Accenture - Customer Sensitivity to Credit Risk Decisions

44 %
56 %
Information about Accenture - Customer Sensitivity to Credit Risk Decisions
Business & Mgmt

Published on March 4, 2014

Author: digitalinsurance

Source: slideshare.net

Customer Sensitivity to Credit Risk Decisions Matthew O’Kane – Senior Manager Accenture Analytics August 2013

Big Data What is an Individual Decision Effect? Use Function Churn Target only people who respond favourably to the retention activity Marketing Target persuadable targets only Risk decisioning Offer the appropriate credit to maximise RoE NBA Comparing multiple options at the customer level to understand customer decision sensitivity Pricing Maximise uptake by offering the right price according to individual price sensitivity Collections Target only people who would have only paid because of this collections activity Digital Attribution Display media only to prospects where the media has a positive influence on conversion Operations Choose the right level of service on a customer by customer basis Alternative A Expected individual effect for a single customer and single decision between multiple alternatives Alternative B Goes by multiple disguises: Uplift / Net lift / Causal effect / Action effect / Maximum differential / Incremental modelling….

Big Data But didn’t we already solve that? “Cosmologists studying a map of the universe from data gathered by the Planck spacecraft have concluded that it shows anomalies that can only have been caused by the gravitational pull of other universes”

Big Data Why predict customer sensitivity? Key issues today What needs to be addressed 1 Competing business objectives (revenue, default, profit, cost to serve, return on equity, capital requirements etc.) Methods that can predict changes in business values due to customer-level decisions 2 Competing customer treatments (limit, authorisations, cross-sell, up-sell, pricing, precollections etc.) Methods that can estimate the Individual Decision Effect of competing treatments 3 Competing methodologies (A/B testing, Champion A method of comparing multiple methodologies / Challenger, online learning, machine learning)

Big Data Why predict customer sensitivity? Method A/B Testing Basic Premise Randomised experiment to estimate the Average Treatment Effect (ATE) More information Kohavi, Ron, et al. "Trustworthy online controlled experiments: Five puzzling outcomes explained." Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2012. Uplift Modelling Maximising the uplift between two or more treatments Rzepakowski, Piotr, and Szymon Jaroszewicz. "Decision trees for uplift modeling." Data Mining (ICDM), 2010 IEEE 10th International Conference on. IEEE, 2010.” Guelman, Leo, Montserrat Guillén, and Ana M. Pérez-Marín. "Random Forests for Uplift Modeling: An Insurance Customer Retention Case." Modeling and Simulation in Engineering, Economics and Management (2012): 123-133.” Heterogeneous treatment effects Segmenting randomised experiments to estimate the Average Treatment Effect (ATE) within segments Grimmer, Justin, Solomon Messing, and Sean J. Westwood. "Estimating Heterogeneous Treatment Effects and the Effects of Heterogeneous Treatments with Ensemble Methods." (2013). Incremental Response Modelling Transforming binary response and treatment variables to measure the incremental response rate Pechyony, Dmitry, Rosie Jones, and Xiaojing Li. "A joint optimization of incrementality and revenue to satisfy both advertiser and publisher." Proceedings of the 22nd international conference on World Wide Web companion. International World Wide Web Conferences Steering Committee, 2013. Model-based Recursive Partitioning Tree-based method where each split looks for structural changes in the treatment / response model Kopf, Julia, Thomas Augustin, and Carolin Strobl. "The potential of model-based recursive partitioning in the social sciencesRevisiting Ockham's Razor." (2010).

Big Data A model for assessing customer sensitivity effectiveness 0.6 • Estimated delta between treatment A and treatment B is calculated • Sorted by descending delta • Actual delta calculated within bins 0.4 0.2 0 -0.2 -0.4 -0.6 Sensitivity  1 n Y i n i 1 Y  i percentile, actual th i sensitivity, by descending sensitivity ˆ Y  i percentile, expected th n 1 ˆ  (Y i  Y i) n i 1 i sensitivity, by descending sensitivity

Customer sensitivity effectiveness resulting under different models - QUIZ Which graph corresponds to a typical result from: • Champion / challenger • Uplift model • Matrix segmentation (action effect) • Differential response model • A/B testing 0.6 0.4 0.2 0 0.6 0.4 0.2 0 -0.2 -0.2 -0.4 -0.6 0.6 -0.4 0.4 -0.6 0.2 Expected 0 Actual -0.2 -0.4 0.6 0.6 -0.6 0.4 0.4 0.2 0.2 0 -0.2 0 -0.2 -0.4 -0.4 -0.6 -0.6

How to scale to Big Data? Value Drivers Decision Levers Pricing Risk decisions Churn management Increase profit per product Campaigns Increase Revenue Increase CLV Increase usage Collections Digital Attribution Decrease risk NBA Operations Decrease Cost to Serve Recover arrears Decrease cost to serve

Big Data A big data solution to address scale and complexity issues Key issues today Observations 1 Competing business objectives (revenue, default, profit, cost to serve, return on equity, capital requirements etc.) Regression rather than classification required due to the requirement to estimate economic impact 2 Competing customer treatments (limit, authorisations, cross-sell, up-sell, pricing, precollections etc.) Methodology needs to scale to multiple treatments, preferably without repeated modelling which increases variance 3 Competing methodologies (A/B testing, Champion Ensemble-based solutions provide the / Challenger, online learning, machine learning) necessary predictive power and scalability

MIT Accenture Research Alliance Accenture and MIT’s Operations Research Center will jointly develop solutions that can be applied to realworld, client specific issues through the development of new business analytics that bring together data, modeling and analysis to achieve a sustained quantum leap in business performance. This Alliance will:  Be the premier program for developing and applying analytics solutions to solve complex and challenging global business problems.  Develop new business analytics and promote the adoption of advanced analytics in all relevant fields.  Support some of MIT’s brightest graduate students on research programs to address strategic business problems through innovative application of analytics.  Jointly develop new tools and techniques focused on leading-edge analytics.  Establish a consortium to provide a platform for invited companies to engage in dialogue on business analytics and impacts.

Big Data Accenture Analytics – a global leader in enterprise-level analytics Accenture Analytics Capabilities Top Talent 13,000 resources providing: Customer, Risk, Claims, Payments, Fraud Detection and Enterprise Analytics Thought Leadership & IP Over 200 patented assets And we are hiring! Global CoEs 15 Global Innovation Centers Innovation consortium 9 Key Alliances & Acquisitions

Add a comment

Related presentations

Related pages

Customer Sensitivity to Credit Risk Decisions

Customer Sensitivity to Credit Risk Decisions ... understand customer decision sensitivity ... Customer, Risk,
Read more

Accenture | Strategy, Consulting, Digital, Technology and ...

Accenture is a leading global professional services company, providing a broad range of services and solutions in strategy, consulting, digital, ...
Read more

Principles for the Management of Credit Risk

Principles for the Management of Credit Risk ... analysis and support credit decisions with simple ... liquidity risk management by the customer or a ...
Read more

Accenture UK | Strategy, Consulting, Digital, Technology ...

Accenture is a leading global professional services company, providing a broad range of services and solutions in strategy, consulting, digital, technology ...
Read more

Sharing The Risks And Rewards Of Outsourcing - InformationWeek

Sharing The Risks And Rewards Of ... behind outsourcing decisions. Ideally, a customer gets good ... Customer will hold Accenture to service ...
Read more

Regulatory Alert | Accenture Finance & Risk Blog

Credit Risk; The PRA has adopted ... Posted in Regulatory Alert Tagged Customer Identification ... The risk charges from sensitivities based method are ...
Read more

Finance and Risk Management | Accenture Finance & Risk ...

Meet Finance and Risk Management ... Accenture disclaims, ... Risks Covered: Credit Risk, Market Risk, ...
Read more

Pricing Intelligence & Insight

... price response curve and regression techniques to conduct pricing sensitivity ... helping businesses take more informed decisions; ... Customer ...
Read more

Accenture, Avanade and Microsoft collaborate on new ...

Accenture, Avanade and Microsoft collaborate on ... more than 30% of banking revenues could be at risk ... “The Accenture Customer Analytic ...
Read more

Newsroom | Accenture Newsroom

Newsroom. Print ... the future of digital customer interactions, Accenture Interactive demonstrated a new ... more informed patient decisions, ...
Read more