Upload
magnify-analytic-solutions
View
345
Download
0
Tags:
Embed Size (px)
DESCRIPTION
The nature of sales in retail banking has changed dramatically. While there is a renewed pressure to grow accounts, the techniques banks have traditionally used to acquire new accounts have become less effective. As consumer preferences continue to shift and non-traditional competitors continue to disrupt the market, the ROI of acquisition techniques like batch mail and branch cross-sell will continue to decline. In order to thrive, banks need to leverage the tremendous amount of data they have on each of their customers to drive more profitable and satisfying customer interactions across all of their channels. This presentation will: • Identify the market trends impacting banks’ growth strategies. • Explore the role of marketing and risk analytics in making better acquisition decisions. • Introduce best practices for implementing a more holistic approach to account acquisition.
Citation preview
Death of a Salesman: Account Acquisition in a New Environment
April 2, 2013
Death of a Salesman: Account Acquisition in a New Environment
April 2, 2013
Zoot Enterprises, Inc. Proprietary & Confidential Information.
2
Tom JohnsonVice President, Strategic Alliances
Zoot Enterprises
Ed O’BrienDirector Banking Channels
Mercator Advisory Group
Keith Shields Chief Analytics Officer, Magnify Chief Credit Officer, Loan Science
3
• Industry Overview• More Intelligent Decisions through Analytics• Next Generation Account Acquisition• Q & A
AGENDA
4
CHANGING MARKET CONDITIONS SIGNIFICANTLY
IMPACT FI GROWTH STRATEGIES
• Financial Institutions are under intense pressure to perform, even though the business fundamentals are challenging• Reduced fee income• Increased costs• Reduced revenues, net interest income, and profitability
• FIs are facing intense pressure to increase their financial performance throughout their LOBs and throughout their portfolios
• They need to find ways to profitably grow their portfolios in new and creative ways
5
CONSUMERS ARE CONSOLIDATING THE NUMBER
OF INSTITUTIONS THEY USE
1.41.6
1.21.31.21.4
1.2
2.3
1.4 1.41.41.41.21.21.31.2
2.1
1.6
4.8
2.72.5
2.7
2.12
1.6
2.6
1.6
OtherOnline brokerage/investment
Online only bankBrokerage firms Auto lendersCredit unionsMortage lendersCredit card banksFull service banks
2012 (4.9 mean of all financial institutions)
2011 (4.9 mean of all financial institutions)
2010 (6.3 mean of all financial institutions)
Mean Number of Financial Institutions Used by Households by Type(Base = Those with FI relationships by type of FI)
6
CONSUMERS ARE MORE LOYAL TO THEIR PRIMARY FI
14%
86%
12%
88%
10%
90%
Yes No
2010
2011
2012
Have You Changed Your Primary Financial Institution in the Past Two Years?(Base = All)
7
MOST PREFER IN-PERSON COMMUNICATION WHEN LEARNING
ABOUT NEW FINANCIAL PRODUCTS
8%
19%
2%
6%
6%
7%
23%
28%
None of the above
Other
Teller-assisted videoconference
Chat online at FI website
Telephone call with account specialist
Electronically at ATM or kiosk
In person with teller or greeter
In person with an account specialist
Preferred Method for Becoming Aware of New Financial Products and Services(Base = All)
8
INDUSTRY OVERVIEW:FI PROFITABILITY SOLUTION
POSITIONING
Database Layer
Application Server Layer
Core Banking System Layer
Integration Layer
Channel Systems
Underlying FI Infrastructure
Customer Analytics Predictive
AnalyticsCRM
Data Cleansing and Quality
Business Intelligence
9
INDUSTRY OVERVIEW:VARYING PROFITABILITY
PERSPECTIVES
FI Profitability
Database/Data
WarehousesLOB and Legacy
Systems
Financials Systems
Operations Systems
Channels Systems
Data Cleansing
and QualityMarketing and CRM Systems
BI, Reports, KPIs, and
Dashboards
Customer Analytics
Profitability Analytics
Best Practices Reviews
Strategy Consulting
FI-Centric Approaches
FI/ISV Partnerships
Consulting Partners
ISV Products and
Consulting Services
10
INDUSTRY OVERVIEW:COMMON CATEGORIES OF
ANALYTICS SYSTEMS
• Metadata
• Master data management
• Data modeling
• Business intelligence
• Dashboards
• Visualization
• Reporting tools
• Querying capabilities
• Branch
• ATM
• Online
• Mobile
• Call centers
• Multichannel
• Databases
• Data warehouses
• Data marts
• Core systems
• CRM
• Web
• Social media
• Predictive analytics• Customer
experience• Profitability models• Risk and compliance
models• Network analytics
• Real-time decisioning
• Content management
• Campaign management
• Event management
BusinessSystems and Data Sources
Data Mgmt
CustomerInsight
DecisioningModels
ChannelMgmtSystems
11
ANALYTICS-DRIVEN DECISIONS
• Why do banks (or any lender) invest in analytics? • Applying analytical techniques, particularly predictive modeling, to
customer data gives forward-looking insight into customer behavior.
• Understanding future customer behavior is integral to making better decisions and driving lender profitability from two primary perspectives:1. Marketing / Pricing – What loan parameters (APR in particular)
acquire the customer’s business?2. Credit Risk Management – Will the customer default on the loan? Is
his business worth having?
• Death of a Salesman? Possibly. • The renewed appetite for profitable growth (note Ed’s
presentation), combined with the explosion of available customer data, make the time right for automatic, realtime, analytically-informed lending to customers.
12
MARKETING AND CREDIT RISK APPLICATIONS
• The need for analytics within the Marketing and Credit Risk Management disciplines is pervasive.
• A recent survey of business technology professionals (see below) indicates that much of the interest in Big Data and Analytics is driven by (or at least correlated with) Marketing or Risk Management needs.
Data: Information Week Analytics, Business Intelligence and Information Management Survey of 417 business technology professionals at companies using or planning to deploy data analytics, BI or statistical analysis software, October 2012
CREDIT RISK NEEDS
MARKETING NEEDS
13
MARKETING ANALYTICS & CREDIT RISK ANALYTICS
• So lenders can make better decisions and drive profitability through “Credit Risk Analytics” and “Marketing Analytics” (not exclusively of course).
• Let’s define these terms that we’ll use colloquially throughout the presentation:• Credit Risk Analytics: empirically-based quantitative
techniques (e.g. statistical models) aimed at understanding, predicting, and controlling the level of credit risk associated with a consumer loan applicant and/or portfolio
• Marketing Analytics: empirically-based quantitative and qualitative techniques (e.g. statistical models, segmentation) aimed at understanding, predicting, and classifying the likely purchase behavior of a consumer or group of consumers
14
THE IMPORTANCE OF CREDIT RISK ANALYTICS
• Let’s show the importance Credit Risk Analytics with an example:• If a lender makes a $100 profit on a paying loan and loses $400 on a
defaulting loan, then it has to book 4 paying loans for every defaulting loan just to break even.
• Another way to state the above bullet is this: a loan applicant should have at least an 80% chance (4:1) of paying as agreed to be considered for approval.
• How do we determine if an applicant has at least an 80% chance of paying as agreed? Empirically-derived, demonstrably and statistically sound models of course. Almost all lenders use these in some form…• Generic credit bureau scores -and/or-• Custom scores derived from contract attributes (LTV, PTI) and credit
bureau attribute libraries (from Zoot of course)
15
RISK MODELS
• Continuing with the example…the importance of robust, predictive “risk models”:• So what if a lender is drawing from a population that is inherently
85% good (85% will pay off a standard loan) and 15% bad (15% will default on a standard loan)? Shouldn’t that lender always be profitable?
• It is crucial that the statistical model (or something equivalent) used by the lender to predict the likelihood of default be able to SEPARATE the good from the bad. Example below:
• If the model is incapable of any separation whatsoever, it will issue a 15% probability of default (PD…in bank terminology) for every proposed contract. This is less than a 20% cutoff, so we approve everything.
• Thus the lender’s profitability is easy to calculate…suppose 1,000 contracts will be booked: P = 850*$100 – 150*$400 = $25,000 … is it that easy to make money? By always saying yes to loan applications? That would be Death of a Credit Analyst. But it’s not that easy…
16
“GOOD AND BAD I DEFINE THESE TERMS, QUITE CLEAR, NO DOUBT, SOMEHOW” --
BOB DYLAN• Risk models must, first and foremost, distinguish between good
and bad (i.e. rank order the risk):• “All models are wrong, but some are useful”. -- George E.P. Box
• No individual customer has a 15% chance of default. All individual customers effectively have a 0% chance or 100% chance of default (they either do or they don’t).
• Profitability is far greater if the model is able to issue higher PD predictions for defaults than for payers. This is what happens when a model is able to RANK ORDER the risk. See the next bullet.
• Back to our example: Suppose the model predicts a PD of 25% for half the defaults, and 10% for the other half. In turn it issues a PD prediction of 25% for ¼ of the payers, and 10% for ¾ of the payers.• Before we calculate profitability we’ll note that the profitability on all
applicants with a predicted PD of greater than 20% is $0. They are declined based on the breakeven calculation on the previous slide. So…
• P = [850*.25*$0 + 850*.75*$100] – [150*.5*$0 + 150*.5*$400] = $33,750
• This is a 35% increase in profitability due to having a better predictive model.
17
MARKETING ANALYTICS
• We’ll continue the discussion with Marketing Analytics, which has become a staple in the retail industry (Target for example)…• The analytical techniques used to predict how “in-market” a
customer is for clothes, diapers, etc… can be the same ones used to predict how “in-market” a customer is for a loan.
• The advantage of using predictive analytics to identify the customers most likely to take up a loan is that it “expands the base of incrementality” associated with a loan offer. • In other words, identifying the groups of customers that are most likely to
buy (take up a loan) is tantamount to identifying the groups that contain the majority of the “incremental” sales (contracts).
• If I can get the most of the incremental sales by making an offer to only a small fraction of the population, then I can squeeze most of the benefit from the offer at a fraction of the cost. Example next slide.
18
ADDING BUSINESS VALUE THROUGH MARKETING ANALYTICS: AN
EXAMPLE• Intelligent use of Marketing Analytics enables lenders to generate
incremental loans cost-effectively and efficiently. An example with assumptions:• An untargeted (no model), incentivized loan offer to 100,000 customers
increases the “take rate” by 10%.• The revenue per incremental loan is $250.• The cost of communicating the offer to 100,000 consumers is $20,000.• The cost of the incentive is $20 per loan.• When we apply a predictive model, we split the population into 4 groups
(1=most likely to take…4=least likely to take)
No Model
Use a Model
Profit = $10,000
Profit = $14,00040% improvement
Population OrganicTake Rate
OfferTake Rate
Incremental “Takers”
100,000 10% 11% 1,000
Incremental Revenue
$250,000
Model Rank Population Organic Take
RateOffer
Take RateIncremental
“Takers”
1 25,000 18% 19.80% 4502 25,000 14% 15.40% 3503 25,000 6% 6.60% 1504 25,000 2% 2.20% 50
IncrementalCost
Incremental Profit
$240,000 $10,000
Incremental Revenue
$112,500 $87,500 $37,500 $12,500
Incremental Cost
Incremental Profit
$104,000 $8,500 $82,000 $5,500 $38,000 ($500)$16,000 ($3,500)
19
THE INTERSECTION OF MARKETING AND CREDIT RISK
• Practically we should not deploy Marketing Analytics in a lending environment without doing sound Credit Risk Analytics at the same time.• Marketing analytic efforts are typically aimed at increasing response (and thus
sales). Doing so can also increase credit risk, which means credit losses can easily wipe out the gains had by improvements in targeted marketing efforts.
• The right loan offer needs to be defined as the one that maximizes incremental profit…after incremental credit losses are factored in.• Making the right loan offer is an analytical exercise that requires the
intersection of Marketing and Credit Risk Analytics. Through the Magnify-Loan Science partnership, we specialize in this type of exercise…and we deploy through Zoot.
• We see pre-approval models as being the perfect example of this intersection. And we will show the work we’ve done in auto…• Common with credit cards…interest rates and credit limits are tested to
determine the impact on response and yearly interest revenue.• Pre-approval models for auto are more complex, because the presence of
collateral means we have to solve for very important variables like loan-to-value and term.
20
DEATH OF REDEFINING A SALESMAN: USE ANALYTICS TO TARGET OFFERS
• The tyranny with almost all pre-approval programs is that the customers who respond to them are the ones you least want to give credit to.
• Customers with low credit bureau scores are generally the ones that respond to pre-approval offers, and the more exposure the lender is willing to take, the better they respond.• See the example below from an auto captive…the data are doctored
somewhat but not to the point where the message is changed:
Population: Existing Customers and Prospects
FICO Score and Control Buy Rate
Target Buy Rate “Incrementality” LiftPre-Approval Amount
FICO <= 680 and Pre-Approval >= $30,000 1.93% 2.17% 0.23% 12.14%
FICO <= 680 and Pre-Approval < $30,000 0.93% 1.52% 0.60% 64.27%
FICO > 680 and Pre-Approval < $30,000 2.34% 2.32% -0.01% -0.46%
FICO > 680 and Pre-Approval >= $30,000 2.64% 2.56% -0.08% -2.85%
21
MAKE THE RIGHT OFFER RESPONSIBLY…
• Can we have the high lift associated with high-risk customers AND control the risk of the pre-approved portfolio? Yes, probably so. Consider turning the traditional PD (probability of default) model on its head:
• Traditional: PD = f(credit score, LTV, PTI, term,…)
• Pre-approval: LTV = f(PD, credit score, PTI, term,…)• See the auto captive example below, where we control for PD and solve for
LTV
Will yield Tier B performance.Tier B rate can be guaranteed in the pre-approval offer.
Will yield Tier C performance.Tier C rate can be guaranteed in the pre-approval offer.
Will yield Tier D performance.Tier D rate can be guaranteed in the pre-approval offer.
Credit Score PD Term LTV Limit581-600 4.0% 60 60%601-620 4.0% 60 75%621-640 4.0% 60 85%641-660 4.0% 60 95%661-680 4.0% 60 100%
Credit Score PD Term LTV Limit581-600 8.0% 60 75%601-620 8.0% 60 85%621-640 8.0% 60 94%641-660 8.0% 60 100%661-680 8.0% 60 105%
Credit Score PD Term LTV Limit581-600 15.0% 60 90%601-620 15.0% 60 98%621-640 15.0% 60 105%641-660 15.0% 60 112%661-680 15.0% 60 120%
22
REDEFINING SALES OCCURS WHEN THE BENEFITS OF ANALYTICS AND TECHNOLOGY
ARE TANGIBLE…
• In the example on the previous slide we achieved two important outcomes:
1. We confined our targeting to the customers who, according to our best Marketing Analytics, would respond to our offer.
2. We confined our offers to those that, according to our best Credit Risk Analytics, would be profitable at a controlled level of risk and price.
• We also achieved a third very important outcome, which I’ll offer as a conclusion: we generated incremental loans that subsequently contributed an additional $8 mils profit per year.
• But what is missing from this good story? DEPLOYMENT. Analytic tools and technologies must be made available to operational systems when:• Credit decisions are made, or when • A list of targeted customers is generated, or when • The parameters of pre-approval offers are specified
• ... and this is where Zoot fits in: realtime deployment of analytic tools so that interactions with the customer are informed, targeted, and profitable…
23
NEXT GENERATION ACCOUNT ACQUISITION
24
CONSISTENT EXPERIENCE ACROSS CHANNELS
25
TYING IT ALL TOGETHER: OFFERS REPOSITORY
26
OPTIMIZED OFFERS
27
BETTER DATA, BETTER DECISIONS
28
QUICKER TIME TO MARKET
29
SUMMARY
“The only thing you got in this world is what you can sell. And the funny thing is that you're a
salesman, and you don't know that.”
~Arthur MillerDeath of a Salesman
• Sales in retail banking isn’t dead, but it has changed.
• New channels and more interactions across all channels.
• Analytics available to make more intelligent decisions.
• Underlying technology must support next generation account acquisition techniques.
30
Ed O’Brien, Director Banking Channels, Mercator Advisory [email protected]
Keith Shields, Chief Analytics Officer at Magnify Analytic Solutions and Chief Credit Officer at Loan [email protected]
Tom Johnson, Vice President, Strategic Alliances, Zoot [email protected]
QUESTIONS ?