In this presentation Mark T. Warren (Director of Decision Science) talks about Big Data with Barclaycard, the foundations they built for it and their goals in the long term for it. Warren also discusses Barclaycard's learnings from building the foundation and how they're using these learnings and coping with market change and other challenges that can affect their long term goals.
3. A starting point
• Credit Card scoring blazes the trail for Big Data
Risk scoring dates to the early sixties
Account management scoring to the early eighties
Direct mail spurs the next wave of innovation/development
4. A starting point
• Credit Card scoring blazes the trail for Big Data
Risk scoring dates to the early sixties
Account management scoring to the early eighties
Direct mail spurs the next wave of innovation/development
• Virtually every customer touch point is highly dependent on statistical models
imbedded in near real time systems fed by a wide variety of data
• The existence of such tools … and the proper use of them by credit managers … is
the foundation of credit card management today.
5. A starting point
• Credit Card scoring blazes the trail for Big Data
Risk scoring dates to the early sixties
Account management scoring to the early eighties
Direct mail spurs the next wave of innovation/development
• Virtually every customer touch point is highly dependent on statistical models
imbedded in near real time systems fed by a wide variety of data
• The existence of such tools … and the proper use of them by credit managers … is
the foundation of credit card management today.
• So … Big Data? Big Deal
8. Building the Foundation – The Goal
• For the past 5 years Barclaycard has pursued a multi-prong approach aimed at
rolling out best-in-class tools that rely on a broad array of data and are embedded in
internally managed systems.
9. Building the Foundation – The Goal
• For the past 5 years Barclaycard has pursued a multi-prong approach aimed at
rolling out best-in-class tools that rely on a broad array of data and are embedded in
internally managed systems.
Scalable
o Central teams supporting geographically dispersed portfolios
Common toolset
o Development tools for analysts
o Scores/models for the business
Integrated
o Card data, retail data, and bureau data give a full view of the customer
o Common platform for risk and marketing purposes
10. Building the Foundation – The Goal
• For the past 5 years Barclaycard has pursued a multi-prong approach aimed at
rolling out best-in-class tools that rely on a broad array of data and are embedded in
internally managed systems.
Scalable
o Central teams supporting geographically dispersed portfolios
Common toolset
o Development tools for analysts
o Scores/models for the business
Integrated
o Card data, retail data, and bureau data give a full view of the customer
o Common platform for risk and marketing purposes
• Today, Barclaycard deploys 200+ predictive scores across its portfolios to manage
touch points throughout the customer life-cycle.
11. Building the Foundation – The Goal
• For the past 5 years Barclaycard has pursued a multi-prong approach aimed at
rolling out best-in-class tools that rely on a broad array of data and are embedded in
internally managed systems.
Scalable
o Central teams supporting geographically dispersed portfolios
Common toolset
o Development tools for analysts
o Scores/models for the business
Integrated
o Card data, retail data, and bureau data give a full view of the customer
o Common platform for risk and marketing purposes
• Today, Barclaycard deploys 200+ predictive scores across its portfolios to manage
touch points throughout the customer life-cycle.
Le t’s take a q uick lo o k at ECMsco ring platfo rm s – the o rig inalbig data so lutio n
12. “Black Box”
Processing Engine
Data management
Score calculations
Decision support
Output
Raw data
Data
Pre-process
Information Scoring Touch-points
Authorization
Module
CLI
Module
Collections
Module
Action
Action
Action
Card
Masterfile
Credit Bureau
Retail
Masterfile?
Extra?
Authorization
Collections
Partner
Third Party
Collections
Customer Service
Building the Foundation – an example
Data Processed
350G – 450G
13. “Black Box”
Processing Engine
Data management
Score calculations
Decision support
Output
Raw data
Data
Pre-process
Information Scoring Touch-points
Authorization
Module
CLI
Module
Collections
Module
Action
Action
Action
Card
Masterfile
Credit Bureau
Retail
Masterfile?
Extra?
Authorization
Collections
Partner
Third Party
Collections
Customer Service
Building the Foundation – an example
Daily Run Time
5-10 hours
14. “Black Box”
Processing Engine
Data management
Score calculations
Decision support
Output
Raw data
Data
Pre-process
Information Scoring Touch-points
Authorization
Module
CLI
Module
Collections
Module
Action
Action
Action
Card
Masterfile
Credit Bureau
Retail
Masterfile?
Extra?
Authorization
Collections
Partner
Third Party
Collections
Customer Service
Building the Foundation – an example
Scale
8-10M Customers
Up to 20 scores
15. Building the Foundation -- Learnings
1. Common operational platforms are key
Without them you can’t get scale
16. Building the Foundation -- Learnings
1. Common operational platforms are key
Without them you can’t get scale
2. Critical role of flexible analytic architecture
Not just a technical capability but a software and licensing capability
17. Building the Foundation -- Learnings
1. Common operational platforms are key
Without them you can’t get scale
2. Critical role of flexible analytic architecture
Not just a technical capability but a software and licensing capability
3. Addressing Data Privacy concerns while making data available to analysts
EU and US regulatory regimes unique and restrictive
18. Building the Foundation -- Learnings
1. Common operational platforms are key
Without them you can’t get scale
2. Critical role of flexible analytic architecture
Not just a technical capability but a software and licensing capability
3. Addressing Data Privacy concerns while making data available to analysts
EU and US regulatory regimes unique and restrictive
4. Quality models depend on market understanding
Since results must be interpretable, context is everything
19. Building the Foundation -- Learnings
1. Common operational platforms are key
Without them you can’t get scale
2. Critical role of flexible analytic architecture
Not just a technical capability but a software and licensing capability
3. Addressing Data Privacy concerns while making data available to analysts
EU and US regulatory regimes unique and restrictive
4. Quality models depend on market understanding
Since results must be interpretable, context is everything
5. Data mining has its pitfalls
Numbers do lie or,
Blindly following numbers yields poor customer experience
21. Market Change
• But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
22. Market Change
• But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
Increased regulatory oversight
o Increased scrutiny often requiring quick turn around time
23. Market Change
• But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
Increased regulatory oversight
o Increased scrutiny often requiring quick turn around time
Reduced margins
o Revenue streams such as fees are increasingly limited
24. Market Change
• But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
Increased regulatory oversight
o Increased scrutiny often requiring quick turn around time
Reduced margins
o Revenue streams such as fees are increasingly limited
Changing customer behaviour
o Reduced appetite for debt and increased demand for quality
• These trends aren’t unique to the US nor are they unique to credit cards
25. Market Change
• But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
Increased regulatory oversight
o Increased scrutiny often requiring quick turn around time
Reduced margins
o Revenue streams such as fees are increasingly limited
Changing customer behaviour
o Reduced appetite for debt and increased demand for quality
• These trends aren’t unique to the US nor are they unique to credit cards
So our goal is to be the ‘Go-To’ bank
26. • But our industry is changing
New competitors (PayPal, etc.)
o PayPal, etc. utilize newer platforms to provide unique services
Increased regulatory oversight
o Increased scrutiny often requiring quick turn around time
Reduced margins
o Revenue streams such as fees are increasingly limited
Changing customer behaviour
o Reduced appetite for debt and increased demand for quality
• These trends aren’t unique to the US nor are they unique to credit cards
So our goal is to be the ‘Go-To’ bank
• In short … if people want to be our customers we’ll have a long-term viable business
model
Market Change
27. Can Big Data help us become the ‘Go-To’ bank?
Big Data solutions are often sold on the following merits:
•Reduced costs
Disk, Processing, Back-up
Open source software
•Faster analytics
MPP/IMP
Real-time/Near Real-time processing
28. Can Big Data help us become the ‘Go-To’ bank?
Big Data solutions are often sold on the following merits:
•Reduced costs
Disk, Processing, Back-up
Open source software
•Faster analytics
MPP/IMP
Real-time/Near Real-time processing
While these savings can be significant there is one simple obstacle …
… we’ve already made significant investments in such technology.
Our costs are already sunk – adopting newer platforms is an incremental cost
29. Can Big Data help us become the ‘Go-To’ bank?
• Getting people to want to be our customers takes way more than keeping our losses
in check
• We need to have a more complete view of the customer
Are we making their lives easy when they use our product?
Are we meeting their needs in a responsible way?
Are we adding to their lives by providing products and services that go beyond
commodity features?
30. • Getting people to want to be our customers takes way more than keeping our losses
in check
• We need to have a more complete view of the customer
Are we making their lives easy when they use our product?
Are we meeting their needs in a responsible way?
Are we adding to their lives by providing products and services that go beyond
commodity features?
• Bureau data, card usage data, and payment data doesn’t give us much insight into
these questions.
So Big Data is not just about adding additional X’s to the mix …
…. It is about creating new Y’s to investigate
Can Big Data help us become the ‘Go-To’ bank?
31. First steps …
2013 Focuses on Proof-of-Concept initiatives:
Hadoop tests (US)
SAS High Power Analytic tests (UK)
Voice of the Customer initiatives using Verint speech-to-text analytics
Customer specific web presentment (UK)
2014 takes these learnings and deploys new solutions
32. … Next steps …
The next 3 years entails:
New data (of course)
o Web logs
o Customer calls
o AID transaction data
New hardware and software to house this data
o Globally available analytic environments where cost isn’t an issue in investigating data
New skills
o Deriving information from unstructured data
o Investigating alternative modelling techniques where feasible
33. … Next steps …
The next 3 years entails:
New data (of course)
o Web logs
o Customer calls
o AID transaction data
New hardware and software to house this data
o Globally available analytic environments where cost isn’t an issue in investigating data
New skills
o Deriving information from unstructured data
o Investigating alternative modelling techniques where feasible
Key challenges:
Market understanding increasingly critical
o Cultural norms more pronounced in unstructured data
Increased complexity of implementations
o Timeliness of results increasingly critical
o Accessing a wide variety of contextual data as customers use our products
34. ... Pivotal change …
• Whereas data intensive statistical analytics has been the mainstay of Risk
Management and Marketing, Big Data opens the door to driving Operations and new
business lines.
• The beauty of this is the following:
Whereas the business case for replacing existing hardware and software that drives
today’s analytics is often weak, the Big Data business case thrives in operations.
Tackling new areas requires new investment.
With that new hardware/software in place, it is then feasible to migrate existing
traditional analytics to that new platform.
35. ... Pivotal change …
• Whereas data intensive statistical analytics has been the mainstay of Risk
Management and Marketing, Big Data opens the door to driving Operations and new
business lines.
• The beauty of this is the following:
Whereas the business case for replacing existing hardware and software that drives
today’s analytics is often weak, the Big Data business case thrives in operations.
Tackling new areas requires new investment.
With that new hardware/software in place, it is then feasible to migrate existing
traditional analytics to that new platform.
So Big Data is a Big De al
36. … the Destination
• So what does the ‘Go-To’ bank look like in 3 years for Barclays?
Seamless customer service
37. … the Destination
• So what does the ‘Go-To’ bank look like in 3 years for Barclays?
Seamless customer service
Products work for the unique needs of our customers
38. … the Destination
• So what does the ‘Go-To’ bank look like in 3 years for Barclays?
Seamless customer service
Products work for the unique needs of our customers
Unique enhancements suited to each customer’s wishes
39. … the Destination
• So what does the ‘Go-To’ bank look like in 3 years for Barclays?
Seamless customer service
Products work for the unique needs of our customers
Unique enhancements suited to each customer’s wishes
Stronger financial position for Barclays given significantly reduced costs