Credit Officer as a Service The Supply Chain Intelligence Company - - PowerPoint PPT Presentation

The Supply Chain Intelligence Company

Credit Officer as a Service

Bill Panak – VP, Data Science, Halo Advisor, LendIt / NSR Invest

Fun Fact and Key Questions

QUANTS! – and in particular, quants who understand the business What risks does this create? How do you, specifically you, address this issue? Where does this rank in your priorities?

At a recent industry conference, when about 30 small business credit risk management executives were asked “what is your toughest personnel management / talent retention issue…?”, the unanimous opinion was:

The Paradox of Big Data and Business Opportunity True in many industry verticals, particularly true in FinTech Situation Complicatio n Solution Data, analytics, computational speed, FinTech revolution all converge to create broad and diverse business opportunitites There is a huge talent / expert knowledge gap, the risks are high, and the regulatory horizon is both unclear and a future cost Build a robust, defined workflow, a hardened asset for doing the key, complex analytical tasks, and lease that into the FinTech ecosystem, driving down complexity management and costs

A defined workflow that ensure routine credit risk analytics is efficient and accurate

• You can customize analysis, building on a trusted foundation

Built on technology that scales, with rigorous validation and ongoing development

• Repurposing existing code and configuration, reducing development time / costs

Enabling one analyst to do the work of 2 or 3

• Less time coding and correcting data, more time for knowledge discovery

Credit Officer as a Service (COaaS) Value Proposition

• Data Ingestation
• Data Quality Assurance (QA) – a single source of truth
• Basic Reporting and Drilldown from a managed data warehouse
• Efficient, defined workflow to build and calibrate a rank-ordering risk model
• Efficient, defined workflow to build and calibrate account level valuations
• Sound actuarial mathematics, programmed, validated, locked down
• Loan product features and financial assumptions build into the valuations system
• Facilitates gaming and optimization, monitoring and adjustment
• Documentation flows into defined locations, following a template, audit-ready

COaaS – What does the solution look like? Basic and Advanced Business Requirements – what would a CO want?

Turns out, 80% of the workflow is already built, at Halo

450 customers worldwide
 10,200 active users Data, Analytics, Compliance, Consulting

San Diego Vienna Auckland

Stats Offices Supply Chain Analytics at Scale

COaaS Schematic – You are invited to add and connect dots

Data Warehouse QA Metrics Machine Learning SAS Microsoft Azure R IBM Salford Systems Rank- Ordering Risk Model Event Hazard Curves Cash Flow Equations Hard-Coded KPI’s Actuarially Sound Math Production Scoring Big Data Forecasting Please send feedback to: bill.panak@halobi.com

Invest in your platform, then products, then your analytics

Invest in UX, the foundation of positive selection Optimize on take-up rates and net promoter index; product + operational execution Pricing tests will teach you where a price drop creates additional positive selection At that point, analytics becomes stable, but you want to get to that point with a plan and a foundation Growth is not always life

In the long game, Positive Selection trumps Advanced Analytics

Look for simple, elegant code

** Fit logistic regression and save scoring wizard input -- note that term was removed from equation because results favored 60 month loans we do not believe this to be valid. LOGISTIC REGRESSION VARIABLES co_event /METHOD=FSTEP(COND) p_co_18m dti annual_inc_000s purpose_n fico emp_length_n t_csb_2 t_csb_8 t_csb_15 t_csb_20 t_csb_35 t t_sq /CRITERIA=PIN(.05) POUT(.10) ITERATE(20) CUT(.5) /OUTFILE= MODEL ('C:\Temp\LC_DATA\XML\cocurve_model_pre2011_12312015.xml'). ** Run scoring wizard. MODEL HANDLE NAME=cocurve_model_pre2011 FILE='C:\Temp\LC_DATA\XML\cocurve_model_pre2011_12312015.xml' /OPTIONS MISSING=SUBSTITUTE. COMPUTE p_co_pre2011=APPLYMODEL(cocurve_model_pre2011, 'PROBABILITY', 1). EXECUTE. MODEL CLOSE NAME=cocurve_model_pre2011.

Code Example: Default event hazard model and scoring on full Lending Club archive

Backtest your method, automate monitoring

Risk and Prepayment Curves, Lending Club, 2007 to 2012 Originatations

It is not done until the cashflow calculator is in production…

Excel based calculators are a starting point, but the destination is hard coded KPI’s Thought Experiment:

• Think about what makes for a

useful cashflow calculator

• Describe those business

requirements verbally to a programmer analyst

• See what they build for you

Hire an Actuary and get it right the first time

Actuarial Math, Embedded, Validated, Audit-Ready First Gen Calculator

• Discounting
• NPV / IRR
• Account Level Prediction
• Scoring Equation Management

Second Gen Calculator

• Programmed in R
• Validated by Sr. Actuary
• Embedded in Halo platform

COaaS Roadmap

Product Design Funding Proof of Concept R Implementation Full Halo Integration Q1 2016 Q2 2016 Q3 2016 Q4 2016 2017

What Success Looks Like

“When you have a chance to review, there is a LOT of content built for the POC…There is SO MUCH available in the data set, though, that my ADD is in tornado mode.”

What Success Looks Like

“I have one like that too. Hopefully you guys can build whatever you need with what I’ve built into the cube”

Mobile Access, Scheduled Alerts, Daily Monitoring

• Distribute reports on a

scheduled basis

• Get notified when a

condition has been triggered

• Take immediate action
• Easy export to Excel

and PDF

What you can expect to see from Halo

Exceptional data management capabilities Solutions that work within your architecture, configurations that enable data innovation World class predictive modeling, actuarial, and data science capabilities Workflows that enable analysts to focus on knowledge discovery Proof of Concept engagements that help you decide on your future

We manage data, automate, drive knowledge discovery, and reduce operating costs

Let’s Keep Talking

bill.panak@halobi.com

Summary

• FinTech is advancing quickly – but there is risk in acquiring and

retaining skills, regulatory shifts, and technology needs

• COaaS is a positive, proactive approach to credit risk management
• COaaS benefits:
• Reduces the risk of Quant dependence
• Reduces costs associated with Advanced Analytics
• Delivers automation and documentation to ensure solid reporting
• n demand