Big Data, Voice Analytics and Unstructured Data Help Tackle the Compliance Conundrum

Big Data, Voice Analytics and Unstructured Data Help Tackle the Compliance Conundrum

Omer SohailLloyd WirshbaDeloitte Consulting LLP

October 17, 2013

The CFPB is a data driven regulator whose rulemaking authority influences the entire consumer finance spectrum

Source (L to R), accessed September 12-17, 2013, web articles

http://www.americanbanker.com/issues/178_176/customers-are-now-banks-greatest-regulatory-threat-1061975-1.htmlhttp://www.americanbanker.com/issues/178_164/lenders-press-cfpb-to-delay-qualified-mortgage-regs-start-date-1061547-1.html http://www.americanbanker.com/issues/178_131/cfpb-to-debt-collectors-were-watching-you-1060478-1.html

The CFPB Complaint Database provides perspective on industry trends and issues

Source: CFPB complaint database, accessed August 28, 2013

• Contains over 130,000 complaints across seven products, where the date of first data collection varies by product

• Mortgage complaints represent both the majority and an increasing trend: 53% overall, from 48% in March 2012 to 63% in March 2013

• The timely response rate is 97.6% (acknowledgement within 15 days)

• Consumers disputed 21.2% of proposed resolutions, while 12.6% are still awaiting response

20278; 15% 3769; 3%

24850; 18%

10012; 7%192; 0%

72436; 53%

4528; 3%

CFPB Complaint Database by Product

Bank account or service (March 2012)Consumer loan (March 2012)Credit card (November 2011)Credit reporting (October 2012)Money transfers (April 2013)Mortgage (December 2011)Student loan (March 2012)

Finding complaint root causes requires identifying predictive indicators within other data sources

Source: CFPB complaint database, accessed August 28, 2013

• Nearly 85% of mortgage complaints relate to loan mods and servicing

• During origination, consumers find another lender instead of complaining

• Consumer awareness is increasing• The mortgage servicing arm of a large

bank sought to identify the root cause of an increase in mortgage complaints

• Each complainant that could be connected to internal records contacted the bank at least three times prior to the CFPB complaint

The CFPB database has no predictive indicators, so it must be joined with other sources to analyze and respond to CFPB complaints

2011 / 12

2012 / 2

2012 / 4

2012 / 6

2012 / 8

2012 / 10

2012 / 12

2013 / 2

2013 / 4

2013 / 60

1000

2000

3000

4000

5000

6000

Mortgage Complaint Issues over Time

Loan modification,collection,foreclosureLoan servicing, payments, escrow accountApplication, originator, mortgage brokerSettlement process and costsOtherCredit decision / underwriting

Com

plai

nts

The CFPB taxonomy provides more issue types for credit card complaints

Source: CFPB complaint database, accessed March 31, 2013 Nilson Report, February 2012 (outstandings at end of 2011)

• The top five of 33 issues constitute 46% of the 24,850 credit card complaints

• Integrate data from complaint handling, process improvement initiatives and product offerings to train a system that suggests a bank-specific root cause

Product Num. Issues

Bank Account/Service 5

Consumer Loan 7

Credit Reporting 5

Money Transfers 6

Student Loan 3

16%

10%

7%

7%

7%6%

48%

Credit Card Complaint Issues

Billing disputesAPR or interest rateCredit reportingIdentity theft / Fraud / EmbezzlementClosing/Cancelling accountOtherNamed other issues (27 categories)

Providing relief is not a primary tactic to reduce consumer disputes

• Despite resolutions in favor of the consumer decreasing, consumer disputes also decreased

(Originally the CFPB only had one category of relief. In May 2012, the options were expanded to monetary and non-monetary relief)

• Better communication with customers may help firms manage consumer expectations and support process improvement

Source: CFPB complaint database, accessed August 28, 2013

2011 / 11

2012 / 1

2012 / 3

2012 / 5

2012 / 7

2012 / 9

2012 / 11

2013 / 1

2013 / 3

2013 / 5

2013 / 70%

5%

10%

15%

20%

25%

30%

35%

Complaint Resolution Statistics

DisputesResponses in favor of consumer

Providing relief has a different effect on dispute rates depending on the product

• For consumer loans and mortgages, providing relief reduces the dispute rate by 7%

• For other products in the original database, providing relief reduces the dispute rate by 13%

• To better allocate resources while reducing complaint volume, it is critical to understand the relationship between providing relief and customer satisfaction. Which opportunity is greater?

Source: CFPB complaint database, accessed August 28, 2013

Consumer

loan

Mortgag

e

Bank a

ccount o

r serv

ice

Credit c

ard

Studen

t loan

0%

5%

10%

15%

20%

25%

30%

Dispute Rate by Product and Response

Closed with relief Closed without relief

Disp

ute

Rate

For bank account and service complaints, disputes are greatly reduced with fee refunds

• Nearly a quarter of bank account and service complaints resolved without relief were disputed

• Compared to providing non monetary relief, providing monetary relief reduces dispute rates from 18% to 9%

• Does the refund amount relative to the fee(s) responsible for the complaint impact the dispute rate?

Source: CFPB complaint database, accessed August 28, 2013 Pennsylvania PIRG CFPB Complaint Database analysis, accessed September 19, 2013 http://www.pennpirg.org/sites/pirg/files/reports/Big%20Banks%2C%20Big%20Complaints%20screen%20vPA.pdf

Closed (7

58)

Closed w

ith ex

planati

on (10550)

Closed w

ith m

onetary

relief

(4473)

Closed w

ith non-m

onetary

relief

(1070)

Closed w

ith re

lief (1

220)

Closed w

ithout r

elief

(1990)0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

66% 65%80%

72%86%

74%

21% 24%9%

18%

14%26%

12% 11% 11% 10%0% 0%

Bank Account or Service Complaint Disputes

Pending ResponseDisputedNot Disputed

Complaints are more likely to originate from zip codes with older and more affluent residents

• Adjust for penetration, separating out each banking product

• Complainants in the database may not precisely represent the customer population at a financial institution

• Banks should understand their exposure, especially which involves consumer segments that are important to the CFPB: underbanked, lower income, or other disadvantaged populations

Source: CFPB complaint database, accessed August 28, 2013

Population Complaints0%

10%20%30%40%50%60%70%80%90%

100%

Mortgage Complaints by Median Age

D (>= 41)C (37.5-40.9)B (33.5-37.4)A (<33.5)

Population Complaints0%

10%20%30%40%50%60%70%80%90%

100%

Mortgage Complaints by Median Income

H (>= $69,000)G ($52,000-$68,999)F ($41,000-$51,999)E (<$41,000)

1.17

1.05

1.00

0.75

1.43

1.08

0.84

0.62 Values > 1 indicate a larger share of complaints from that age/income group than is represented in the population

Deriving value from complaint data analysis

Components of an enterprise-wide complaint analytics capability

Social Media

News

Call Transcripts and Agent Notes

Focus Groups

Online Chat

Email

With structured data, collect unstructured and external data

Integrate with a cross-functional complaint management program

Data and Analytics

Governance and Controls

Escalation Process

Employee Training

Regulatory Reporting

Independent Compliance Audit

Perform historical reporting and advanced data analytics

Data aggregation and cleansing

Create relevant variables

Develop predictive models

Develop reason codes and business rules

Refine modeling outcomes

Building and deploying leading analytics requires a combination of domain, data management, data intuition, statistics and technology skills

Complaint analytics drive value across all bank functions

Operations Consumers

Compliance Marketing

Improve Customer Service

Decrease C

hurn

Pro

duct

P

rom

otio

n

Product Design

CFPB G

uidelines

Com

plai

nts

Man

agem

ent

Improve customer

outcomes (e.g. loyalty, spend)

Enhance investment to

improve satisfaction

Proactive monitoring of

customer reactions

Identify and mitigate high

risk complaints

Refine channel marketing strategy

Proactive approach to regulators

ComplaintAnalytics

A complaint analytics system includes multiple use cases for each function

Complaint Analytics Applications

1

2

3

Text analytics to understand the voice of customer

Voice analytics to address complaint escalation

External lifestyle data to support complaint handling

Consumer treatments post fraud, disputes, complaints4

Text analytics can be used to understand the voice of the customer

Social Media

News

Call Transcripts and Agent Notes

Focus Groups

Online Chat

Email

Collect structured and unstructured data streams across products and channels

1

Classify Text to Quantify Known Issues

Cluster Documents to Identify Emerging Issues

Analyze Sentiment and Top Keywords to Improve Predictive

Models

Integrate with Dashboards• Hypothesis Testing

• Emerging Trends/Surges

• Product Monitoring

Text analytics reveals origination experience drivers for wealth management customers1

A difficult process with poor communication leads to dissatisfaction. The bank in blue should focus on process

The top factors driving satisfaction and likelihood to recommend depend on competent bank staff

Source: Deloitte survey and analysis

Near real time voice analytics enables integration with predictive modeling and reporting

Customer Call

CSR

Customer

QueueRecordCalls

Voice Analytics (phonemes, text)

Contact Center

Customer Warehouse Prediction

Score

Sequence Based Predictive Model

CRM

Demo-graphic, Lifestyle Data

Customer Operations

Accelerators

Outbound or Transfer Call Center

Sample metrics• Wait time• Transfer rate• First contact resolution• Abandon rate• Escalation rate

IVR and Switch Logs

2

Escalated Complaints

Post Complaint Churn Repeat Callers CFPB Escalation

Near real time propensity model gives agents feedback

Sequence analysis identifies churn and reduced spending factors

Root cause and metric driven analysis to reduce callbacks

Join internal data with CFPB database to preempt escalation

Approach

• Agent desktop integration

• Specialist call center agents

• Outbound call center

• Real time retention offer generation

• Decisions based on customer lifetime value forecasts

• Cross channel context for agents

• Volume forecast and IVR changes

• Test and learn

• Reperform tests in CFPB examination manual

• Integrate with complaint responses

Implement

Model

Voice analytics can provide data for predictive models that prevent complaint escalation2

External lifestyle data can enhance the power of predictive analytics3

AcxiomAM BestAMAAmerican Housing SurveyAmerican Tort Reform FoundationBurueau of Labor StatisticsCap IndexCarfaxCDS Hail DatabaseCensus PointChoicepointCorporate Research BoardDataListerDirectory of US HospitalsDun & BradstreetEASI AnalyticsEEOCEquifaxESRIExperianFastcase Legal Research SystemFlorida Tax Assessment RecordsFulbright Lititgation Trends SurveyInsurance Information InstituteInsurance Institue for Highway SafetyInternal Renvue ServiceKnowlege Based Marketing (KBM)Lawyer Data – Florida & CaliforniaLexisNexisMartindale/Hubble Attorney ListingMRI Purchasing PropensitiesNFIRS – National Fire Reporting NHTSAOSHAUS Census

Wage DataWealth IndicatorsUnemployment StatsEEOC Complaints DBEc. Freedom Index Aggregated IRS DataOccupational Codes

Real Property DataAffluent HomeownersHome Equity BorrowerGovt. Housing SurveyHome Value ScoringForeclosure Data

Purchase BehaviorsPurchase PropensitiesSpend by Category DTC Spend by RetailerBrand Usage StatisticsRetailer Trans DataPurchase Triggers

Crime StatisticsHail Vector DataStorm Events DBClimate Data Geographic MappingCollege RankingsFirehouse Data Fire Incident Data

Judicial HellholesFed. Case Law DBFlorida Tax RecordsLit. Trends SurveyLawsuit Climate DataDUI/DWI LawsCA/FL Lawyer DataTort Liability Index

Bus. Hazard GradeBus.Insurance ScoresBus.Financial StatisticsUCC FilingsSmall Bus. DataBus. Credit ScoreOSHA Bus. Data Tax Liens & Bankruptcy

Disability DataUS Hospital DirectoryNursing Home DataMedical Provider DataHosptial Visit StatisticsDoctor Practice DataHealth Interest Data

Auto DataCarfax Vehicle HistoryMotor Vehicle ReportsAuto Injury / Loss DataDriver Device UsageRoad Rage SurveyVIN Decoding Data

Lifestyle and Life TraitsWorking MothersActive SeniorsHigh-Tech SegmentsLife Stage ClusteringDemo. Census Data

Data Vendors Data Vendors Data CategoriesData Categories

Augmenting outcomes monitoring with external data can improve complaint handling policy3

Source: Building Consumer Trust in Retail Payments, available at http://www.deloitte.com/assets/Dcom-UnitedStates/Local%20Assets/Documents/us_fsi_Bank_ConsumerTrustPayments_July08.pdf

Integrating analytics with operations• Proactive pre-complaint issue handling.

Some segments tend to have certain complaint issues.

• Perform segment based complaint resolution subject to fair treatment

• Develop a communication approach to reduce complaints and disputes

Example: customer A expressed interest in a simple resolution process without a notarized form. Subject to fraud loss risk guidelines, if Customer B is has similar lifestyle attributes to Customer A, omitting the notarized form can help build trust

Predicting how customers react after fraud handling is key to reducing complaints and churn4

Collect Data Build Analytics

Integrate with

Operations• Convert model results to reason

codes that correspond to process changes

• Provide service reps and complaint investigators with relevant context from previous interactions• Expedite a fraud investigation• Issue a credit for the

transaction(s) under investigation

• Retention offer as appropriate• Make recommendations about

the fraud handling policy based on predicted customer profitability and behaviors

• Customer data • Demographics• Product relationships• Profitability

• Contact event data • Potentially fraudulent

transactions• Consumer trust indices and

survey data derived from an existing Deloitte study

• Deloitte administered focus group (for updated data)

• Identify indicators that explain consumer behavior following a fraud incident

• Develop a propensity model that predicts customer churn, reduced spending, or inactive accounts• Who is likely to churn?• What are the risk factors?

• Analyze free form survey and focus group data to identify trends in unmet customer needs. For example• Ensured I didn’t lose money• Stopped transactions quickly• Limited paper forms• Went beyond minimum legal

requirements

Summary• The CFPB Complaint Database provides a starting point consumer complaint

analysis, but it can provide greater insights when properly integrated with internal process and complaint data

• Big data technologies are a key component of an enterprise-wide complaint analytics and response capability

• Text analytics can be applied to identify emerging issues and focus areas for improving customer satisfaction

• Using voice analytics to support predictive modeling represents an emerging area that can provide substantial incremental benefits to complaint analytics

• External data sources can augment a predictive model or provide context for customer interactions following complaints

Copyright © 2013 Deloitte Development LLC. All rights reserved.

As used in this document, "Deloitte" means Deloitte Consulting LLP, a subsidiary of Deloitte LLP. Please see www.deloitte.com/us/about for a detailed description of the legal structure of Deloitte LLP and its subsidiaries. Certain services may not be available to attest clients under the rules and regulations of public accounting.

This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.