Modeling the U.S. Postal Network December 8, 2010 Princeton, NJ Agenda ● ● ● ● ● ● Introduction The Postal Industry The U.S. Postal Service & its Network Infrastructure Postal Distribution Concepts Modeling the USPS Network Q&A decision/analysis partners ● Founded 11 years ago ● Three practices Logistics & supply chain Information & communications technologies Postal service ● Technical & management consulting Operations management Postal & Mailing Industry Mailers Service Providers Strategy Business Plan Other Operators Services Postal Sector Policy Laws Regulations Postal Enterprise Network Delivery Postal Regulator Operations The Mail ● Mail vs Parcels -- Letters vs Flats ● C2C: Personal Communications First class mail Subject to electronic diversion ● B2C: Transaction & Advertisement First & “standard” mail Standard: 0.5%$ growth over next 10 years ● B2B: Transactions First & Express Eroded ● C2B: Bill payment & reverse logistics Bill presentment in mail but payment online. United States Postal Service ● 500 million pieces of mail daily UPS: 15M Worldwide FedEx: 2.6M ● $68 billion Revenue First-Class Mail $36 b Ad $17 b Others $15B ● Large infrastructure 269 Processing and distribution plants 218,684 vehicles ● Evolving Network Volume subject to the economy and electronic diversion 923,595 new delivery points added to the network in 2009 43.8 million address changes processed in 2009 USPS Postal Products MARKET DOMINANT PRODUCTS ● First-Class Mail: Single-Piece Letters, Cards Presort Letters, Cards Flats Parcels ● Standard Mail: High Density and Saturation Letters, Flats & Parcels Carrier Route, Letters, Flats Not Flat-Machinables and Parcels ● Periodicals Mail ● Package Services Mail COMPETITIVE PRODUCTS ● ● ● ● Express Mail Priority Mail Parcel Select Mail International Mail: Expedited/Priority Air Parcel Post How the mail moves (simplified version) Delivery point sequence in Trenton 085 to Princeton DDU To delivery carrier 1 Mail collected Sacramento CA 95818 To Post Office 95815 7 5 PHL 2 SFO 4 6 MEM 3 Mail sorted at 958 to 94Z then to SFO Primary Sort Delivery Sort Mailers How the mail moves Platform Operations Mail Processing Advance Facer Canceller System (AFCS) Letters Mail Prep Delivery Bar Code Sorter (DBCS) DBCS Input/Output Subsystem (DIOSS) Flats Automated Flat Sorting Machine (AFSM 100) Flat Sequencing System (FSS) Small Parcel / Roll (SPRS) Small Parcel & Bundle Sorter (SPBS) Packages Automated Package Processing System (APPS) BMEU Hierarchy of Facilities ● National Distribution Centers (21) Bulk Mail and Parcels Processing Two-tier Regional Distribution Secondary Sort ● Processing & Distribution Centers (About 300) Facing & cancelling mail Outbound primary sort Destination Delivery Sequencing ● Surface Transfer Center (STC) Tray & Container Cross-docking Lettermail Plant Parcel Sorter Flow Analysis (Montreal) Material Handling Simulation (Montreal) Network & Distribution Concepts Why do we need a network? ● Transporting each mail envelope from its origin to its destination with its own driver would be prohibitively expensive, so… ● We stage the mail and we bundle the mail for transport and delivery We collect the mail and stage it to process it and deliver it once a day We sort the mail in order to bundle it for transport and delivery Bundling & staging are critical distribution processes Mode of Transport Bundling/Staging Freight Rail blocks/unit trains staged/switched in flat or hump yards LTL Trucking Pallets, pallet positions staged in warehouses Container Shipping Stacks in ships staged in container ports. The process of staging and sorting is common to transportation and distribution. Flat objects such as envelopes provide a significant economic opportunity to bundle. Postal processing represents about 82% of operating costs – transport about 12%. The Network The Network Network Layer Elements Time Horizon Real Estate Plants, Facilities 20 to 30 years Equipment, Fleet Sorting & material handling systems, trucks, planes 5 to 20 years People, Skills Clerks, drivers, planners, managers, trainers. 2 to 20 years Sort Plans & Schedules Sort & operating plans, transportation & other schedules. Real time to 1 year Measurement, Mgt & Planning Evaluation, planning & forecasting. Management Real time to 3 months What controls the flow of mail? ● In the short run, mail flows are under the control of Sort plans: bundling the mail into trays for further processing Operating plans: staging the mail for processing, transport or delivery, and Schedules: Transport (trucks, rail, planes), people, facilities, etc. Network Topology ● Layout pattern of interconnection of the elements of the network ● Topologies and hierarchy of nodes imply a distribution strategy: Star: All P&DC connected to one central mega-plant Full mesh: Each P&DC connected to each P&DC Tree: Each P&DC is connected to a regional center What impacts the network’s shape? Alternative A Alternative B City 1 City 2 City 1 City 2 Vol V1 Vol V2 Vol V1 Vol V2 City 3 City 4 City 3 City 4 Vol V3 Vol V4 Vol V3 Vol V4 one day one day What impacts the network’s shape? Alternative Processing Transport A ● Several plants ● Smaller facilities ● Lower productivity per plant ● Additional handling for some mail ● Higher total processing cost ● Shorter trips ● Lower transport volumes ● Lower circuity (perhaps) B ● One larger plant ● Higher productivity in the plant (perhaps) ● Longer processing times (perhaps) ● Longer trips ● Higher volumes ● More transportation Network Modifiers DPS Letter and Cased Volume History (City Delivery) 100.0 Cased Letters DPS Letters 90.0 80.0 70.0 • FY 2007 USPS Sequenced 130 Billion Letters Percent 60.0 • Resulting in Over $5 Billion Annual Savings 50.0 40.0 30.0 20.0 10.0 0.0 1993 1994 1995 1996 1997 1998 1999 2000 2001 Fiscal Year 2002 2003 2004 2005 2006 Cased Letter % DPS Letter % 2007 2008 YTD Plant Productivity Hours per Piece Handled as a function of Total Piece Handled per Plant per Year (M) U-SHAPED PREDICTED WH / TPH AS A FUNCTION OF TPH 0.0200 0.0180 0.0160 WH / TPH 0.0140 0.0120 0.0100 0.0080 0.0060 0.0040 50 100 150 TPH 200 250 300 Network Modifiers Processing Productivity Improvements Barcoding APPS MLOCR AFCS RBCS DBCS AFSM ATHS IDR DIOSS-EC PARS FSS RCS Other Changes ● Volume changes Volume/electr. diversion/recession Mix ● Population changes New addresses ● Mailer network induction sites ● Transport Aviation economics Air security & handling Fuel costs ● Increase in volatility ● International volumes What impacts the network’s shape? ● The network is constrained by time One-day service areas constrain operating plans and facility locations Three-day and some two-day areas require the use of air transport ● Larger processing plants are more efficient up to a point Mail can be sorted more continuously Efficiency per letter handled plateaus at a certain volume Mail can be sourced from longer distances Modeling the Postal Netowrk Objective ● Support USPS Office of Inspector General Engage in public discourse about USPS network infrastructure Educate public policy debate and the policy establishment Provide benchmarks for public policy ● Introduce USPS management to new concepts “Appreciate” the impact of initiatives on network topology Plan for the future at all five layers Approach ● A “complex” problem Previous efforts at closed modeling failed Too many details – not enough details. ● Simulation approach Stay away from complex mathematical constructs Emulate & evaluate distribution approaches & constraints ● Technology Repast (Recursive Porus Agent Simulation Toolkit) Symphony: Advanced, free, and open source agent-based modeling and simulation platform @ sourceforge.net Douglas Samuelson and Charles Macal, "Agent-based Simulation Comes of Age," OR/MS Today, Vol. 33, Number 4, pp. 34-38, Lionheart Publishing, Marietta, GA, USA (August 2006). http://www.lionhrtpub.com/orms/orms-8-06/fragent.html Simulation ● Agent Class: Facility Subclasses: Consolidator and Non-consolidator ● An initial number of facilities are seeded ● Mail is directed from zips to facilities. Operations in each plants are simulated based on mail input and output Cost are tallied: processing & transport Mail performance is tallied Abstra ction Simulation Agent Class Hierarchy Abstract Agent Processing Agent Standard Plant Consolidatio n Plant (Hub) Transport Agent Truck (Surface) Plane (Air) Model Structure Local ZIP Codes Other Processing Agents Route New d Mail Mail Incoming Mail Mail To Deliver Surface Transport Agent Routed Mail Processing Agent Surface / Air Transport Agents Routing Rules Processing Cost Functions Transport Cost Transport Cost Total Cost Processing & Transport Cost Parameters ● Processing Costs Workhours (labor) computer based on plant productivity statistics Workhours include Primary Outgoing, Secondary Outgoing, or Incoming sorts Using average labor cost per hour ● Transportation Costs Ground: $0.009 per cubic foot mile Air: $0.0006-8 per lb per mile flown depending on type of mail ● Statistical conversion factors are used 33 Distribution Strategies ● Point-to-Point Mail is sent directly from origin plant to destination plant No intermediate stops, By truck if <500 miles, by air if >500miles. ● Peer-to-Peer Mail takes the shortest-path route between origin and destination (Dijkstra's algorithm) No surface transportation leg exceeds 500 miles. ● Hub-and-Spoke Consolidation All mail is routed through Consolidation hubs origin to destination Except mail with the same O-D, which is just delivered locally). ● Hybrid Consolidation O-Ds that are <500mi apart exchange mail directly Mail between all other O-D pairs (distance >500mi) is routed through Consolidation hubs. Facility Placement Using 90mi Service Radius Service Radius: 90mi Facility Count: 170 (includes 16 Consolidators) 35 Facility Placement Approach #1: Maximum Service Radius, Example Computed facility placements based on 150mi max service radius (for illustration): Max Service Radius: 150mi Facility Count: 82 (includes 15 Consolidators) 36 Distribution Strategies ● Peer-to-Peer Strategy Facilities attempt to send mail directly to the destination. Transportation inefficient - many trucks with small loads. ● Consolidation Strategy Select facilities are used as Consolidation points in the network Mail is routed through these hubs to achieve processing economies of scale Longer transportation distances, but trucks will be more full 37 Modeled Mail Types Modeled Mail Types Express Priority 1st Class Standard Periodicals Package Letters Flats Parcels Non-Presorted Presorted 38 Sample Model Results Strategy Radius # Facilities P2P 90 Miles Consolidation 200 Miles 90 Miles 200 miles 170 57 170 57 248K 277K 212K 211K Ground Transport $13.3M $22.8M $21.0M $30.4M Late Mail 0% 10% 3% 13% Work Hours Transforming the Postal Network Preparing the Postal Network for the Future The World will be increasingly… ● Digital: Use of data -- Mixed media ● Volatile: More rapid changes in volumes, mix, O/Ds ● Uncertain: Less predictable volumes Harder to forecast ● Complex: Non-linear, difficult causal relationships More difficult decision making ● Ambiguous: Lack of clarity Role of the Model ● Use the model as a didactic tool Model can never be complex enough to be realistic Use the model to develop cooperation and consensus ● Incorporate the model lessons in a larger context The five layers Preparing the Postal Network for the Future Real Estate ● Re-map/increase plant capture areas ● Create/consolidate megaplants outside large metropolitan areas ● Increase footprint flexibility: Consider leasing/renting space for processing in low volume areas ● Develop R/E strategy to deal with transportation congestion and fuel costs Equipment ● Design plants with standard work centers ● Use multi-purpose MLOCR equipment for increased flexibility ● Leverage intelligent mail technology to improve productivity ● Create on-demand transportation contracts and fleet (power by the hour) Preparing the Postal Network for the Future People ● Match manpower to needs (increased use of part-time or flex-time labor) ● More decision-oriented culture - Cell-production systems ● Improve how people communicate, seek innovation, and address conflict ● Promote network-friendly cooperation among managers Sort Plans & Schedules ● Implement dynamic network management system ● Enable operating plans – perhaps sort plans – to be adapted in the near/real time. To take advantage of excess capacity Improve performance Reduce costs Preparing the Postal Network for the Future Measurement, Planning & Mgt ● Develop business analytics Simulation Forecasting Predictive Modeling ● Institutionalize network control Network cooperation and collaboration ● Develop dynamic network management capabilities Ability to react ● Questions?