Predictive Analytics For Operational Decisions Dr. Rado Kotorov
The Red and Black Side of Forecasts
We all have the data, but some organizations take the analysis one step further.
Use Cases Use Cases
Dealer Services Corporation Used Car Dealer Financing Big, High-Risk Industry DSC, Société de financement pour concessionnaires de voitures d occasion Secteur d activité étendu, à haut risque 18,700 used car dealerships vs. 41,250 franchises Franchises: -- 50% of sales new cars -- Higher profit margins -- Additional service revenue -- Large well trained staff Used Car Dealers: -- Stand alone -- Slim profit margins -- Less trained staff 18 700 concessionnaires de voitures d occasion vs. 41 250 franchises Franchises: -- 50% des véhicules vendus sont des voitures neuves -- Des marges bénéficiaires plus importantes -- Une proposition de services assurant des revenus complémentaires -- Une équipe compétente, en nombre Concessionnaires de voitures d occasion -- Propriétaires indépendants -- Marges bénéficiaires réduites -- Un personnel moins bien formé
How it works: Dealer Floor Plan Financing Comment ça marche? Le financement des stocks des concessionnaires -- A specialized niche of asset based landing -- Lender get financing from banks -- Lender finance inventory by making an advance against the wholesale price of a specific piece of collateral (car) -- Individual loans are originated for each vehicle -- As each piece of collateral is sold by the dealer, the loan advance against that piece of collateral is repaid -- Une niche spécialisée dans le financement d actifs -- Le concessionnaire obtient le financement de DSC -- Le prêteur (DSC) finance le stock en procédant à une avance du prix du véhicule -- Des prêts sont émis pour chacun des véhicules -- Le prêt est immédiatement remboursé auprès de DSC lorsque le concessionnaire vend le véhicule
The Business Issue: Reduce Write-Offs Problématique: Réduire les impayés DSC Management Observations: -- Loan originators rely on personal relationships -- Subjective risk assessment -- Inconsistent performance Management by Objective: What gets measured, gets managed! Management by objective works - if you know the objectives. Ninety percent of the time you don't. Peter Drucker Challenge: How do you predict risk in order to set objectives? Les observations du Management de DSC: --Les initiateurs de prêts favorisent des relations de proximité -- L évaluation du risque est subjective -- Les performances sont inégales Management par objectif: Ce qui est mesuré, est encadré! Le Management par objectif fonctionne si vous connaissez ces objectifs. Dans 90% des cas, vous ne les connaissez pas. Peter Drucker Challenge: Comment prévoir le risque afin de définir les objectifs?
Explore/Explorer The Solution: RStat Predictive Modeling Solution: RStat Predictive Modeling Model/Modèle Deploy/Déployer Implement Dealer Loan Risk Scoring Applications: (1) Use WebFOCUS to access DSC dealer loan histories ERP data (2) Use RStat logistic regression to create a model and scoring routine for deployment (3) Use WebFOCUS to build web based scoring application for operational users and write scores to ERP (4) Use iway Data Quality Management to cleanse the data Mettre en œuvre des applications de notation pour évaluer les risques liés aux prêts octroyés aux concessionnaires : (1) Utiliser WebFOCUS pour accéder aux données ERP correspondantes à l historique des prêts octroyés par DSC aux concessionnaires (2) Utiliser RStat pour créer un modèle statistique et le déployer (3) Utiliser WebFOCUS pour diffuser une application statistique en mode web pour les utilisateurs opérationnels et écrire les scores dans l ERP (4) Utiliser iway Data Quality pour nettoyer les données
How It Works: Easy & Intuitive Comment ça marche? Simple & Intuitif - Initial 27% loan default rate - Taux initial de défaillance de prêt: 27% - Model differentiates between 10 segments - Best segment has 8% default rate - Worst segment has 52%% default rate - Le modèle segmente en 10 catégories les concessionnaires - Le meilleur segment a un taux de défaillance de 8% - La catégorie à risque maximal a un taux de défaillance de 52% - Model determines the contribution of each variable to final score - Users input application, model calculates score - Le modèle détermine la contribution de chaque variable au score final - Les utilisateurs entrent les données, le modèle calcule le score
Financial Impact from Deploying the RStat Model Les bénéfices financiers liés au déploiement du modèle RStat Avoiding bottom 10% customers could bring down the default rate by 5.1% (from 27.2% to 22.1%) Assuming $ 5,000 average write-off per default customer, avoiding bottom 20% customer could save $ 5.2 million Réduire de 10% la catégorie de clients la plus risquée permet de diminuer de 5,1% le taux de défaut de remboursement (de 27,2% à 22,1%) Réduire les créances individuelles de $ 5000 auprès des 20% de clients les plus risqués permet un gain de $ 5.2 million New Portfolio Default Rate Current Portfolio Default Rate 22.1% 18.0% 14.3% 11.1% 8.3% 27.2% 10% 20% 30% 40% 50% $ 2.9 MM $5.2 MM $ 7.2 MM $ 9.0 MM $ 10.5 MM Credit Loss Saved
Niche Retailing: Business Issues The Business Environment: Challenging environment Slow consumer spending Competition from the Web International competition The Problem: Loyalty due to location and convenience has decreased The Issue: How to leverage current loyalty program to increase loyalty and campaign responsiveness 65% of customers participate in the program
How It Works: Segmentation Segment customers based on shopping behavior Value, Recency, Preferences Top 3 deciles capture 80% of respondents Result: 20% incremental sales
Operationalizing it with ongoing monitoring
Retail Solutions Portfolio Strategy & Finance Sales planning and forecasting Performance Management Marketing Effectiveness Promotion Effectiveness Cross Channel marketing Merchandising Planning, Pricing, Assortment Size/Space, Planogram Shrinkage minimization Customer Management CRM and Loyalty Analytics Consumer Insights Customer satisfaction
Telecommunications: The Prepaid Subscriber The Industry: Prepaid phone subscribers is the new growth industry About 65% of new subscribers are prepaids About 22% of all mobile subscribers are prepaids. Highly competitive business driven by price conscious consumer: $45 vs $100 Customer Acquisition Incentives: Phones, free minutes, etc. Monetization Strategies : Retention Bite and switch to postpaid
Prepaid Subscribers: Only 45% Retention Rate The Paradox: Customers who do not accept promotional offers have already reduced their usage significantly and have made up their minds to churn out. Catch customers much in advance to increase the retention rates? Targeting customers in advance means more wasted offers on customers who were anyway going to renew the validity. The Question: How to target only real CHURN prospects?
Retention Solution For Prepaid Subscribers The Model: Use logistic regression to identify likely CHURNERS. The regression assigns probability to each subscriber The Data Analyzed: Usage Data: Call Volume, Call Duration, Call Frequency, Call Diversity / Density, Number Equity. SMS Behavior Usage of Value Added services Recharge Behavior Product / Plan / Package subscribed Promotional Campaign data Call center information
Results & Impact Monthly Campaign Metrics Without Analytics With Analytics Monthly Camapign Size (# of customers) 185,185 100,000 # of customers retained Avg. life of retained customer in months 19,493 26,810 7.05 10.2 Additional gross margin / month Telecalling Costs Cost of offer 5,789,474 7,962,570 1,500,000 900,000 3,055,556 2,543,973 Overall Cost Structure 4,555,556 3,443,973 Incremental Profits Generated by Analytics over a year 39,416,142
Telecom Solutions Portfolio Churn Management Managing Voluntary Churn Reactive Retention Programs Usage Enhancement Pricing Strategy Cross Sell / Up sell Programs Customer Lifetime Value Risk Management Managing Postpaid Collections Dunning Strategy Marketing Effectiveness Marketing Mix Modeling Campaign Management