root cause analysis and exploration of unplanned costs in ... · tmc is a division of c.h. robinson...
TRANSCRIPT
Root Cause Analysis and Exploration of Unplanned Costs in Truckload TransportationCapstone Advisor : Dr. Chris CapliceCapstone Partner: C.H.RobinsonTeam members: Nishitha Aemireddy, Xiyang Yuan
Background
Motivation
Methodology
Model Results
Business Insights
Further Research
2
Agenda
TMC is a division of C.H. Robinson that offers Managed TMS®, a unique combination of global transportation management system (TMS) software, logistics process management, and consulting services.
3
Operated as an independent division of C.H.Robinson since 1999
Offers managed TMS®
Operates from 5 different locations across globe
10.4 Million shipments $3.1b in Freight under management
Background – TMC
Tender to Primary Carrier
Tender to Backup
Carriers in Routing guide
Auction in Spot Market
Primary tender
Backup tender
Spot tender
AcceptAccept Accept
Reject Reject
Reject
Background – Load Tendering Process
TMC uses routing guides for their load tendering. Loads are tendered in sequential order where the sequence of carriers is determined based on prices, performance levels and capacities.
Tender rejections by primary carrier increased from 19% in 2015-16 to 37% in 2017-18.
2015-16 is a soft market, 2016-17 is a transitional market and 2017-18 is a tight market
5
Motivation
80.60%73.11%
63.12%
15.87%20.77%
27.49%
3.53% 6.12% 9.39%0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
2015-16 2016-17 2017-18
Percentage distribution of loads accepted primary, backup and spot
Primary Backup Spot
Primary acceptance rate decreased by 22%
Backup acceptance rate increased by 73%
Spot acceptance rate nearly tripled
Backup premium : Average of percentage difference in backup carrier rates in routing guide and primary carrier rate over all lanes
6
Increase in premiums with Routing guide depth
Premiums paid increased from 2015 to 2018
Premiums increased at a decreasing rate with
sequence number
0%
2%
4%
6%
8%
10%
12%
14%
16%
1 2 3 4 5
% in
crea
se o
ver P
rimar
y Ca
rrie
r Rat
e
SeqNum
Percentage increase in backup premiums with sequence number
2015 - 2016 2016 - 2017 2017 - 2018
! = −0.0015() + 0.0304( + 0.0146.) = 0.9967
! = −0.0021() + 0.0368( + 0.0099.) = 0.9916
! = −0.0043() + 0.0504( + 0.0074.) = 0.9999
Spot premium : Average of percentage difference in spot rate and primary carrier rate over all lanes
7
Routing guide and Spot pricing premiums
9%12% 12%
23%
32%35%
0%5%
10%15%20%25%30%35%40%
2015 - 2016 2016 - 2017 2017 - 2018Perc
enta
ge In
crea
se o
ver P
rimar
y Ca
rrie
r Rat
eRouting Guide & Spot Percentage Premium
Routing Guide Spot
Spot premium increased significantly in 2016-17
Backup premiums did not increase as much
Not a significant increase from 2016-17
to 2017-18
The project uses 3 years of load and tender data from October 2015 to September 2018. The dataset contains only full truckload (TL) dry van shipments with a minimum length of haul of 250 miles
8
Methodology
Data collection
Data Cleansing
Data Profiling
Hypothesis generation
OLS & Logistic
Regression modeling
Business Insights
• Longer the distance, higher the cost.
• Distance has no clear impact on tender acceptance rates.
9
Data Modeling Results - Distance
200
300
400
500
600
700
800
900
1000
1100
1200
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
0
500
1000
1500
2000
2500
3000
3500Avg.CostPerLoad(Dollars)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
AcceptanceRate
ImpactofDistance
MeasuresAvg.CPL
AcceptanceRatebyCarrierCategoryPrimaryCarrier BackupCarrier Spot
Distance(Miles)
2015-2016 2016-2017 2017-2018
Low
Volume
Medium
Volume
High
Volume
Low
Volume
Medium
Volume
High
Volume
Low
Volume
Medium
Volume
High
Volume
0
200
400
600
800
1000
1200
1400
1600
Avg.CostPerLoad(Dollars)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
PrimaryAcceptanceRate
1,356
1,161
1,048
1,360
1,134
984
1,573
1,276
1,086
ImpactofCorridorVolume
PrimaryAcceptanceRate Avg.CPL
• Corridor volume: Monthly average volume for each 3-digit to 3-digit zip code combination.
• Higher the corridor volume, lower the cost per load.
• Increasing corridor volume increases primary carrier acceptance rate and reduces the chance of going to spot market.
10
Data Modeling Results – Corridor Volume
Soft MarketDifference: 23%
Tight MarketDifference: 45%
ConsistencyFactor
0.0-0.2 0.2-0.4 0.4-0.6 >0.6
HighConsistency
MediumConsistency
LowConsistency
0
500
1000
1500
Avg.CPL(Dollars)
0
500
1000
1500
Avg.CPL(Dollars)
0
500
1000
1500
Avg.CPL(Dollars)
0.6
0.7
0.8
0.9
PrimaryAcceptanceRate
0.6
0.7
0.8
0.9
PrimaryAcceptanceRate
0.6
0.7
0.8
0.9
PrimaryAcceptanceRate
0.850.81
0.96
0.78
0.74
0.79
0.69
0.81
0.64
0.63
0.66
0.62
ImpactofLaneConsistencyandVolatility
PrimaryAcceptanceRate Avg.CPL
CV
• Lane Consistency: Number of weeks that a lane has at least one load in a year
• Lane Volatility: Coefficient of Variation (CV) of the weekly volume on a lane when volume is present
11
Data Modeling Results – Lane Consistency & Volatility
↑ Consistency↓ Cost Per Load
↑ Primary Acceptance Rate
↑ Volatility↑ Cost Per Load
↓ Primary Acceptance Rate
The impact of volatility is more profound in high consistency lanes.
• Weekend: Friday, Saturday, Sunday• Quarter Ends: Last 5 days of each quarter
12
Data Modeling Results – Weekends & Quarter Ends
Cost Per Load
Weekday $1195
Weekend$1302
Acceptance Rate
Weekday0.75
Weekend 0.73
↑ 9%
↓ 3%
Cost Per Load
None Quarter Ends $1222
Quarter Ends$1237
Acceptance Rate
None Quarter Ends 0.75
Quarter Ends0.72
↑ 1%
↓ 4%
• Lead time: the time between tendered date and pickup date• Shorter the lead time, higher the cost per load.
• Increasing lead time decreases acceptance rates and increases the likelihood of going to spot market.
Data Modeling Results – Lead Time
2015-2016 2016-2017 2017-2018
0-48hrs 48-120hrs
>120hrs 0-48hrs 48-120hrs
>120hrs 0-48hrs 48-120hrs
>120hrs
0
200
400
600
800
1000
1200
1400
Avg.CostPerLoad(Dollars)
0.0
0.2
0.4
0.6
0.8
1.0
AcceptanceRate
ImpactofLeadTime
Measure
Avg.CPL
CarrierCategory
PrimaryCarrier BackupCarrier Spot
LeadTime
Soft MarketDifference: 13%
Tight MarketDifference: 15%
• 135 distinct key market areas• If a key market area is a cheaper origin, it is likely to be an expensive destination and vice-versa.
14
Data Modeling Results – Origin & Destination
Miami, FLOrigin Value: -$439Destination Value: $270
Chicago, ILOrigin Value: $338Destination Value: -$511
y = -0.8016x - 45.004R² = 0.3894
-800
-600
-400
-200
0
200
400
600
800
-600 -400 -200 0 200 400 600 800
Dest
inat
ion
Origin
Origin & Destination Market Value
15
Data Modeling Summary
Variables Cost Per Load Primary Acceptance Routing Guide Failure
Distance ↑ - -
Corridor Volume ↓ ↑ ↓
Lane Consistency ↓ ↑ ↓
Lane Volatility ↑ ↓ ↑
Weekends ↑ ↓ ↑
Quarter Ends ↑ ↓ ↑
Lead Time ↓ ↓ ↓
Regions Various
16
Managerial Implications
Focus attention on low volume corridors
Improve lane consistency and control volatility
Avoid weekends and quarter ends shipments
Consider regional values when constructing new networks
Increase lead time
17
Further Research
• Lead time: carriers’ capability of managing short lead time loads• Market index: impact on tender acceptance rates• Seasonality in demand• Network design with regional sensitivity
Suggestions
• The impact of lead time on primary tender acceptance rate• The combined effects of different factors• Regression models capture the correlation but not the causation
Limitations
Thank YouCapstone Advisor: Dr. Chris CapliceCapstone Partner: C.H.RobinsonTeam members: Nishitha Aemireddy, Xiyang Yuan