identifying inventory excess and service risk in...
Post on 10-Aug-2020
1 Views
Preview:
TRANSCRIPT
IdentifyingInventoryExcessAndServiceRiskInMedicalDevices:
ASimulationApproach
AlbertXu MariaRey
Agenda• Motivation&Background
• GoalStatements
• Methodology• DataAnalysis
• Simulationmodel
• Results
• Insights
• Conclusion
Motivation• Distributioncenters• Highinventoryholdingcosts
• Medicaldevices• Highvalue• Non-interchangeable• Criticality
Background• MedCorecentlycollectedlargeamountsoftransactionaldata• Inventory• Demand• Supply• Forecastaccuracy• Location
• Needtoreduceinventorywithoutaffectingservice• Currentlyuseclassical(s,S)inventorymodel,assumingnormality
GoalStatements•Determineinventorylevelforamaterialnumber,thatensuresaLIFRof98%
•Gaininsightsfromthedata• Understandwhatdrivesserviceperformance
𝐿𝐼𝐹𝑅 = 𝑁𝑢𝑚𝑏𝑒𝑟𝑜𝑓𝑙𝑖𝑛𝑒𝑠𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑒𝑑𝑡𝑜𝑖𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦
𝑇𝑜𝑡𝑎𝑙𝑛𝑢𝑚𝑏𝑒𝑟𝑜𝑓𝑜𝑟𝑑𝑒𝑟𝑙𝑖𝑛𝑒𝑠
Methodology
Demand/SupplyCharacterization• Usetransactionaldatatofindrealdemanddistribution
• FindmetricsthataffectLIFRperformance
Summarystatistics
Clustering• DemandCoefficientofVariation(COV)• NumberofcountriestheSKUisshippedto(Countries)• Averageordersizequantity(OrderQty)• Daysofsupply(DOS)• Orderfrequency(Transactions)
Cluster N Mean Min Max Mean Min Max Mean Min Max Mean Min Max Mean Min Max Mean Min Max1 987 0.929 0.00 1.00 23,784 28 180,480 8 1 22 91 1 1269 0.45 0.12 2 255 3 13372 13 0.943 0.88 0.97 1,128,427 608,436 1,625,940 30 21 41 34 19 113 0.39 0.18 1 4552 1105 93523 168 0.822 0.00 1.00 3,588 3 121,500 3 1 12 824 0 7160 2.10 1.09 4 21 1 2224 11 0.994 0.94 1.00 656 12 2,700 2 1 5 15648 8660 31839 3.06 1.27 3 10 1 595 230 0.939 0.63 1.00 188,832 22,452 600,876 21 8 35 45 14 139 0.33 0.12 1 1433 291 4850
LIFR OrderQuantity Countries DOS DemandCOV Transactions
Commodities HighVolume ServiceRisk
Sparsedemand HighVolumeCommodities
CLUSTERS
1. Commodities2. HighVolume3. Servicerisk4. Sparsedemand5. Highvolumecommodities
CLUSTERS- MEANCOMPARISON
SimulationModel
OrderQuantity
OrderFrequency
ReplenishmentLT
LIFRTarget:98%
IOHLevel
1 2
SimulationModel
Cluster1 Cluster2 Cluster3 Cluster4 Cluster51stPlace Pareto Gamma Gamma Pareto Exponential2ndPlace Gamma Pareto Pareto Triangular Triangular3rdPlace Triangular Triangular Gamma
OrderQuantitydistribution:~Gamma
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0
20
40
60
80
100
120
140
160
180
200
12 211 410 610 809 1008 1207 1406 1606 1805 2004 2203 2402 2602 2801More
Frequency
Bin
Cluster1:XCW9567
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
0
5
10
15
20
25
30
7 15 22 30 38 46 53 More
Frequency
Bin
Histogram
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
0
2
4
6
8
10
12
1 2 3 4 5 6 7 More
Frequency
Bin
Histogram
Orderfrequencydistribution:~Normal
Issueswithmaterialnumberswithfewdemandpoints- Hardtofitadistribution- TheyhavehighDOSlevels
1 Distributions
MaterialNumber Cluster Shape Scale Shift Mean StDev Mean StDevXCW9567 1 0.714 175.83 12 4.7 1.9 7 2XAVCP359H 2 0.331 2,761.20 36 49.4 8.3 3 1XAV475G 3 0.569 147.68 36 1.0 - 3 1XNMIC572H 4 0.232 125.00 36 1.2 0.4 12 3XZW9252 5 0.421 1,367.10 12 7.5 2.2 3 1
GammaDistributionOrderFrequency ReplenishmentLeadTimeOrderQuantity
NormalDistribution NormalDistribution
SimulationModelAssumptions• Customerordersareexogenousandindependentofinventorylevel.Orderscancomeinevenifthereisnoinventoryonhand.• Customerordersarenotcorrelated• Assuminglotforlotpolicy.• Assumingsupplierisnotoutofstock.• Inventoryavailabilityistheonlyfactoraffectingservice.• Rawdatadonothavesystematictrend/fluctuationacrossseasons.Therandomnessofthevariablesusedinthesimulationaccountforanyseasonalitypresentintherawdata.• Assumingcountries’demandsareindependent.• Unfilledinventorywillbebacklogged.Customerswilleventuallyacceptinventoryregardlessofwhentheyreceiveit.• Orderallocationissequential,onafirst-come-first-servebasis.
SimulationProcess
0 10050 98LIFR
Inv=1k Inv=2kInv=3k
2
Createdemand
distribution
Initializeparameters
Simulateweek-by-week
orderallocationand
receipt
Placeorder
Advancetime
ReportLIFR
Results
0
200000
400000
600000
800000
1000000
1200000
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52
Inventoryo
nHa
nd
WeekNumber
IOHComparisonCluster2:MaterialB
OrderQty ActualIOH Base SS TotalEntitlement Simulated
0
10000
20000
30000
40000
50000
60000
70000
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52
Inventoryo
nHa
nd
WeekNumber
IOHComparisonCluster5:MaterialE
OrderQty ActualIOH Base SS TotalEntitlement Simulated
SKU ACTUALLIFR AVERAGEIOH WEEKSW/OINV. CURRENT INV.LEVEL SIMULATEDLEVEL
A 85.83% 3,160 2 4,100 3,500
B 96.94% 740,928 0 800,000 180,000
C 53.85% 632 1 300 1,000
D 100% 1,142 0 1,500 800
E 90.05% 49,923 0 61,000 26,000
Samplematerialnumberpercluster
Insights
COUNTRYA(BLUE) COUNTRYB(ORANGE)
NO.OFORDER 620 45
TOTALUNITSORDERED 51,696 121,860
LARGESTORDERSIZE 504 23,040
DEMANDCOV 0.21 1.07
Hypothesis:Erraticorderingpatternscausesstraininthesupplychainperformance.Itleadstohigherinventoryrequirementsgiventhesameriskexposure.
Insight1:DemandPatternAnalysis
0
5000
10000
15000
20000
25000
30000
35000
40000
1 2 3 4 5 6 7 8 9 10 11 12
OrderQ
ty
Month
Country ComparisonMaterialB
CountryA CountryB
SimulatedOrderPatternChanges
• Recommendation:reduceordervariability,advancednoticesbeforeorderingabovethreshold
Insight2:Forecasting• Weincrease1periodofforecastingbyassumingweknow80%ofthenextperiod’sdemand
• Resultsshowitmoderatelyimprovesinventoryperformanceundercurrent1weekleadtime
Insight3:Leadtime• Wesimulateda‘supplyshock’whereourreplenishmentlead-timesuddenlyinflatesfrom1weekto1month
• Inventoryperformancedropsdrastically.Weneedtodoublethebaselineinventorytoguaranteeservicelevel
Conclusion• Data-drivenapproachtoinventorymanagement
• Understandingdemandcharacteristics• Clustering• Machinelearning
• Applicabletomanyindustries• Criticalityofproductavailability• RiskManagement• Manageorderpatterns
Q&A
top related