Credit and Collections optimization and decision support An example of the intersection of decision tool development and management consulting Includes experiences from work with PA Consulting Group
Me Kevin Ross Assistant Professor, Technology and Information Management, UCSC From New Zealand PhD, Management Science & Engineering At UCSC Since 2004 Research areas: Scheduling, optimization, networks, pricing Worked with: Eli Lilly & Company Bell Labs NASA Thomson Reuters London Councils Fitness First PA Consulting Group Electrical Utility Companies
Service Management Air Traffic ControlCommunication Networks OptimizationMinimizing cost to deliver services Minimizing total delay in US Airspace Maximizing capacity of a communication network SchedulingAllocating people and machines to tasks Allocating arrival and departure slots Switching and routing rules for rapid scheduling PricingValuing resources in call centers Utilizing auctions to allocate arrivals Pricing models for online advertising Queueing Theory Trading off waiting time with service quality Minimizing air holding time Managing buffers and backlogs Kevin Ross: Network and Service Management Professor Ross develops analytical models for three application areas
Today Walk-through of a real consulting assignment with a decision tool focus Details limited due to confidentiality
Situation Big utility provider in a major US City Has a monopoly on business Is owned by taxpayers Heavily regulated “Too many of our customers don’t pay their bills on time. What should we do?”
Step 1: Understand the real problem What is the objective? Who cares? What can be changed? What cannot be changed? What data is available? How available is it? Who will use what I produce?
Step 2: Propose an approach to find a solution Much better than just proposing a solution! Nothing an outsider produces will be adopted unless an insider is excited and owning it Agree a plan of action Deliverables Timeline Obligations & expectations
Sample timeline for this project
Step 3: Make it happen
Understanding the dynamics… Customer up to date with payments Phase 1 Customer behind on payments Phase 2 Customer received final bill Phase 3 Charged off account Regular bill cycle Deposits Deposit Review Proactive telephone calls Disconnect notices Field notifications Disconnection Final bill issued Final bill collections agency Charge off collections agency Final recoveries Due date missed Power is cut Account Charged off Receivable is increasing Recovery rates are decreasing This project focused on understanding customer payment activity and collection activity timing in order to understand the implications of changes to the collection cycle
Designing the optimization problem Tool functionality objectives Optimized collection timeline – “when” Optimized cut strategy – “how many” Prioritized cut list – “what accounts” Scenario testing Total expected cost Expected charge off Expected operational cost Minimum total Optimal days until next action x x Cost Time Optimization is finding the most cost effective collection strategies for all collection actions and for all customers
Interjection “Wow that’s cool. If you can do that then can you also add these features…”
Decision support system Right actions at the right time based on consistent data with accurate reporting and ease of execution. Customer Analytics Dashboard Forecasting Timeline Optimization Scenario testing Prioritized cut list Data mining (eg, ROI, etc) Optimization Tool Optimization is finding the most cost effective collection strategies for all collection actions and for all customers
Final tool delivered… DataPurposeModules Customer input files Information on the customers in collections Current Customer List Historic Collections Activity Forecast Entries Parameters Describe the basic activities, segments and rates used in the model Collection Actions Risk Segments Payment Probabilities Transfer Rates Global Inputs Options User selected information for different scenarios, timelines and cut limits Cut limits Timelines Scenarios Outputs Summary Dashboard Excel Reports Comparison Dashboard Timeline Optimization Prioritized Cut List
Step 4: Handover Someone has to own a decision tool This is different from the designer
Lessons Lots of opportunities for people with the right skills in data mining, analysis and optimization The people who use the tools we develop will not usually be experts in these areas Unless someone owns a tool, it will never get used