Cost Optimisation for Unilever.pdf
0 likes34 views
This project developed an automated data analytics toolkit to optimize logistics costs for Unilever. The toolkit uses data mining and visualization techniques on Unilever's logistics data extracted from SAP. It consolidates shipping data, applies cost savings algorithms, and generates an interactive dashboard for visualizing results. The dashboard allows dynamic criteria selection to compute potential cost savings in minutes, improving over manual methods. Recommendations include standardizing input data, providing programming documentation and adopting an agile development approach for future enhancements.
![National University
of Singapore
Cost Optimisation for Unilever's Logistic
Scheduling through Data Mining and Visualization
IE3100M Systems Design Project (Group 12)
Team Members: Chan Hua Han I Lim Cheng Kiat I Darren Pang I Liu Peng I Chow Jun Hao Jacky
Supervising Professors: Prof. Andrew Lim I Dr. Li Chongshou
Unilever Supervisors: Ms. Zoey Zhang I Mr. Daniel Cortez
Course Co-ordinator: Dr. Bok Shung Hwee
Department of Industrial Systems
Engineering and Management
(ISEM)
INTRODUCTION PROBLEM DESCRIPTION
With increasing demand for fast moving consumer goods,
this incurs rising supply chain operation costs as a result
of sub-optimal logistic shipment allocations. Unilever is
furthering their analytics capabilities to increase efficiency. Scheduling Analytics Order Fulfilment
Incoming Order Dispatchment
* Unilever currently relies on SAP Enterprise Resource Planning
SHIPMENTS LOGISTICS COSTS ANALYTICS TOOLKIT * Data Extraction is confined to reporting and ad-hoc usage only
This project aims to develop a data mining tool to enable deep dive
and analytics for identifying potential areas of improvements in the
logistics costs.
* Increasing pressure to manage complexity and scale of logistic operations
* Need for digitalizing operations framework for continuous business insights
Problem Formulation:
Understanding the planning
and processes behind the
data to identify the problem
STEP 1: CONCEPTUALIZATION
Data Consolidation: Descriptive Analytics:
Consolidate the extracted Generate first layer of
data to derive information performace indicators to
required for analytics display trends and
anomalies
STEP 2: PRELIMINARY DESIGN
...-,..coo,..,...,......,,""
--=-鐃"鐃-"'
0--- -揃"'
..,
鐃:-
a:: 鐃-
__/lo,..__.,___
""
'属"'"-"-.port"'
'"'"属'"...
揃揃-揃
,,,,,..,..,,,,.,_,.,
.......,...,,.,,,....,,
.....,,...,,,,.,._.,
鐃-...-鐃盾,_,.......,,
.......""'.....,,,.....,
""'"''"'""'"""
........,.,,,...-
' ,..:-,11.'7
S l>W1'1
l."50.m
"'-""'揃'"' W+M@
IOmSl<TOI Sl<IP"""" [DY>)
Criteria Formulation:
Formulate cost savings
methods to evaluate and
compare between different
computation
methods
STEP 3: ALGORITHMS
Method 1: Right Truck Sizing
- KEY OBJECTIVES
Utilize existing data to provide better
analytical tools for business analytics
Provide cost savings methods to improve
overall logistic excellence
Integrate above features into a single
executable dashboard toolkit for ease of
use
Ensure accessibility of usage through
multiple platforms
DEVISE
KPls to empower
stakeholders in
everyday operations
ANALYZE
Current system
performance to
maintain and
improve quality
OPTIMIZE
I Propose strategies
to augment current
performance
Visualization: Results Analysis:
Generate dashboard Validation of results
toolkit for visualization on to evaluate their
the front end and backend effectiveness
formulation of program for
automation
STEP 4: AUTOMATED TOOL
Each shipment is checked for loadfill percentage Database
Local database
---揃揃揃揃揃----- '" ..._ 揃-揃
00WMt 0F 5H PM<N1'1W<D<LY]
.:鐃
..
..
-
----
"" Olllo, >l<IPMeNn t'N""''
鐃緒申-
----
If smaller vehicle available, consider weight and
volume
If found, highlight and calculate the percentage
difference in cost
Method 2: Truck Consolidation
-
Initial toolkit used Excel VBA to attain graphical trends
but this was felt to be too slow in producing required
results and further development would be hindered
Data Manipulation DataTransition DataVisualization
Based on the design values, the programming
language and toolkit was then narrowed down.
JavaScript
Criteria Evaluation
(1) Parameters
1. Temperature Conditions
2. Transport Scope
3. Transport Mode
Identify consecutive shipments of same description
based on user input
Check for suitable truck combination to hold total
loadfill
Cost and
shipment level
dataset from
Unilever SAP
Comma-separated
value file
conversion
C)
fython
Cleaning of raw
dataset
Sanitisation of
non-conforming sets
Cost profile
calculations
Apply cost savings
techniques C)
Output from
python in form of
comma-separated
value file
Analytics
extractions into
JavaScript
C)
財
JavaScript
Interactive Dashboard
with multiple criteria
filter
Able to deep dive into
individual shipments
for investigation
C)
4. Sanitized Data
(2) Processing Time
Approximately 30 seconds per click per criterion
Compare new cost against originial sum. If lower,
compute the cost saving
The toolkit works in a single pass format by passing all the input data
computations from Python into Javascript for dashboard display
DASHBOARD IMPACT OF AUTOMATED TOOLKIT
PRODUCT IMPLEMENTATION
Day-of-weekJ
and
Month-of-year
Selection
Cost Sevings Dela Table
Date Ran11e: 31/01/2017 to 30/04/201鐃
33,500sel8'Cted out of33,500 records I Rwlll.Ai
L Data Records
Report can now be generated
completely within minutes compared to
a few hours previously
U...Flloflo:a,mr.,:11,10鐃盾
30.600-..00UI.........,_,_,..
Da-;-<Jf-Wee,: Profile ShiD11ents Overve,i
0 ,,ooo <,000 "-''" 0,000 10.000 12,000 o 1,om e,ooo ,,om ,,ooo ,,ooa ,..o, """" ..,,o ..... "'"'"" ...,,, ""''属 "'"'"" ApO> ""'"' ,..,. """" "'
Dynamic
Pie Chart 7 "-"""""'
Parameters L_
Mode ol Tfal3P(lt
Dr"!! to a.e1,,ct 1ime dur"1ion
Terr鐃ti..re 0.:n:mlon
Criteria Selection Interface
Final product design takes in data input from the user and enables dynamic criteria
selection which computes the cost saving based on the the selection concurrently
KEY SKILLSETS
Human Factors Engineering (HFE) principles
were applied to improve the visualization of the
model
Statistics knowledge and data analytics skills
were applied to interpret and evaluate the
significance of the data
Engineering communications and human
resourcemanagement wereadopted to facilitate
the interaction with stakeholders and team
Software engineering techniques used to enable
automation of toolkit
LIMITATIONS
J- End-user has to learn python programming for
effectiveuse and future modifications of the new
dashboard reporting tool
!- Certain components of the dashboard tool
require manual adjustments
RECOMMENDATIONS
t+
Standardization of input data to reduce need for
manual adjustments
t+ Python and Javascript programming for end-user
as well as documentation for code
., !
--h.d<--
,. ,- DAS!-IBOARD
L Dynamic
Date
Selection
This increases productivity and allows the
company to focus on mitigating cost
savings in the shipment logistics rather
l. . . . ,. . . .
;:揃1
1
揃1' 揃1'....
Percentage
Weight Programming Logic
' . ' ' ' '
., :
"" - .;. ... ...
---鐃-
-
: 揃揃揃n揃揃i揃揃a揃揃i揃
I鐃:i,_u 揃,.揃揃=-(e=.i:
- l_ .:.. 鐃..-,
,,. .:..
----
F
i:
-鐃渋 .
. .
..'
- -
l..
Based on the output charts and cost savings summary, identify
trends and potentialcost saving allocations from the cost saving
methods j≒1'揃揃1'揃揃揃.揃
l. . . . ,. . .
. .
' ': . ', ' '
1i 揃-lll
' '
' . ' '
C:W<>"揃--- -----
I Month-of-year/ Day-of-week Chart
1. Shipment Profile 4, Volume Profile
2. Weight Profile 5, Truck Size Profile
3. Loadfill Profile
I Cost Savings - Before and After
1. Daily Truck Size
2. MonthlyTruck Size
3. Percentage Savings
f+
Currrently adopting an agile methodology framework for
this proJect Future work can focus on using a waterfall
programming methodology framework to scale up the
dashboard
鐃
Waterfall
software
鐃 development
鐃
cycle
鐃
Q]llll9
.. -=--
Cost Savings Graphs and Charts Interface
VALIDATION OF PRODUCT TOOLKIT
Criterion
Practicality
Workability
Accessibility
Flexibility
Description
Dashboard is deemed a practical visualization tool by the company. since visuals are easy
to comprehend and computation of required information is quick and efficient
Dashboard displays results in an ergonomic manner and allows user to easily modify the layout
according to specific analysis needs
Dashboard requires less time and cost to gain access than conventional methods since data
extraction is separate from the analytics toolkit
Dashboard allows for further development to incude additional features to be included
D,:o_,,,,.,,.,,.,-,,,,,
鐃-""属-""'"'"'-""",_I_"'
Remark
*** Data shown on this poster are fabricated for demonstration purposes only.](https://image.slidesharecdn.com/costoptimisationforunilever-230610122053-e7fa4cc7/85/Cost-Optimisation-for-Unilever-pdf-1-320.jpg)
Cost Optimisation for Unilever.pdf
- 1. National University of Singapore Cost Optimisation for Unilever's Logistic Scheduling through Data Mining and Visualization IE3100M Systems Design Project (Group 12) Team Members: Chan Hua Han I Lim Cheng Kiat I Darren Pang I Liu Peng I Chow Jun Hao Jacky Supervising Professors: Prof. Andrew Lim I Dr. Li Chongshou Unilever Supervisors: Ms. Zoey Zhang I Mr. Daniel Cortez Course Co-ordinator: Dr. Bok Shung Hwee Department of Industrial Systems Engineering and Management (ISEM) INTRODUCTION PROBLEM DESCRIPTION With increasing demand for fast moving consumer goods, this incurs rising supply chain operation costs as a result of sub-optimal logistic shipment allocations. Unilever is furthering their analytics capabilities to increase efficiency. Scheduling Analytics Order Fulfilment Incoming Order Dispatchment * Unilever currently relies on SAP Enterprise Resource Planning SHIPMENTS LOGISTICS COSTS ANALYTICS TOOLKIT * Data Extraction is confined to reporting and ad-hoc usage only This project aims to develop a data mining tool to enable deep dive and analytics for identifying potential areas of improvements in the logistics costs. * Increasing pressure to manage complexity and scale of logistic operations * Need for digitalizing operations framework for continuous business insights Problem Formulation: Understanding the planning and processes behind the data to identify the problem STEP 1: CONCEPTUALIZATION Data Consolidation: Descriptive Analytics: Consolidate the extracted Generate first layer of data to derive information performace indicators to required for analytics display trends and anomalies STEP 2: PRELIMINARY DESIGN ...-,..coo,..,...,......,,"" --=-鐃"鐃-"' 0--- -揃"' .., 鐃:- a:: 鐃- __/lo,..__.,___ "" '属"'"-"-.port"' '"'"属'"... 揃揃-揃 ,,,,,..,..,,,,.,_,., .......,...,,.,,,....,, .....,,...,,,,.,._., 鐃-...-鐃盾,_,.......,, .......""'.....,,,....., ""'"''"'""'""" ........,.,,,...- ' ,..:-,11.'7 S l>W1'1 l."50.m "'-""'揃'"' W+M@ IOmSl<TOI Sl<IP"""" [DY>) Criteria Formulation: Formulate cost savings methods to evaluate and compare between different computation methods STEP 3: ALGORITHMS Method 1: Right Truck Sizing - KEY OBJECTIVES Utilize existing data to provide better analytical tools for business analytics Provide cost savings methods to improve overall logistic excellence Integrate above features into a single executable dashboard toolkit for ease of use Ensure accessibility of usage through multiple platforms DEVISE KPls to empower stakeholders in everyday operations ANALYZE Current system performance to maintain and improve quality OPTIMIZE I Propose strategies to augment current performance Visualization: Results Analysis: Generate dashboard Validation of results toolkit for visualization on to evaluate their the front end and backend effectiveness formulation of program for automation STEP 4: AUTOMATED TOOL Each shipment is checked for loadfill percentage Database Local database ---揃揃揃揃揃----- '" ..._ 揃-揃 00WMt 0F 5H PM<N1'1W<D<LY] .:鐃 .. .. - ---- "" Olllo, >l<IPMeNn t'N""'' 鐃緒申- ---- If smaller vehicle available, consider weight and volume If found, highlight and calculate the percentage difference in cost Method 2: Truck Consolidation - Initial toolkit used Excel VBA to attain graphical trends but this was felt to be too slow in producing required results and further development would be hindered Data Manipulation DataTransition DataVisualization Based on the design values, the programming language and toolkit was then narrowed down. JavaScript Criteria Evaluation (1) Parameters 1. Temperature Conditions 2. Transport Scope 3. Transport Mode Identify consecutive shipments of same description based on user input Check for suitable truck combination to hold total loadfill Cost and shipment level dataset from Unilever SAP Comma-separated value file conversion C) fython Cleaning of raw dataset Sanitisation of non-conforming sets Cost profile calculations Apply cost savings techniques C) Output from python in form of comma-separated value file Analytics extractions into JavaScript C) 財 JavaScript Interactive Dashboard with multiple criteria filter Able to deep dive into individual shipments for investigation C) 4. Sanitized Data (2) Processing Time Approximately 30 seconds per click per criterion Compare new cost against originial sum. If lower, compute the cost saving The toolkit works in a single pass format by passing all the input data computations from Python into Javascript for dashboard display DASHBOARD IMPACT OF AUTOMATED TOOLKIT PRODUCT IMPLEMENTATION Day-of-weekJ and Month-of-year Selection Cost Sevings Dela Table Date Ran11e: 31/01/2017 to 30/04/201鐃 33,500sel8'Cted out of33,500 records I Rwlll.Ai L Data Records Report can now be generated completely within minutes compared to a few hours previously U...Flloflo:a,mr.,:11,10鐃盾 30.600-..00UI.........,_,_,.. Da-;-<Jf-Wee,: Profile ShiD11ents Overve,i 0 ,,ooo <,000 "-''" 0,000 10.000 12,000 o 1,om e,ooo ,,om ,,ooo ,,ooa ,..o, """" ..,,o ..... "'"'"" ...,,, ""''属 "'"'"" ApO> ""'"' ,..,. """" "' Dynamic Pie Chart 7 "-"""""' Parameters L_ Mode ol Tfal3P(lt Dr"!! to a.e1,,ct 1ime dur"1ion Terr鐃ti..re 0.:n:mlon Criteria Selection Interface Final product design takes in data input from the user and enables dynamic criteria selection which computes the cost saving based on the the selection concurrently KEY SKILLSETS Human Factors Engineering (HFE) principles were applied to improve the visualization of the model Statistics knowledge and data analytics skills were applied to interpret and evaluate the significance of the data Engineering communications and human resourcemanagement wereadopted to facilitate the interaction with stakeholders and team Software engineering techniques used to enable automation of toolkit LIMITATIONS J- End-user has to learn python programming for effectiveuse and future modifications of the new dashboard reporting tool !- Certain components of the dashboard tool require manual adjustments RECOMMENDATIONS t+ Standardization of input data to reduce need for manual adjustments t+ Python and Javascript programming for end-user as well as documentation for code ., ! --h.d<-- ,. ,- DAS!-IBOARD L Dynamic Date Selection This increases productivity and allows the company to focus on mitigating cost savings in the shipment logistics rather l. . . . ,. . . . ;:揃1 1 揃1' 揃1'.... Percentage Weight Programming Logic ' . ' ' ' ' ., : "" - .;. ... ... ---鐃- - : 揃揃揃n揃揃i揃揃a揃揃i揃 I鐃:i,_u 揃,.揃揃=-(e=.i: - l_ .:.. 鐃..-, ,,. .:.. ---- F i: -鐃渋 . . . ..' - - l.. Based on the output charts and cost savings summary, identify trends and potentialcost saving allocations from the cost saving methods j≒1'揃揃1'揃揃揃.揃 l. . . . ,. . . . . ' ': . ', ' ' 1i 揃-lll ' ' ' . ' ' C:W<>"揃--- ----- I Month-of-year/ Day-of-week Chart 1. Shipment Profile 4, Volume Profile 2. Weight Profile 5, Truck Size Profile 3. Loadfill Profile I Cost Savings - Before and After 1. Daily Truck Size 2. MonthlyTruck Size 3. Percentage Savings f+ Currrently adopting an agile methodology framework for this proJect Future work can focus on using a waterfall programming methodology framework to scale up the dashboard 鐃 Waterfall software 鐃 development 鐃 cycle 鐃 Q]llll9 .. -=-- Cost Savings Graphs and Charts Interface VALIDATION OF PRODUCT TOOLKIT Criterion Practicality Workability Accessibility Flexibility Description Dashboard is deemed a practical visualization tool by the company. since visuals are easy to comprehend and computation of required information is quick and efficient Dashboard displays results in an ergonomic manner and allows user to easily modify the layout according to specific analysis needs Dashboard requires less time and cost to gain access than conventional methods since data extraction is separate from the analytics toolkit Dashboard allows for further development to incude additional features to be included D,:o_,,,,.,,.,,.,-,,,,, 鐃-""属-""'"'"'-""",_I_"' Remark *** Data shown on this poster are fabricated for demonstration purposes only.