SlideShare ist ein Scribd-Unternehmen logo
1 von 26
Downloaden Sie, um offline zu lesen
Real Time Fulfilment Planning at
Flipkart Scale
Jagadeesh Huliyar
jagadeesh.huliyar@gmail.com
The Fifth Elephant 2016
29th July 2016
What is it?
➔ Flipkart stores and sells
millions of unique items
through its Fulfillment
Centers (FCs) and Sellers.
➔ These items need to be
picked from FCs or Seller
Locations and delivered to
End Customers.
➔ Real Time Fulfilment
Planner is responsible for
planning the schedule for
these activities.
What exactly happens
Decisions
Are there any decisions to be made?
When should pick start in the FC? What is the order or batch for pick?
When should other activities be performed? Should items be held somewhere?
Which transport connection and mode should be taken?
Which route should be taken during transportation?
Do we really need to take these decisions? What happens if we process as and
when order are placed or as and when shipments arrive at a particular location?
Transportation Schedule
Downstream Capacity
Storage Usage
➔ Hour 1
◆ Operation-1 processes 75 small units.
◆ Operation-2 will be able to process 50 small units and remaining 25 will remain in storage.
➔ Hour 2
◆ Operation-1 processes 50 large units.
◆ Operation-2 can processes these 50 large units plus 25 small present in storage.
➔ This average processing over two hours is 62 units and at the end of two hours storage is empty.
Transportation Batching and Routing
Priority, Breach Cost and Capacity Consumption
Case 1 Case 2
Supply 2 units 2 units
Demand 2 COD and 2 Prepaid. 2 COD and 2 Prepaid.
Service Type All Regular COD are NDD.
Pre-paid are Regular
Consumption COD = 2 units.
Prepaid = 1 unit
COD = 2 units.
Prepaid = 1 unit
Breach Cost Regular = 20 rupees
NDD = 70 rupees
Regular = 20 rupees
NDD = 70 rupees
Decision 2 Prepaid 1 COD
Priority, Breach Cost and Capacity Consumption
➔ Consumption Factory vary as per Stage of
processing
◆ Sortation Station
● Size
◆ FC
● Proximity to packing stations
● Packaging Type
◆ Delivery Centre
● Payment Type
● Address Type
➔ Breach cost or Profit on timely delivery
◆ Service Type
◆ Customer Type
Gigantic Graph
➔ Items : Different dimensions, Volume and Weight → Different
restrictions and shipping costs.
➔ Constraints
◆ Transport Connection Time
◆ Priority of Shipments
◆ SLA
Need to do all this at Lowest Possible Cost
Variability
➔ Demand Variability
◆ Destination Mix Variability
◆ Product Mix Variability
◆ Seller Mix Variability
◆ Service Mix Variability
➔ Supply Variability
◆ Dependency on temporary work
force
◆ Unreliable Transport Capacity
◆ Network and infrastructure issues
◆ Changing rules and regulations
Modelling
Planned Storage
Unplanned Storage
Operations
Capacity
Lead Time
Graph - Capacity and Lead Time
Capacity Modelling with Storage
Graph
Plan Creation
➔ Computation Graph Creation
➔ Start and Terminal Nodes Identification
➔ Feasible Set
➔ Optimisation Function
➔ Allocation
➔ Learning
➔ Usage of Learning
➔ Repeat
➔ Terminating Conditions
➔ Match
Scalability
➔ Nodes - Millions
➔ Edges - Tens of millions
➔ Need to distribute the
computation
Bulk synchronous parallel and Pregel
➔ BSP system consists of
◆ Local Memory Transactions.
◆ Message routing across Components.
◆ Synchronisation
➔ Pregel - Inspired by BSP
◆ Graph is divided into Partitions
◆ Algorithm is modelled as Computation
at Vertex and Message Passing across
Vertices.
◆ Sequence of supersteps
Example
Giraph
Giraph Job Lifetime
Real Time Fulfillment Plan
➔ Vertices - Time Slots in a Facility or Transport
➔ Edges - Movement of Shipments from one Time Slot in a Facility/Transport to
another Time Slot in a Facility/Transport
➔ Input - Requests Received and Forecast grouped by Dimensions
➔ Messages - Allocation groups and Learnings (Bottlenecked Dimensions and
Variable Capacity Requirement)
➔ Output - Set of Plans : Dimension Group : Time Slot : Number of Shipments
Simulated Annealing - Global Optima and Local Optima
➔ Starting with an initial solution. Learn better
allocation strategies. Move to a better
neighbouring solution.
➔ It can lead to situations where you're stuck at
a sub-optimal place.
➔ Simulated annealing injects right amount of
randomness into things to escape local
maxima.
Local Planning
Once Global Plan is ready there are some local decisions to be taken. These do
not affect the Global Plan and are hence taken independently
➔ Pick Path Optimisation
➔ Sorting Configuration
➔ Last Mile Vehicle Routing Problem
References
➔ Distributed multi-agent optimization
◆ http://www.cds.caltech.edu/~murray/preprints/ttm11-ifac_s.pdf
➔ Apache Giraph
◆ http://www.slideshare.net/ClaudioMartella/giraph-at-hadoop-summit-2014
◆ http://giraph.apache.org/
◆ http://www.apress.com/9781484212523
➔ BSP
◆ https://en.wikipedia.org/wiki/Bulk_synchronous_parallel
➔ Reactive Search and Intelligent Optimization
◆ REACTIVE SEARCH AND INTELLIGENT OPTIMIZATION Roberto Battiti, Mauro Brunato, and Franco Mascia
➔ Flow shop scheduling
◆ http://faculty.ksu.edu.sa/ialharkan/IE428/Chapter_4.pdf
◆ https://en.wikipedia.org/wiki/Flow_shop_scheduling
➔ Ant colony optimization
◆ http://www.scholarpedia.org/article/Ant_colony_optimization
◆ M. Dorigo, M. Birattari & T. Stützle, 2006 Ant Colony Optimization: Artificial Ants as a Computational Intelligence
Technique. TR/IRIDIA/2006-023
➔ Simulated Annealing
◆ https://en.wikipedia.org/wiki/Simulated_annealing
➔ Images have been picked by searching Google Images

Weitere ähnliche Inhalte

Andere mochten auch

Herramientas de la reingeniería
Herramientas de la reingenieríaHerramientas de la reingeniería
Herramientas de la reingenieríaHermes Niño
 
Super Engineering Works Delhi India
Super Engineering Works Delhi IndiaSuper Engineering Works Delhi India
Super Engineering Works Delhi IndiaPuja Pal
 
Omni channel fulfilment and supply chain management analytic
Omni channel fulfilment and supply chain management analyticOmni channel fulfilment and supply chain management analytic
Omni channel fulfilment and supply chain management analyticAmit Kumar Garg
 
Types of advertising media
Types of advertising mediaTypes of advertising media
Types of advertising mediadeepu2000
 
Easy Serve, an Online Services Marketplace Startup in India
Easy Serve, an Online Services Marketplace Startup in IndiaEasy Serve, an Online Services Marketplace Startup in India
Easy Serve, an Online Services Marketplace Startup in IndiaUpamanyu Acharya
 
Iain at vdepot, ecommerce fulfilment company
Iain at vdepot, ecommerce fulfilment companyIain at vdepot, ecommerce fulfilment company
Iain at vdepot, ecommerce fulfilment companyDaytodayebay
 
Advertising Agencies & Functioning
Advertising Agencies  & FunctioningAdvertising Agencies  & Functioning
Advertising Agencies & FunctioningAnubha Rastogi
 
Manoj(Java Developer)_Resume
Manoj(Java Developer)_ResumeManoj(Java Developer)_Resume
Manoj(Java Developer)_ResumeVamsi Manoj
 
Supply Chain Management of Amazon India
Supply Chain Management of Amazon IndiaSupply Chain Management of Amazon India
Supply Chain Management of Amazon IndiaAbhisek Khatua
 
E-Commerce Models and Web 2.0 in Supply Chain
E-Commerce Models and Web 2.0 in Supply ChainE-Commerce Models and Web 2.0 in Supply Chain
E-Commerce Models and Web 2.0 in Supply ChainArgha Ray
 

Andere mochten auch (15)

Vinyl records
Vinyl recordsVinyl records
Vinyl records
 
Herramientas de la reingeniería
Herramientas de la reingenieríaHerramientas de la reingeniería
Herramientas de la reingeniería
 
Salma's CV
Salma's CVSalma's CV
Salma's CV
 
Super Engineering Works Delhi India
Super Engineering Works Delhi IndiaSuper Engineering Works Delhi India
Super Engineering Works Delhi India
 
Omni channel fulfilment and supply chain management analytic
Omni channel fulfilment and supply chain management analyticOmni channel fulfilment and supply chain management analytic
Omni channel fulfilment and supply chain management analytic
 
Types of advertising media
Types of advertising mediaTypes of advertising media
Types of advertising media
 
Easy Serve, an Online Services Marketplace Startup in India
Easy Serve, an Online Services Marketplace Startup in IndiaEasy Serve, an Online Services Marketplace Startup in India
Easy Serve, an Online Services Marketplace Startup in India
 
MP Seller Starter Kit - Malaysia
MP Seller Starter Kit - MalaysiaMP Seller Starter Kit - Malaysia
MP Seller Starter Kit - Malaysia
 
Iain at vdepot, ecommerce fulfilment company
Iain at vdepot, ecommerce fulfilment companyIain at vdepot, ecommerce fulfilment company
Iain at vdepot, ecommerce fulfilment company
 
Advertising Agencies & Functioning
Advertising Agencies  & FunctioningAdvertising Agencies  & Functioning
Advertising Agencies & Functioning
 
Flipkart
FlipkartFlipkart
Flipkart
 
Manoj(Java Developer)_Resume
Manoj(Java Developer)_ResumeManoj(Java Developer)_Resume
Manoj(Java Developer)_Resume
 
Supply Chain Management of Amazon India
Supply Chain Management of Amazon IndiaSupply Chain Management of Amazon India
Supply Chain Management of Amazon India
 
E-Commerce Models and Web 2.0 in Supply Chain
E-Commerce Models and Web 2.0 in Supply ChainE-Commerce Models and Web 2.0 in Supply Chain
E-Commerce Models and Web 2.0 in Supply Chain
 
Amazon supply chain
Amazon supply chainAmazon supply chain
Amazon supply chain
 

Ähnlich wie Real Time Fulfilment Planning

Supply optimization instacart
Supply optimization instacartSupply optimization instacart
Supply optimization instacartJagannath Putrevu
 
Logistics - Operational Planning - for XLRI PGCLSM
Logistics - Operational Planning - for XLRI PGCLSMLogistics - Operational Planning - for XLRI PGCLSM
Logistics - Operational Planning - for XLRI PGCLSMVinodh Soundarajan
 
Project presentation som group 1
Project presentation som group 1Project presentation som group 1
Project presentation som group 1blackbat999
 
DNow Internship Presentation 2014
DNow Internship Presentation 2014DNow Internship Presentation 2014
DNow Internship Presentation 2014Jonathan Scott
 
4. Joe Kraft, Minemax - Minemax Scheduler Applications
4. Joe Kraft, Minemax - Minemax Scheduler Applications4. Joe Kraft, Minemax - Minemax Scheduler Applications
4. Joe Kraft, Minemax - Minemax Scheduler ApplicationsKristy Marshall
 
Capacity
CapacityCapacity
Capacityarggarg
 
Cindi Hane – Logistics Management “Show Me the Time Management Money”
Cindi Hane – Logistics Management  “Show Me the Time Management Money”Cindi Hane – Logistics Management  “Show Me the Time Management Money”
Cindi Hane – Logistics Management “Show Me the Time Management Money”Elemica
 
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics Program
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics ProgramCSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics Program
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics ProgramArrowStream
 
Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019samjune
 
Akshaya Patra Foundation.pptx
Akshaya Patra Foundation.pptxAkshaya Patra Foundation.pptx
Akshaya Patra Foundation.pptxSudipto Mukherjee
 
Physical Distribution Basics & New Trends
Physical Distribution Basics & New TrendsPhysical Distribution Basics & New Trends
Physical Distribution Basics & New TrendsAbdelghafar Gamil
 
Introduction to Management Science and Linear Programming
 Introduction to Management Science and Linear Programming  Introduction to Management Science and Linear Programming
Introduction to Management Science and Linear Programming Kishore Morya PhD.
 
Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019Akash Maurya
 
TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP
 TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP
TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIPiQHub
 
Real-World Resiliency: Surviving Datacenter Disaster
Real-World Resiliency: Surviving Datacenter DisasterReal-World Resiliency: Surviving Datacenter Disaster
Real-World Resiliency: Surviving Datacenter DisasterScyllaDB
 
Lean logistics and warehousing final
Lean logistics and warehousing finalLean logistics and warehousing final
Lean logistics and warehousing finalCody White
 

Ähnlich wie Real Time Fulfilment Planning (20)

Supply optimization instacart
Supply optimization instacartSupply optimization instacart
Supply optimization instacart
 
Deck
DeckDeck
Deck
 
Logistics - Operational Planning - for XLRI PGCLSM
Logistics - Operational Planning - for XLRI PGCLSMLogistics - Operational Planning - for XLRI PGCLSM
Logistics - Operational Planning - for XLRI PGCLSM
 
Project presentation som group 1
Project presentation som group 1Project presentation som group 1
Project presentation som group 1
 
DNow Internship Presentation 2014
DNow Internship Presentation 2014DNow Internship Presentation 2014
DNow Internship Presentation 2014
 
4. Joe Kraft, Minemax - Minemax Scheduler Applications
4. Joe Kraft, Minemax - Minemax Scheduler Applications4. Joe Kraft, Minemax - Minemax Scheduler Applications
4. Joe Kraft, Minemax - Minemax Scheduler Applications
 
Capacity
CapacityCapacity
Capacity
 
Cindi Hane – Logistics Management “Show Me the Time Management Money”
Cindi Hane – Logistics Management  “Show Me the Time Management Money”Cindi Hane – Logistics Management  “Show Me the Time Management Money”
Cindi Hane – Logistics Management “Show Me the Time Management Money”
 
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics Program
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics ProgramCSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics Program
CSCMP 2014 Breaking the Barrier to a Lower Cost Inbound Logistics Program
 
desk top
desk topdesk top
desk top
 
Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019
 
Tqm final case
Tqm final caseTqm final case
Tqm final case
 
Akshaya Patra Foundation.pptx
Akshaya Patra Foundation.pptxAkshaya Patra Foundation.pptx
Akshaya Patra Foundation.pptx
 
Physical Distribution Basics & New Trends
Physical Distribution Basics & New TrendsPhysical Distribution Basics & New Trends
Physical Distribution Basics & New Trends
 
Introduction to Management Science and Linear Programming
 Introduction to Management Science and Linear Programming  Introduction to Management Science and Linear Programming
Introduction to Management Science and Linear Programming
 
Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019Capacity requirement planning sure 12mt07ind019
Capacity requirement planning sure 12mt07ind019
 
TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP
 TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP
TRANSPORT & STORAGE ADVANCEMENTS ACROSS RAIL, TRUCK, TRANSLOAD, BARGE, AND SHIP
 
Solve my Problem When I Want
Solve my Problem When I Want   Solve my Problem When I Want
Solve my Problem When I Want
 
Real-World Resiliency: Surviving Datacenter Disaster
Real-World Resiliency: Surviving Datacenter DisasterReal-World Resiliency: Surviving Datacenter Disaster
Real-World Resiliency: Surviving Datacenter Disaster
 
Lean logistics and warehousing final
Lean logistics and warehousing finalLean logistics and warehousing final
Lean logistics and warehousing final
 

Real Time Fulfilment Planning

  • 1. Real Time Fulfilment Planning at Flipkart Scale Jagadeesh Huliyar jagadeesh.huliyar@gmail.com The Fifth Elephant 2016 29th July 2016
  • 2. What is it? ➔ Flipkart stores and sells millions of unique items through its Fulfillment Centers (FCs) and Sellers. ➔ These items need to be picked from FCs or Seller Locations and delivered to End Customers. ➔ Real Time Fulfilment Planner is responsible for planning the schedule for these activities.
  • 4. Decisions Are there any decisions to be made? When should pick start in the FC? What is the order or batch for pick? When should other activities be performed? Should items be held somewhere? Which transport connection and mode should be taken? Which route should be taken during transportation? Do we really need to take these decisions? What happens if we process as and when order are placed or as and when shipments arrive at a particular location?
  • 7. Storage Usage ➔ Hour 1 ◆ Operation-1 processes 75 small units. ◆ Operation-2 will be able to process 50 small units and remaining 25 will remain in storage. ➔ Hour 2 ◆ Operation-1 processes 50 large units. ◆ Operation-2 can processes these 50 large units plus 25 small present in storage. ➔ This average processing over two hours is 62 units and at the end of two hours storage is empty.
  • 9. Priority, Breach Cost and Capacity Consumption Case 1 Case 2 Supply 2 units 2 units Demand 2 COD and 2 Prepaid. 2 COD and 2 Prepaid. Service Type All Regular COD are NDD. Pre-paid are Regular Consumption COD = 2 units. Prepaid = 1 unit COD = 2 units. Prepaid = 1 unit Breach Cost Regular = 20 rupees NDD = 70 rupees Regular = 20 rupees NDD = 70 rupees Decision 2 Prepaid 1 COD
  • 10. Priority, Breach Cost and Capacity Consumption ➔ Consumption Factory vary as per Stage of processing ◆ Sortation Station ● Size ◆ FC ● Proximity to packing stations ● Packaging Type ◆ Delivery Centre ● Payment Type ● Address Type ➔ Breach cost or Profit on timely delivery ◆ Service Type ◆ Customer Type
  • 11. Gigantic Graph ➔ Items : Different dimensions, Volume and Weight → Different restrictions and shipping costs. ➔ Constraints ◆ Transport Connection Time ◆ Priority of Shipments ◆ SLA Need to do all this at Lowest Possible Cost
  • 12. Variability ➔ Demand Variability ◆ Destination Mix Variability ◆ Product Mix Variability ◆ Seller Mix Variability ◆ Service Mix Variability ➔ Supply Variability ◆ Dependency on temporary work force ◆ Unreliable Transport Capacity ◆ Network and infrastructure issues ◆ Changing rules and regulations
  • 14. Graph - Capacity and Lead Time
  • 16. Graph
  • 17. Plan Creation ➔ Computation Graph Creation ➔ Start and Terminal Nodes Identification ➔ Feasible Set ➔ Optimisation Function ➔ Allocation ➔ Learning ➔ Usage of Learning ➔ Repeat ➔ Terminating Conditions ➔ Match
  • 18. Scalability ➔ Nodes - Millions ➔ Edges - Tens of millions ➔ Need to distribute the computation
  • 19. Bulk synchronous parallel and Pregel ➔ BSP system consists of ◆ Local Memory Transactions. ◆ Message routing across Components. ◆ Synchronisation ➔ Pregel - Inspired by BSP ◆ Graph is divided into Partitions ◆ Algorithm is modelled as Computation at Vertex and Message Passing across Vertices. ◆ Sequence of supersteps
  • 23. Real Time Fulfillment Plan ➔ Vertices - Time Slots in a Facility or Transport ➔ Edges - Movement of Shipments from one Time Slot in a Facility/Transport to another Time Slot in a Facility/Transport ➔ Input - Requests Received and Forecast grouped by Dimensions ➔ Messages - Allocation groups and Learnings (Bottlenecked Dimensions and Variable Capacity Requirement) ➔ Output - Set of Plans : Dimension Group : Time Slot : Number of Shipments
  • 24. Simulated Annealing - Global Optima and Local Optima ➔ Starting with an initial solution. Learn better allocation strategies. Move to a better neighbouring solution. ➔ It can lead to situations where you're stuck at a sub-optimal place. ➔ Simulated annealing injects right amount of randomness into things to escape local maxima.
  • 25. Local Planning Once Global Plan is ready there are some local decisions to be taken. These do not affect the Global Plan and are hence taken independently ➔ Pick Path Optimisation ➔ Sorting Configuration ➔ Last Mile Vehicle Routing Problem
  • 26. References ➔ Distributed multi-agent optimization ◆ http://www.cds.caltech.edu/~murray/preprints/ttm11-ifac_s.pdf ➔ Apache Giraph ◆ http://www.slideshare.net/ClaudioMartella/giraph-at-hadoop-summit-2014 ◆ http://giraph.apache.org/ ◆ http://www.apress.com/9781484212523 ➔ BSP ◆ https://en.wikipedia.org/wiki/Bulk_synchronous_parallel ➔ Reactive Search and Intelligent Optimization ◆ REACTIVE SEARCH AND INTELLIGENT OPTIMIZATION Roberto Battiti, Mauro Brunato, and Franco Mascia ➔ Flow shop scheduling ◆ http://faculty.ksu.edu.sa/ialharkan/IE428/Chapter_4.pdf ◆ https://en.wikipedia.org/wiki/Flow_shop_scheduling ➔ Ant colony optimization ◆ http://www.scholarpedia.org/article/Ant_colony_optimization ◆ M. Dorigo, M. Birattari & T. Stützle, 2006 Ant Colony Optimization: Artificial Ants as a Computational Intelligence Technique. TR/IRIDIA/2006-023 ➔ Simulated Annealing ◆ https://en.wikipedia.org/wiki/Simulated_annealing ➔ Images have been picked by searching Google Images