voice
Response
Matt Ideas “Consumer “My Bank “Geospatial
Intelligence” Looks Out for Apps” to
budget / Me” alerts drive
planning apps customer
education
5
6. Scene 1: Problems at StagnoBank!
Meeting of Simon (CMO) and Jodice (BSI)
Simon and Jodice … in her office talking
• Simon: “I’m the new CMO, only on the job for 3 months, but
everywhere I turn, we have problems”
• Biggest issue – we’re a big, old bank, perceived as “behind the
times”. No appeal to younger households.
• Asks Jodice to do a quick BSI project to come with turnaround
ideas
Simon,
StagnoBank’s
CMO
6
7. Summary of StagnoBank’s Problems
Business KPIs
• Assets dropping
• Margins eroding
• Customer count dropping
• Losing market share
Customer KPIs
• Average age of customer is increasing
• Decrease in take rates for offers
Jodice agrees to
• Bad customer service scores
take on the
Channel KPIs assignment, will
• Branch services under-utilized have her team do
• Long wait times at the call center interviews and
• Weak mobile and online banking offers brainstorming
7
8. KPIs Not Good: Assets and Margins Dropping
StagnoBank Assets By
Quarter ($B)
120
100
80
60
40
20
0
ROA = return on assets
8
9. Bank Results Are Not Good:
Number of Accounts and Market Share
# Consumer Accounts
(Millions)
2.35
2.3
2.25
2.2
2.15
2.1
Market Share (%)
20
19
18
17
16
15
14
13
12
11
10
9
10. Bad Take Rates for Car and Credit Card Offers
10
Goal is to reach 2.5% take rates for all campaigns
11. Jodice Charters the Team
Jodice puts together a young team:
Mathieu, Max, and Mercedes
She commissions them to:
• Come up with Ideas for StagnoBank
– 3 each – 9 total
• Go interview StagnoBank customers Max
• Look at the bank’s data for yourself
• Interview some customers Mathieu
• Work hard and come back in 4
weeks with your best ideas!
Mercedes
11
12. The Team Divides Up the Brainstorming
Assignments
• Max: Better Marketing
• Mercedes: Better Customer Service
• Mathieu: Consumer Mobile Apps, Alerts, Geo-Spatial
12
13. Readouts
• 4 WEEKS LATER SIMON COMES OVER TO THE BSI HQ FOR A
READOUT WITH THE BSI Team
• Each one of the 3 team members gets their timeslot to show
off their best 3 ideas for their area. That makes 9 “ideas” in
total for Simon.
13
16. How We Did This Report
• This chart is from Aprimo’s integrated analytics suite-
specifically Behavior Trend Analysis- which can show the
behavior of any customer segment over time.
• Though not shown, Drill Down to the individuals included in
any of these segments is available at any time. By merely
pointing and clicking on these value bands, it is very quick
and simple to generate a list of customers that have dropped
out of the highest 10% of contributors to lower levels
between any two time periods.
• This would allow you to either do further analysis on these
customers, or quickly target them with promotions to re-
engage them.
16
17. The Problem
Cross-Channel Campaigns – Also Not Good
Any Household Age
MONTH APR MAY JUN JUL AUG SEP OCT NOV DEC
11 11 11 11 11 11 11 11 11
4219031 4390888 4309933 4220493 4390222 4109253 4239803 4440982 4590363
# Contacts No
E-Mail Intersection 30798
# Responses 28540 30169 29543 31170 28353 28830 35083 31214
% Response
0.73% 0.65% 0.70% 0.70% 0.71% 0.69% 0.68% 0.79% 0.68%
# Contacts
1652341
# Contacts 1567820
762516 1429033 1520987 1459092 1340964 1590202 1490341 1509231
Direct # Contacts/ yr. 261665
# Responses 12392 10661 11432 10190 8608 10157 11608 10432 10262
Mail % Response
# Contacts/ mo. 308996
0.75%
Response % 0.68%.076 0.80% 0.67% 0.59% 0.75% 0.73% 0.70% 0.68%
CAMPAIGN CONVERSION RATES
Monthly response rates across channels – all segments: 1.3 – 1.5%
Way too many emails and direct mail pieces – about 3 per month
per customer – and take rates are horrible
17
18. How We Did This Report
• This slide shows another of Aprimo’s integrated analytics-
Cross Segment Analysis.
• Here you can easily see the performance of various channels
over time, and could also quickly change this chart to show
the performance of any segment of customers, across
channels, over time.
• So, for example, you could quickly substitute customer age
ranges across the top and show the performance of different
communications channels by age segments- or customer
value segment, or by any other customer attribute.
18
19. What 3 Ideas Did Max Come Up With for
BETTER MARKETING?
Max Ideas Event Based GoldenPath Attribution
Campaigns to Analysis to Analytics and
increase increase Digital
relevance channel Marketing
effectiveness Optimization
19
20. Max Idea #1: Move to Event Based Campaigns
Example: Large Withdrawal Triggers Phone Call
20
21. How We Did It
• In this screen shot of Aprimo Relationship Manager, you can
see an example of an event based (or complex trigger
based) campaign. Event based campaigns allow you to
watch for specific behaviors, or combinations of behaviors,
by customers so that you can quickly respond with an
appropriate message or offer.
• The Large Withdrawal which is the primary characteristic of
this segment of customers actually implements a fairly
complex rule to identify customer that have exhibited a
specific behavior (or combination of behaviors) in the last x
time period.
21
22. How We Did It
• For example, a large deposit may be defined based on
individual characteristics- so it might be calculated to identify
customers who have made a deposit that is at least 500%
greater than their individual average deposits over the last
12 months.
• This provides much greater accuracy and relevance than
stipulating a set amount of deposit- so a $10,000 deposit
may be a “large” deposit for one person, but might not be a
big deal for someone else.
22
24. How We Did It
• Likewise, an event trigger could be se tup for anyone who
initiates an automatic deposit into their account- eliciting an
automatic email from the bank, thanking them for signing up
for direct deposit
• We could then possibly cross-sell other offers that have been
found through analysis to be attractive to people who just
started automatic deposits. The offers can be different, and
even use different channels, based on any attribute of the
new depositors.
• For example, for people in the targeted younger age group
just starting a new job, we might offer
> Consolidation of student loans
> Car loans
> New credit cards
24
25. Event-Based Campaigns Are Run By Aprimo
• See www.aprimo.com for tutorials and examples. The
technology illustrated here is called ARM – Aprimo
Relationship Manager
• Each industry at Teradata has built a set of interesting
“Events” – the two events here are on the list of 200 interest
events in the Banking Industry, and also are based on the
Teradata Financial Logical Data Model (next page)
• The events are detected often during the ETL or ELT phase
when loading data from a front-end transaction processing
(OLTP) system
• Teradata then hands the event to Aprimo for “action” (or
not), and it launches multi-channel, multi-step dialogues or
campaigns
25
27. SQL
• A fragment of pseudo SQL, for example:
SITUATION: LIKELY ACCOUNT CANCEL
AT-RISK EVENT: Unusually-Large-Withdrawal:
DEFINED AS
Current WithdrawalAmt > 5 * AVG(All Withdrawals)
27
28. Creating Customer Segments with Aprimo
• Aprimo Relationship Manager allows you to create segments
in 5 different ways:
- Segments can be created directly from analytics, as we saw
earlier
- Segments can be imported from a third party, such as an
analytics group, or MSP
- Segments can be created with a simple, point and click user
interface, known as Selection Manager, that is a standard
component of ARM
- Segments can be created by selecting tables and fields from
the database, or
- Segments can be created from custom SQL that is written to
address very complex scenarios
28
29. Max Idea #2: Use GoldenPath Analysis
• Golden Path Analysis – once we agree to doing more event-
based campaigns and aiming at new segments, we have to
optimize their experiences.
• What is the PATH TO PURCHASE? How many steps?
Which channels? How long does it take?
• Younger people will NOT put up with what you have now in
terms of the mobile web experience … too many clicks
29
30. What products are most popular with young
adults in the last month?
30
31. Response Rates By Channel (Across All Offers)
For Younger Households are Poor
31
32. An Aside: Aster
• The technology for Goldenpath and Attribution Analytics is
based on Teradata Aster, an acquisition Teradata made in
2011
• This technology is designed for use by “Data Scientists” who
are familiar with SQL MapReduce and Hadoop technologies,
especially suited in deriving insights from non-traditional
data (e.g., data not easily structured in relational database
tables)
• Web graph analytics fit into this class of BI, along with other
categories not in this episode like finding fraud patterns
• Aster and Teradata sit “side by side”, as shown in the next
page
32
33. Aster Data Analytic Platform Complements
an Existing Teradata System
Brings data science to the masses
Aster Data Teradata Integrated
Analytic Platform Data Warehouse
)
(or Appliance)
Example Apps
Investigative Analysis Example Apps
Social Media Data
SQL-MapReduce
Integrated Web
Retention & Analysis Intelligence OLAP
Scoring and Behavioral Investigate in
Relationship Scoring
Anomaly Analysis Aster Data, Management
Integrate &
Operationalize Analytics
Fraud/Cheating Fraud
Detection in the Data
Prevention
Warehouse
Reporting
Marketing Insights Process
Optimization
33
34. Aster GoldenPath Analysis
Analyze behaviors – across all Cross-Channel Customer Interactions
channels
17,000 Customers, 1 Month
Watch paths to purchase, and
look for / fix problems in the
paths to purchase 34,000 Branch Visits 25,000 ATM Sessions
userID event time userID event time
50001 Withdraw 12:00 PM 40001 Inquiry 12:00 PM
With Aster Data 30001 Deposit 1:45 PM 40001 Deposit 1:45 PM
• SQL-MapReduce for pattern matching 10001 Inquiry 3:00 PM 20001 Withdraw 3:00 PM
can identify the “last mile”
30001 Deposit 12:20 PM 20001 Home 12:20 PM
> E.g. Identify all interaction
patterns prior to an event of 5,000 Call Center Sessions
interest – like taking out a loan –
4300 E-mails 92,000 Online Sessions
and time spent on each channel
userID event time userID page time
Impact 30001 Sent 12:00 PM 10001 Home 12:00 PM
• With 10-300x less effort, know when 20001 Click 1:45 PM 50001 Banking 1:45 PM
customers are in the “last mile” of 30001 Open 3:00 PM 40001 Mortgage 3:00 PM
consideration
40001 Click 12:20 PM 50001 Home 12:20 PM
34
35. Sample Insights - GoldenPath
• Paid Ads on websites: average number of ad impressions to
drive customer to our savings website: 10.8
• On the Stagnobank web: Number of
web clicks to research (pre-app): 10
• Number of web fields to
fill out a simple savings
account application: 25
> Competitor Alpha:
12
> Competitor Bravo:
14
35
36. Where Do People Drop out when Opening a
Savings Account on the Website?
36
37. How We Did It: Aster – Teradata Adapter
Operational and Strategic
Big Data Analytics Intelligence
Business Objects, etc
Queries
Queries
Queen
Workers Teradata
Integrated
SQL/MR Data Warehouse
Loaders/Exporters
2- way Aster/TD
Connector
Big Data Sources Aster Analytic Platform Teradata Integrated Data Sources
Data Warehourse
37
38. How Aster and Teradata Work Together
CookieID UserID Attribution_Path
Aster Discovery Platform Teradata
Analytics Development Integrated Data Warehouse
Analytic Processing Structured Insights
(examples)
Parallel Data Storage • Campaign/Media Costs
• Marketing ROI Calculation
• Customer Value
Raw Web Logs Social Media 3rd Party Data SQL
OLAP Reporting
APIs Analytics
ERP E-POS Legacy Consumer
38
39. How We Did It: Aster - Teradata Adapter Usage
• Customer Profile: StagnoBank is an existing Teradata Customer
interested in doing detailed pattern and path analysis on clickstream
data. Max installed an Aster system to do his analysis.
• Use Case: How to use an Aster Data system with Teradata to support
Digital Marketing Optimization and Attribution
• Analytics Workflow:
1. Load: Load data feeds from weblogs, Omniture, Doubleclick, etc to Aster
2. Analyze: Use SQL-MapReduce to perform pathing, attribution on the
clickstream
3. Enrich: Enrich pathing analysis on clickstream with dimensional
information from Teradata EDW
4. Implement: Move high-value customer ids to Teradata EDW. Implement
marketing campaign using Aprimo Relationship Manager
39
40. Conclusion #2: Fix Your Web Site
• Redesign it!
• Pay attention to
what people are
doing, how long
it takes
• Optimize,
especially
compared to the
competition
40
41. Idea #3: Optimize Marketing Spend
• Attribution Analytics
> Do you know what it costs to cause consumer behavior (like a
purchase)?
> Can you attribute the cost to each channel (or previous
campaign)?
• Digital Marketing Optimization
– Do you know how much to spend on the various elements of driving
consumer behavior?
– Are your investments the right ones?
41
42. Attribution Analysis and DMO –
Web /paid search
Web / organic search
Call Center / agent
Web / organic
Branch /banker
Web / application
Call Center / agent
Branch/banker
42
43. How We Did It
• Analyzing complex sequences of customer behavior is
another good use of Aster
• Those insights – what influenced sales of products or what
behavior predict attrition – can then be fed into Aprimo and
used to do Digital Marketing Optimization (DMO), which is
part of Integrated Marketing Management (IMM)
• Unlike the campaign/dialogues parts of Aprimo, IMM focuses
on optimizing marketing spend, or more to the point in this
story, correlating spending and impact
• Putting this all together requires the complex behavior
analytics from Aster, the historical context from Teradata,
and the spending analytics from Aprimo
43
44. Digital Marketing Attribution – Aster and Aprimo
Functional Overview
Social Digital Marketing
Attribution Spend
Mobile (Aster Appliance) Management
(Aprimo)
Web Teradata Integrated
Multi-Channel
Channel Intelligence
Physical & Logical Model Execution
POS (Aprimo)
Call
Center Teradata Marketing
Operations
Integrated Database
Media Customer Hub
44 44
I/F to
Other Apps
45. How Teradata Aster, Teradata, and Aprimo Fit Together
in a Logical Banking Architecture
Data Sources ETL Data Platforms Analytics/Reporting Users
High
Performance Aster MPP
Unstructured SQL, SQL-MapReduce
Data Direct Loading Analytic DBMS Data
- Text (social In-Database Investigative Scientists
media, email) Analytics Analysis
- Sensor
Diverse
Data nPath Analysis
Semi- Customer Business
(customer data, metadata, …)
structured Data Management,
- Machine logs
Users
Risk, Fraud,
SIntegrated for 360°
- Clickstream FPM, Operations
Dimensional Data
- Tick-data
APRIMO Marketing Studio
Reporting
Core Banking Multi-Channel
Campaign
System Data Marketing
ETL Management
Infrastructure “ARM”
(Structured &
Relational Data)
Business Customers
Applications Mobile/Web
3rd Party Data (Online & Mobile)
-
-
Credit Bureau
SaaS Provider Teradata SAS IN-DB
Data Cloud
BI Tools Business Users
(Microstrategy, IBI, SAS Analyst
Mobile/Web
45 Tableau, Cognos)
47. Max Ideas on Better Marketing
• Which idea would you vote for?
Max Ideas Event Based GoldenPath Attribution
Campaigns to Analysis to Analytics to
increase increase focus
relevance channel Marketing
effectiveness spending
Max #1 Max #2 Max #3
47
49. For More Product Information
• If you’re in the banking industry, you may want to look at
Teradata offers at http://www.teradata.com/industry-
expertise/financial-services/
• For more Teradata and Aster Data product information, see:
www.teradata.com,
www.asterdata.com
• A good attribution paper is “Integrated Marketing Management:
Using Multi-Touch Attribution for Deeper Insight into the
Customer Journey”
• For more information on Aprimo Relationship Manager, see:
http://www.aprimo.com/Products_.aspx?id=2265
• For more information on Aprimo Real Time Interaction Manager,
see: http://www.aprimo.com/Products_.aspx?id=2266
49
50. Check Out Mercedes’ and Matt’s Ideas, Too!
See Part 2 and 3 StagnoBank videos on YouTube.com
3 - Matt:
Consumer Apps,
Alerts, Geo
2-
Mercedes:
Better
Customer
Service
50
51. Other BSI Episodes???
You can find more episodes at www.bsi-teradata.com or on
YouTube (keywords: BSI Teradata Case):
> Case of the Defecting Telco Customers
> Case of the Misconnecting Passengers
> Case of the Credit Card Breach
> Case of the Retail Tweeters
> Case of the Fragrant Sleeper Hit
> Case of the Dropped Mobile Calls
Corresponding “How We Did It” PowerPoints are available, too, at
www.slideshare.net (keywords: BSI Teradata Case)
51
This slide shows another of Aprimo’s integrated analytics- Cross Segment Analysis. Here you can easily see the performance of various channels over time, and could also quickly change this chart to show the performance of any segment of customers, across channels, over time. So, for example, you could quickly substitute customer age ranges across the top and show the performance of different communications channels by age segments- or customer value segment, or by any other customer attribute.
In this screen shot of Aprimo Relationship Manager, you can see an example of an event based (or complex trigger based) campaign. Event based campaigns allow you to watch for specific behaviors, or combinations of behaviors, by customers so that you can quickly respond with an appropriate message or offer.The Large Withdrawal which is the primary characteristic of this segment of customers actually implements a fairly complex rule to identify customer that have exhibited a specific behavior (or combination of behaviors) in the last x time period. For example, a large deposit may be defined based on individual characteristics- so it might be calculated to identify customers who have made a deposit that is at least 500% greater than their individual average deposits over the last 12 months. This provides much greater accuracy and relevance than stipulating a set amount of deposit- so a $10,000 deposit may be a “large” deposit for one person, but might not be a big deal for someone else.
Likewise, an event trigger could be setup for anyone who initiates an automatic deposit into their account- eliciting an automatic email from the bank, thanking them for signing up, and possibly cross-selling other offers that have been found through analysis to be attractive to people who just started automatic deposits. The offers can be different, and even use different channels, based on any attribute of the new depositors.
Notice that Teradata’s FSLDM already includes this event trigger marketing capability, and links it to both individual customers and campaigns (not shown on this screen).
Aprimo Relationship Manager allows you to create segments in 5 different ways:Segments can be created directly from analytics, as we saw earlierSegments can be imported from a third party, such as an analytics group, or MSPSegments can be created with a simple, point and click user interface, known as Selection Manager, that is a standard component of ARMSegments can be created by selecting tables and fields from the database, orSegments can be created from custom SQL that is written to address very complex scenarios
Purchase of specific products can also quickly be analyzed by channel effectiveness- which illustrates which channels are most used to purchase various products by various segments- such as the 20’s segment.
An existing TD customer, Flyers has invested in a new Aster system to support Digital Marketing Optimization/Attribution for its Personal Insurance Business. Flyers sells Home and Auto Insurance to prospects via the flyers.com. To drive business through the website, the Marketing department runs campaigns via various online channels like Search, Display Advertising, email, newsletters and social media (Twitter, Facebook, etc). Flyers, like most insurers, is facing stiff competition from Progressive when reaching out to individual prospects via the online channel. The Aster system is being brought in to understand referral channels and their effectiveness, attribution of conversions to referring channels for optimizing future spend, and understanding customer behavior through various sections of the flyers.com website. The Aster system will receive data feeds from weblogs, Omniture data files and Doubleclick logs. In addition to this, the ‘user’ demographic data which is stored in the EDW also needs to be loaded into Aster for further analysis of marketing spend by demographics. To perform attribution analysis, the customer needs to perform URL decoding and sessionization on the raw logs and then calculate weighted contributions of the conversion to referring channels based on user interaction (search click, display ad view, email newsletter, etc) Once attribution is complete, these results need to be combined with the user demographic data to understanding channel effectiveness across different demographics. Understanding customer behavior on flyers.com requires path and pattern analysis on the weblog data to calculate click-through rates, golden path to conversion and drop-off rates. Additionally, once the analysis in Aster is complete, the refined results from Aster need to populate the TD warehouse and become part of operational reports and dashboards which are served to the MIS system running on IBM Cognos over Teradata.
Channel analysis can also easily include individual branches, thus allowing you to identify transactions, deposits, or services that are provided by each branch and TO WHOM they are provided. By segmenting your customers by value (or profitability), you can see which branches are used by your most valuable customers, and you would naturally want to avoid closing those branches, unless you individually contact those high value customers to help them find another branch or encourage them to use other channels for the types of transactions that THEY perform most often.Event Based Marketing is generating response rates of over 50% at leading banks around the world, and many of them are reallocating budget from lower performing channels or programs to this higher performing approach