How to design the most efficient pick
and pass system
Alina Stroie
MSc Supply Chain Management
Rotterdam School of Management
Warehouse operating costs
Other
costs;
45%
Source: Tompkins et al.,2003
Order
picking;
55%
2
Warehouse operating costs
Other
costs;
45%
Source: Tompkins et al.,2003
Order
picking;
55%
Significant impact on bottom line
Labor intensive
3
Breakdown of order picker time
Setup
10%
Pick
15%
Other
5%
Travel
50%
Search
20%
Source: Tompkins et al.,2003
4
Breakdown of order picker time
Setup
10%
Pick
15%
Other
5%
Travel
50%
How can we
reduce non-value
added time?
Search
20%
Source: Tompkins et al.,2003
5
Multiple strategies
Popular due to simplicity
Pick-and-pass
Simultaneous picking
Bucket brigades
Relatively low implementation cost
High demand rate
High SKU variety
Small-medium products
6
Pick and Pass Zone Picking
OUT
• To reduce order picking costs, the
storage area is divided into zones,
covered by one or more pickers; this
reduces travel and search time.
IN
Zone D
Zone A
Zone C
Zone B
• Pick-and-pass is an order picking
strategy:
– Each customer order is assigned a tote;
– Totes visit zones in the system;
– At each zone, a picker will retrieve products
from storage and fill the tote before sending
the tote back on the conveyor.
Slide 7
What is the most efficient pick and pass
system?
8
Designing a pick and pass system
Number of segments
Strategic
Tactical
Operational
Number of zones / segment
Use of shortcuts
Block and recirculate protocol
Class-based versus random storage
9
What is the most efficient pick and pass
system in terms of these 5 variables?
10
Modelling pick and pass systems
O
I
Number of segments 4, 6, 8
Number of zones/segments 2, 4, 6
Allowing totes to recirculate Yes or No
Allowing totes to use shortcuts Yes or No
Storage policy Random or
Class-based
storage
Modelled in Material Handling
Simulation Package (MHSP)
Slide 11
Simulation Demo
Material Handling Simulation Program
Simulation parameters
Reasoning and constraints
User determined parameters
Order size UNIF (1,5)
•
The number of orders per simulation
run and the number of simulations per
design are selected based on the
standard deviation of the average
throughput rate. At this setup, the
confidence intervals are sufficiently
small.
•
The order generation rate cannot be
higher, otherwise a blockage occurs in
specific designs.
Order generation EXP (0.0167)
rate totes/second
Number of orders per
1,000
simulation run
Number of simulations
20
per policy set
Picking time EXP (0.2)
picks/second
Slide 13
Finding the most efficient design
Choice to use Data
Envelopment
Analysis (DEA)
Minimum set of assumptions to evaluate models
along different measurement units (in this case
time and cost).
The goal is to compare policy sets with one
another, rather than to find the optimal design for
a pick-and-pass system; DEA achieves this by
ranking policy sets according to their relative
efficiency within the group.
14
Finding the most efficient design
DEA calculates the efficiency of a particular
policy set based on its ability to generate output
given a specific input.
How does DEA
work?
In this case, DEA assigns an efficiency score
based on the ability to achieve the lowest average
throughput time with the lowest total costs
possible.
• DEA input is the total cost of a policy set
• DEA output is based on the average
throughput time of a policy set
15
How did the designs perform?
2200
Top 10 least
efficient have
in total 36 or
48 zones
(2 exceptions)
Total Costs (per 1000 orders) (€)
2000
1800
1600
Top 10 most
efficient have
in total 12 or
16
zones
Most
efficient:
6 segments
2 zones/segm.
4 segments
with shortcuts
4 zones/segm.
class-based
storage
with
shortcuts
no recirculation
1400
6 segments
2 zones/segm.
with shortcuts
1200
1000
800
600
400
200
9,00
9,50
10,00
10,50
11,00
11,50
12,00
12,50
13,00
Average Throughput Time (minutes)
Slide 16
What did I find?
• Designs with few zones and shortcuts are most efficient in the set
• Across all metrics studied (total, investment, and operational cost, average
throughput time, and make span), a smaller number of zones seems to perform
better than designs with many zones.
Relative efficiency (outcome of DEA)
100,00%
80,00%
60,00%
40,00%
20,00%
0,00%
8
12
16
24
32
Total number of zones
36
48
Slide 17
Some interesting clusters…
2200
Allowing
totestotes
to
use
7.4%
decrease in
inaverage
average
Class-based
Allowing
storage
effect
to 0.9%
0.3% increase
shortcuts
over random
recirculate
storage throughput time
Total Costs (per 1000 orders) (€)
2000
1800
1600
1400
1200
1000
800
600
400
200
9,00
9,50
10,00
10,50
11,00
11,50
12,00
12,50
13,00
Average Throughput Time (minutes)
Slide 18
When designing your system…
• No trade-off between total cost and average throughput time
performance: fewer zones perform better on both metrics.
• Number of zones needs to be able to handle demand, otherwise the
system becomes unstable and performs poorly.
• If possible, shortcuts should be implemented as these significantly shorten
the time travelled by totes.
• Storage policy and allowing recirculation should be decided on a case-bycase basis.
Slide 19
Thank you
Appendix
Calculation of average throughput time
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 𝑡𝑖𝑚𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑗
1000
𝑖=1 𝑇ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 𝑡𝑖𝑚𝑒 𝑡𝑜𝑡𝑒𝑖
=
1000
𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑝𝑜𝑙𝑖𝑐𝑦 𝑠𝑒𝑡 𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 𝑡𝑖𝑚𝑒
72
𝑗=1 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑡ℎ𝑟𝑜𝑢𝑔ℎ𝑝𝑢𝑡 𝑡𝑖𝑚𝑒 𝑠𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑗
=
72
22
DEA output
To reverse the goal of maximization of the output, for each DMU, the
average throughput time of that DMU is subtracted from the maximum
average throughput time of all policy sets:
Let 𝑎𝑣𝑔𝑖 denote the average throughput time of 𝐷𝑀𝑈𝑖 .
Let max = max(𝑎𝑣𝑔1 , 𝑎𝑣𝑔2, … , 𝑎𝑣𝑔72)
Then for each DMU the output variable is given by:
(max − 𝑎𝑣𝑔𝑖 )
23
Effect of average throughput time on
the efficiency of the models
Total cost
Efficiency
€ 2.500,00
120,00%
€ 2.000,00
100,00%
80,00%
€ 1.500,00
60,00%
€ 1.000,00
40,00%
€ 500,00
20,00%
€ 0,00
0
5
10
15
20
25
30
35
40
0,00%
0
10
20
30
40
50
60
70
24
80
Results: recirculation
Comparison of policy sets by number of zones and segments, and recirculation
policy
Total Costs (per 1000 orders) (€)
2500,00
2000,00
1500,00
1000,00
500,00
0,00
9,00
9,50
10,00
10,50
11,00
11,50
Average Throughput Time (min)
12,00
12,50
4S 2Z no recirculation
4S 2Z recirculation
4S 4Z no recirculation
4S 4Z recirculation
4S 6Z no recirculation
4S 6Z recirculation
6S 2Z no recirculation
6S 2Z recirculation
6S 4Z no recirculation
6S 4Z recirculation
6S 6Z no recirculation
6S 6Z recirculation
8S 2Z no recirculation
8S 2Z recirculation
8S 4Z no recirculation
8S 4Z recirculation
8S 6Z no recirculation
8S 6Z recirculation
13,00
Slide 25
Results: shortcuts
Comparison of policy sets by number of zones and segments,
and shortcut allowance
Total Costs (per 1000 orders) (€)
2500,00
2000,00
1500,00
1000,00
500,00
0,00
9,00
9,50
10,00
10,50
11,00
11,50
Average Throughput Time (min)
12,00
12,50
4S 2Z no shortcuts
4S 2Z shortcuts
4S 4Z no shortcuts
4S 4Z shortcuts
4S 6Z no shortcuts
4S 6Z shortcuts
6S 2Z no shortcuts
6S 2Z shortcuts
6S 4Z no shortcuts
6S 4Z shortcuts
6S 6Z no shortcuts
6S 6Z shortcuts
8S 2Z no shortcuts
8S 2Z shortcuts
8S 4Z no shortcuts
8S 4Z shortcuts
8S 6Z no shortcuts
8S 6Z shortcuts
13,00
Slide 26
Results: storage policy
Comparison of policy sets by number of zones and segments,
and storage policy
Total Costs (per 1000 orders)
2500,00
2000,00
1500,00
1000,00
500,00
0,00
9,00
9,50
10,00
10,50
11,00
11,50
Average Throughput Time (min)
12,00
12,50
4S 2Z random storage
4S 2Z class-based storage
4S 4Z random storage
4S 4Z class-based storage
4S 6Z random storage
4S 6Z class-based storage
6S 2Z random storage
6S 2Z class-based storage
6S 4Z random storage
6S 4Z class-based storage
6S 6Z random storage
6S 6Z class-based storage
8S 2Z random storage
8S 2Z class-based storage
8S 4Z random storage
8S 4Z class-based storage
8S 6Z random storage
8S 6Z class-based storage
13,00
Slide 27
Effect of increasing the launch rate
• Initial scenario: launch rate = EXP (0.0167) totes/second
• New scenario: launch rate = EXP (0.03) totes/second
– Constraint: models with recirculation and congestion
• Effect:
– Higher average throughput rate
– Lower make span
– Recirculation still increases average throughput time
Make span (hours)
Comparison of make span
based on different launch rate
4
Average throughput time
(seconds)
3000,00
3
2500,00
2
2000,00
1
1500,00
0
1 3 9 11 17 19 25 27 33 35 41 43 49 51 57 59 65 67
DMU
Launch rate = exp(0.03) totes/second
Launch rate = exp(0.0167) totes/second
1000,00
500,00
0,00
1
3
9 11 17 19 25 27 33 35 41 43 49 51 57 59 65 67
Launch rate = exp(0.0167)
Launch rate = exp(0.03)
Slide 28
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Zone
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
Shelf Shelf
29
Most efficient model:
12 zones, shortcuts, class-based storage, no recirculation
Slide 30