Newgistics Case Study 04 19 2006 Lei Deng Timothy Gumm Jacqueline Jones Adam Levin Leihong Li Yang Song 1 Agenda Problem Overview Current vs Proposed Situation Pre Modeling Issues The Model Results Analysis 2 Problem Statement and Goal Problem Statement The postal network accumulates and consolidates the return volumes at Return Delivery Units RDU and Return Bulk Mail Centers RBMC At some point it becomes faster and less expensive to pull the consolidated volumes out of the postal stream and ship it directly to Newgistics returns centers Goal Develop a model that Newgistics can use under changing conditions to understand where to best extract the returns from the postal network 10 20 and 30 million volume cases 3 Pull from BMC Current State RDU i 70 Cost by postal zone Post Office 10 20 or 30 Million Packages Legend i 1 to 6115 j 1 to 22 k ATL CHI LAX DFW PHY Return Centers k 30 RBMC j Cost FTL cost Constraints maximum density minimum pick up frequency 4 Pull from RDU or BMC Possible Future State Cost 2 parcel DDU cost Constraints Pick up at least 3 times week RDU i 75 70 Cost by postal zone Post Office 10 20 or 30 Million Packages Legend i 1 to 6115 j 1 to 22 k ATL CHI LAX DFW PHY items bag 30 Return Centers k 25 RBMC j Cost by FTL cost Constraints truck capacity minimum pick up frequency 5 Issues Before Modeling SUM RDU dist 70 100 where does the other 30 go The 30 go directly to BMC from postal office outside of our control So the uncontrollable volume at a single BMC is 30 annual volume BMC dist Any problems 30 fixed volume does not apply to an individual BMC The right answer is BMCi dist Sum RDUs in BMCi dist annual volume 6 Issues Before Modeling Truckload What if 3 TL week do we still pay for 3 FTL cost Package and bag weight distribution One bag can contain 1 2 3 4 or 5 packages We are given weight distribution for packages 10 lbs but no distribution for packages 10 lbs cost What distribution to assume Exponential or Uniform What is the bag weight distribution Simulation needed 7 DDU Bag Weight Simulation the distribtuion of bag weight 5 packages per bag 0 5 weight distribtuion weight distribtuion the distribtuion of bag weight 5 packages per bag 0 4 0 3 0 2 0 1 0 0 5 10 15 20 weight lbs 25 30 0 2 0 35 simulation of bag weight 5 packages per bag 10000 0 4 0 5 10000 10 15 20 25 weight lbs simulation of bag weight 5 packages per bag 30 35 6000 No case No case 8000 4000 5000 2000 0 0 20 40 60 80 weight lbs 100 120 140 0 0 5 10 15 20 25 30 weight lbs 35 40 45 50 Not realistic Exponential is better estimate 8 DDU Bag Weight Optimization Understand optimal bag configuration to determine the financial value in understanding cubic dimensions of packages Developed optimization model Results indicate that significant savings exist if bag configurations are optimized Suggestion Work with vendor to improve pricing contract for DDU Pickup Cost based on Bag Opt Conclusions 9 The Model Uses Excel and GAMS General Algebraic Modeling System Computed costs in Excel Formulated problem as an Integer Program in GAMS Excel Model Transfer costs to GAMS Model Transfer solution to Excel Model 10 Costs of Pulling from RDU For 75 captured packages DDU cost for each RDU Expected DDU cost bag with i packages i 1 2 3 4 5 Cost i sum cost per pound weight P weight Total cost for each RDU M cost 5 cost N 2 5M N M full bags rounddown packages at RDU 5 N packages in last bag packages at RDU 5 M BagWeight 1 2 3 4 5 6 7 8 9 10 11 12 13 Cost per lbs 3 1 5 1 0 75 0 65 0 55 0 5 0 45 0 4 0 38 0 35 0 34 0 32 11 Costs of Pulling from RDU 25 Non captured Packages Expected postal cost for each RDU Expected postal cost by zone noncaptured pkgs at each RDU Expected postal cost for each zone sum over all weights weight distribution the postal cost for each weight Non captured pkgs at each RDU ROUND RDU distribution total annual volume 0 25 52 3 Weight 1 2 3 4 5 6 7 8 9 10 10 Distribution 50 20 10 5 4 3 2 1 1 1 3 12 Costs of Pulling from BMC Expected Expected postal cost by zone pkgs at each RDU Truckload postal cost for each RDU cost Includes pkgs that go from post office BMC and those that go from RDU BMC Max 3 total weekly volume at BMC max density of truck FTL cost 52 13 Model Formulation n n n i 1 i 1 i 1 Min Weekly Total Cost xi C R 2 Bi 1 xi C R 2 BN i xi C R 2 RCi truckload Subject to n n i 1 i 1 P2 BMC 1 xi PR 2 BN i xi PR 2 Bi totPackBMC totPackBMC numt MaxDensity numt 3 truckload FTLCost numt xi binary eq totalBMC eq numtruck eq numsent eq FTL 14 Results Total RDUs pulled 10 million 5276 or 86 3 20 million 5574 or 91 2 30 million 5625 or 92 Savings from proposed solution 10 million 8 294 499 20 million 17 800 244 30 million 29 618 935 15 DDU Cost Sensitivity Analysis Detroit and Springfield DETROIT SPRINGFIELD Scenario RDUs Pulled Fraction of Total RDUs Pulled Fraction of Total 10 million 1 Cost 1 2 Cost 1 5 Cost 2 Cost 206 206 206 187 93 63 93 63 93 63 85 00 319 319 270 245 82 64 82 64 69 95 63 47 20 million 1 Cost 1 2 Cost 1 5 Cost 2 Cost 216 216 216 198 98 18 98 18 98 18 90 00 326 326 326 277 84 46 84 46 84 46 71 76 30 million 1 Cost 1 2 Cost 1 5 Cost 2 Cost 216 216 216 200 98 18 98 18 98 18 90 91 331 331 331 303 85 75 85 75 85 75 78 50 16 Postal Cost Sensitivity Analysis Detroit Scenario RDUs Pulled Fraction of Total 10 million 1 Cost 0 9 Cost 0 8 Cost 0 7 Cost 0 6 Cost 206 206 206 183 36 93 64 93 64 93 64 83 18 16 36 20 million 1 Cost 0 9 Cost 0 8 Cost 0 7 Cost 0 6 Cost 216 216 216 189 39 98 18 98 18 98 18 85 91 17 73 30 million 1 Cost 0 9 Cost 0 8 Cost 0 7 Cost 0 6 Cost 216 216 216 190 73 98 18 98 18 98 18 86 36 33 18 17 Overall Results Pull from RDU except when volumes are very low Very low volumes increase DDU cost cheaper to go through BMC As annual …