Transcript Warehouse Activity Profiling

Warehouse Activity Profiling
Based on Bartholdi & Hackman
Chpt 5
Warehouse Activity Profiling
•The careful measurement and statistical analysis of the warehouse
activity.
•The process of understanding the customer orders that drive the
system
•Sifting through historical data for opportunities and insights that
might confer advantage.
Summary statistics
SKU data
Order data
Location
data
WAP
Distributions
“Structural”
Characterizations, e.g.,
• prevailing patterns/trends
• relations
• dominant elements
SKU-related data
(distributed over a set of data-bases)
• SKU ID
• text description
• product family (product families are defined for each industry and
suggest certain types of storage and handling)
• Addresses of storage location in the warehouse (zone, aisle, section,
shelf, position on the shelf)
• For each location storing the SKU:
–
–
–
–
storage unit
physical dimensions of the storage unit (length, width, height, weight)
scale of the selling unit
number of selling units per storage unit
• Date the SKU was introduced (for assessing growth of the
corresponding activity)
• Max inventory level by month or week (for assessing space needs)
Order-related data
(coming from sales-transactions databases)
•
•
•
•
•
•
•
Order ID
SKU ID
Customer ID
Any needs for special handling
Date/time order was picked
Quantity ordered
Quantity shipped
Remark: This set of data can be really large (the corresponding datafile
might exceed the 100M) => Needs processing through some
specialized Database software.
Data Mining
• Handling a set of tables in a relational database
management system
• Table rows: Records with instances of the object/entity
stored in that table (e.g., SKU’s, order lines, etc.)
• Table columns: Attributes characterizing the considered
entity
• Typical functionality involved in data-mining
– sorting the rows of a table by a certain attribute
– selecting a subset of rows of a table, s.t. all isolated entities satisfy
a certain property
– counting distinct entries in a table meeting a certain condition
– performing joins, i.e., combining the information one table with
that of another table to create a new table with a different set of
attributes
– graphing the results
• SQL: Structured Query Language
Some basic summary statistics
• Order-related
–
–
–
–
–
average number of SKU’s involved (work and storage complexity)
average number of orders shipped per day (volume of activity)
average number of lines (SKU’s) per order (picking complexity)
average number of units per line
seasonalities (Seasonal Indices: What percentage of a cycle
corresponds to a period in the cycle - temporal distribution of the
work)
• Facility-related data
– area of the warehouse
– average number of shipments received per day(the “backend”
activity)
– average rate of introduction of new SKU’s (operational stability)
– average number of SKU’s in the warehouse (volume and scope of
operations)
– distribution of the personnel to the various activities (labor-related
costs and opportunities)
A closer characterization of the
warehouse workload
• What drives the entire warehouse activity is the order/pick
lines!
• Need to understand how these lines are distributed among
–
–
–
–
–
SKU’s
product families
storage locations
warehouse zones
time
• Activity analysis
• Results are communicated as
– discrete distributions
– Pareto curves, i.e., cumulative distributions where the items on the
horizontal axis are arranged in a decreasing order w.r.t. the
corresponding value of the distribution.
– other plots (e.g., bird’s eye view for characterizing location activity)
Graphing the results of the
Activity Analysis
Discrete Distribution
% picks
1.0
A
B
C
zone
D
Pareto curve
% picks
1.0
10K
20K
SKU’s
Pareto Effect and ABC Analysis
• Pareto Effect: A small percentage of the considered entities
account for the largest fraction of the activity (20/80 rule)
• ABC analysis: Exploit the Pareto effects in order to
classify the considered entities into (typically three: A, B
and C) categories, such that
– the entities in the first category are the ones responsible for most of
the activity, and therefore, more closely managed;
– the entities in the second category account for most of the
remaining part, and therefore, are moderately important;
– the entities in the third category are the largest bulk responsible for
only a small part of the activity, and therefore, insignificant.
• Remark: ABC classification of the same set of entities will
differ from activity to activity (c.f. Bartholdi & Hackman,
Tables 5.1 - 5.5)
Work Patterns and their Implications
• Distribution of lines per order: What percentage of orders
have a single line, two lines, etc. (Reveals possibilities for
batching and/or zoning)
• Distribution of picks by order-size: What fraction of picks
comes from single-line orders, two-line orders, etc.
(reveals whether most work is generated by small or large
orders, shipping activity)
• Distribution of families/zones per order: What fraction of
orders involves a single family/zone, two families/zones,
etc. (identifies coupling which can be exploited by the
picking process)
• Family pairs analysis / “order-crossings” (for zones):
identify pairs of families/zones with correlated demand
(this correlation should be exploited by putting items in
each pair close to each other)
“Case Study”:
Profiling the Activity of a Wholesales
Distributor of Office Products
•Problem description:
•http://www.isye.gatech.edu/people/faculty/John_Bartholdi/wh/book/profile/projects/projects.html
•Problem Solution:
•http://www.isye.gatech.edu/~spyros/courses/IE6202/WAP-cs.pdf