synergies between risk modeling and customer analytics · synergies between risk modeling and...

Synergies between Risk Modeling andCustomer AnalyticsEY – SAS Forum, Stockholm

18 September 2014Lena Mörk and Ramona Klein

Agenda

► Introduction

► Modeling in the financial sector

► Consequences from lack of alignment between riskmodeling and customer analytics

► Achieving synergies in the organization

► Wrap up

Synergies between Risk Modeling and Customer Analytics

1

2

3

4

5

Introduction

► Risk models► Used in most financial services domains to ultimately predict profitability

► Customer analytics► Marketing campaigns aim to increase customer base► “Customer analytics” team models customer behavior in order to create target marketing

► Synergies► Organizations allocate significant IT and human resources to preparing data and building

models► When the risk modeling and marketing analytics are performed in silos, organizations

waste resources, time and do not achieve optimal benefits► Effective data repositories and manipulation, appropriate modeling practices and model

risk management can create synergies between risk modeling and customer analytics


► Introduction




► Wrap up


1

2

3

4

5

Models are used regularly in the financial sector

Features of predictive models:► Various techniques, e.g.:

► Generalized Linear Models(GLMs) loss cost andconversion (insurance)

► GLM logistic regression(banking)

► Statistical model diagnostics► Validation techniques► Ability to test predictiveness► Ability to incorporate business

judgment► Not a black box (can follow

model development, statistics,validation process)


►Predict fraud►Predict reserves►Evaluate sales force

►Predict loss cost►Predict credit risk►Build rate structure►Build tiers►Loss given default

EnhancedDecisionMaking

CustomerValue

Pricing andRisk Analysis

►Predict demand►Predict retention

(e.g. loyalty scores)►Risk segmentation

CustomerAnalytics Others

Model development is similar across products anddepartments

Documentation

Quality assurance

Experience and knowledge

Preparation Data Single factoranalysis

Multi factoranalysis Validation Implementation

► Project plan► Establish

scope► Involve all

stakeholders

► Gather data► Prepare files

for modeling► Check and

clean data► Reconcile

against othersources

► Initial dataexploration

► Univariateanalysis

► Correlationstatistics

► Buildpredictivemodel

► Regressionanalysis (e.g.,GLMs)

► Iterativeprocess

► Statisticaltechniques tovalidate modelstructure

► Holdoutsamples tovalidatepredictiveness

► Analyzecompetitive /profitabilityimpact

► Incorporateconstraints(e.g. business)

► Implement

Monitoring


High quality data remains the #1 priority for buildingaccurate models


In market behavior

Interaction

Lifestyle

Life stage

Life events

Geo-demographics

Company relations

External data

Marketing

Billing

CRM

Other sys

Internal data

Product

Modelstructure Variable parameters

Validation

customer behaviore,g. default or

claim history, billpayment on time,

existing loans

e.g. postal code,proximity to coastproximity to fire

dept. (insurance)

Credit ratings

e.g. age,education,

marital status

e.g. elasticity ofdemand

e.g. loan size,insurance lineof business,

other productspurchased

Pred

ictiv

eM

odel

Setu

p

Working on the model inputs should be a collaborativeeffort across the organization


►Create a common data platform, whileadhering to customer privacy and dataprotection guidelines►More complete information►Efficiencies

►Maintain a common data dictionaryacross the organization►Reduce risk of errors►Reduce data misuse

►Document the data sources and datamanipulation

In market behavior

Interaction

Lifestyle

Life stage

Life events

Geo-demographics

Company relations

External data

Real-timedata

Marketing

Billing

CRM

Other sys

Internal data

Product

Credit ratings

► Introduction




► Wrap up


1

2

3

4

5

Financial (short-term, long-term)Accounting, Reputation, Poor decision making

There are numerous sources of model risk and potentialadverse business consequences

Sources of model risk:► Inputs► Design► Use/implementation in silos:

► Inadequate knowledge of model purpose,processes and controls, e.g. key person risk, lackof training

► Errors in the end-to-end process, e.g.unauthorized and incorrect model changes

► Overreliance on models, e.g. limitations beingignored

► Old-generation models unreliable as a result ofchanges in market conditions

Possible adverse consequences:

...indicating the need for appropriate model risk management (MRM) and collaboration


► Ineffective marketing (limited up/cross-selling)

► Customer loss► Inadequate quantification of risk and

capital requirements► Incorrectly designed and priced products► Poor strategic decisions► Poor operations (planning, investment

decisions and resourcing)► Financial reporting errors and

restatements

Examples and consequences of model management in silos

► Marketing campaign attracts large number ofcustomers who are disqualified due to bad credit

► Marketing attracts customers based on low riskof default without regard to their profitability

► Newly developed customer analytics departmentstarts data aggregation process in new IT system

► Sales force and underwriters focus on high riskmarket segment to increase sales volume, butprice them incorrectly


Collaboration and alignment between modeling and customer analyticswould have reduced the risk of model misuse

► Increased costs and “bad will” as creditdepartments spend time on rejections

► Resources tied up on customers with little profitmargin potential

► New department would gain efficiency bystarting with modeling data from risk department

► Adverse selection for the insurer: the increase in“bad risks” in the book of business leads tolower profitability

Examples Consequences

Ban

king

Insu

ranc

e

► Introduction




► Wrap up


1

2

3

4

5

Three cornerstones of synergies between risk modeling andcustomer analytics


Understandyour models

Break downthe silos

Understandyour customers

Efficiency,profitability &

customer valuemaximization

Create synergies by effective risk managementof business process and model risks


Financialprocesses

Riskmanagement

processes

Operationalprocesses

Other modelinputs

Businessdecisions

MgmtreportingExternalreporting

Bus

ines

spr

oces

s

Adjustments Outputs

Internal dataExternal dataAssumptionsOther model

outputs

Transformand cleanse

inputs

Inputs Outputs

Modeloperation

Changemanagement

Implementation

Development

Businesspurpose

Validation

Risk management should focus on business process (e.g., resource pool, model results communication andimplementation) and model life cycle (e.g., maintain model inventory, results documentation)


Break downthe silos




Create synergies by better understandingand using your models

Consistentapproach tomanaging

models

► Adopt consistent development standards for new models and model changes acrossthe organization

► Use resources efficiently for model review and validation

Increaseawareness ofmodel usage

and materiality

► Create an enterprise-wide understanding of what models are used, where they areused and for what purpose

► Understand the range of model usage in the risk and customer analytics departments


Better decisionmaking

► Bring together the objectives of the risk and customer analytics departments tooptimize pricing efficiencies and marketing spend

► Improve management understanding of key models, assumptions and limitations


Break downthe silos





360° View

Customer

Customer-Centric AnalyticsCustomer-Centric Analytics► Segmentation► Campaigning

► Customer Acquisition► Customer Retention

► Cross-sell flags► Up-sell flags

In market behavior

Interaction

Lifestyle

Life stage

Life events

Geo-demographics

Company relations

External data

► Regression models► Proactive models

BenefitsBenefits

► Agile analytics► Accurate calculations

Real-timedata

Marketing

Billing

CRM

Other sys

Internal data

Product

► Increased customer revenue► Increased cross/up-selling► Reduced time to market

► Improving risk management► Improving customer satisfaction► Improve customer retention

► Focus efforts on profitablecustomers

► One version of the truth

► Compliance with legislation

Understand and leverage the data availableacross the organization


Break downthe silos




Credit ratings

Customer information is scattered across systemsand the organization - Break down the silos!► Align incentives with enterprise-wide value maximization rather than rewarding individual

business units for volume generated

”Siloed” organization

RiskModeling

Mgmt.(Strategy)

MarketingDepartment

Customer

Finance andOperations

A single view of the customer seeks to realize the financial benefitsof the offerings by tailoring them to the customer needs

Integrated business intelligenceand analytics vision

“Customer-oriented organization”

Agen

tcha

nnel

Inte

rnet

chan

nel

Mob

ilech

anne

l

Oth

erC

hann

els



Break downthe silos





► Introduction




► Wrap up

1

2

3

4

5

Create a model inventory andensure that it is used

One of the most important toolswhen defining and recalibrating

your strategy

Align incentives towardscustomer-valuemaximization

i.e analytics vision acrossproduct and line of

business

Secure a 360˚ viewof your customer

through shared datamanagement

Wrap up



Break downthe silos


Profitabilitythrough


Thank you!

EY | Assurance | Tax | Transactions | Advisory

About EY

EY is a global leader in assurance, tax, transaction and advisoryservices. The insights and quality services we deliver help build trustand confidence in the capital markets and in economies the worldover. We develop outstanding leaders who team to deliver on ourpromises to all of our stakeholders. In so doing, we play a critical rolein building a better working world for our people, for our clients andfor our communities.

EY refers to the global organization, and may refer to one or more, ofthe member firms of Ernst & Young Global Limited, each of which is aseparate legal entity. Ernst & Young Global Limited, a UK companylimited by guarantee, does not provide services to clients. For moreinformation about our organization, please visit ey.com.

© 2014 Ernst & Young AB. All Rights Reserved.

ED 01.2015

synergies between risk modeling and customer analytics · synergies between risk modeling and...

Documents