Teaching Note

Farmaid Tractors Limited

Abstract
FarmAid Tractors Limited (FTL) is a tractor company with a 20% market share and aiming
to be the market leader within five years. The tractor industry in India has become very
competitive, with growth in capacity outstripping the growth in demand. Customer
preferences and demands have changed in the context of the competitive environment. The
role of the dealer and servicing the dealer by the company are recognised by FTL as key
success variables in this business. The key issue of concern is to develop a partnership with
the dealers by bringing in a service focus, through improvements in supply chain
management and outbound logistics. Two critical areas for analysis are to be taken up,
namely (i) order processing and inventory planning and (ii) distribution structure including
stockyard location. The issue that is taken up specifically is of determining optimal
stockyard location for states in India. The state of Gujarat has been given as an example for
analysis.

Questions for Discussion

1. What are the factors leading to the perceived 'lack of control' and poor delivery
quality?
2. What forecasting technique should be used for inventory planning at the plant and
the stockyards?
3. Is there a need for a central dispatch yard? What are the pros and cons?
4. What is a good model for determining optimal location of stockyards and the
associated allocation of dealers to the stockyards in the state of Gujarat?
5. Interpret the implications of the recommended locations versus the existing
locations as given in Exhibit 11.

Analysis
Lack of Control and Poor Delivery Quality
The present scenario, especially the distribution system needs a detailed discussion
regarding the pros and cons. Currently, all despatches were made from the factory to the
stockyards through a daily allocation process that took into account the unmet demand at
stockyards, ready for despatch inventory and availability of the special-purpose trucks
based on the transporters inputs. Once assigned to a transporter, they moved the tractors
to their godowns, since there was no space in the factory for holding finished stocks.
Thus the tractor no longer remained under the control of the company. It was often
noticed that the transporters actually moved the tractors out of their godowns after an
average of two days, primarily because of the non-availability of the intended trucks for
despatch. Transporters tried to hedge against under-utilization of their special-purpose
1

trucks by reporting availability of trucks ahead of when they actually were. This not only
results in additional inventory carrying cost, but also resulted in additional loss of control
over the tractors on the part of the company. More over the percentage of tractors
damaged during transportation is also questionable. Transit/storage resulting in
repair/replacement/replenishment (70% of tractors received a 'yellow' card on receipt at
dealers - implying not ready for sale. 75% of these were set right in the first week. The
remaining sometimes got complicated in 'investigations', resulting in non settlement of
claims/dues even upto four years).
Forecasting Technique for Inventory Planning
The key concerns in inventory planning were to enable high service levels to the dealers,
and at the plant to respond to seasonality
Forecasting:
Forecasting can be done at three levels as follows:
(i) Company level:
This should enable aggregate and seasonality planning.
(ii) Regional office level:
This is required for cross-checking the periodic consolidated dealer forecasts for
placing orders from the factory. An important factor is choosing a suitable
established forecasting model and to validating the model using standard
techniques like Root Mean Squared Error.
(iii) Dealer level:
This should result in enabling regions to position inventory in the stockyards and
place orders from the factory. At the dealer level, developing models for the
disaggregate level of forecasting would have been difficult. However, it became
clear that tracking of potential customers would be a reasonably robust
mechanism for assessing demand since a customer went through many predictable
stages of the buying process, before finally purchasing a tractor. The model would
also have to incorporate marketing decisions.
Inventory Planning:
The monthwise sales at an aggregate level could be forecasted with a high level of
accuracy. Given an average sales of 5000 tractors per month, the lowest sales of 4000
tractors were expected in February and the highest sales of 6000 tractors in April. A
policy of production plan can be evaluated that followed the demand versus one that
would have a uniform monthly production. The demand-driven production plan would
have no inventory because of seasonality, although there would be costs because of
overtime, subcontracting, supervision, etc.; the uniform production plan would have
inventory costs because of seasonality. It was easier to assess the inventory costs of the

uniform production plan rather than the costs of the demand-driven production plan.
Exhibit TN-1 (also refer Exhibit 5 of case) gives the inventory because of seasonality
under the context of a uniform production plan. The seasonality inventory would clear
towards the end of June, which is at the end of the AprilMayJune peak. The overall
inventory cost because of seasonality was Rs 47.4 million, which amounted to Rs 790 per
tractor. It was the managements judgement that the cost because of the demand-driven
production plan would be higher, especially since it involved overtime, which often had
long-term fixed cost implications. Consequently, no attempt was made to quantify the
costs because of the demand-driven production plan.
Central Dispatch Yard
The problem of extended period of 'lack of control' and poor delivery quality could be
solved by a central despatch yard. As there was no space adjacent to or near the existing
plant for such expansion, the possible location of the yard was at a suitable highway
location, 20 km away from the plant.
The analysis for the economics of a central despatch yard indicates that at an inventory
saving of two days per tractor, the annual savings would be Rs 12 million. The annual
operating cost would be Rs 2 million (including the additional transportation cost to move
the tractors to the central despatch yard rather than the payment to the transporters to
move the tractors to their godowns), thus offering a net saving of Rs 10 million per year.
This was very good compared with the investment cost of Rs 15 million. There were
issues as to whether the two days would be entirely saved, just because the allocation
would now be made after physically seeing the truck that would transport the tractors.
The qualitative benefit of the increased flexibility in allocation and reduction in losses
because of being able to inspect the transporting truck were considered significant. It was
also felt that the transporters would welcome this move, since they would save on the
storage space in their godowns, while of course giving up the margin on the payment for
moving the tractors to their godowns.

Mathematical Programming Model for Stockyard Location Analysis

Stockyard location analysis can be effectively considered as a typical linear programming
transportation problem. Transportation problem deals with optimal transportation and
allocation of resources where there are sources with a supply of some commodity is
available and destinations where the commodity is demanded.
The application of this model for the state of Gujarat is illustrated. Five potential
locations for stockyards were considered (the current stockyard location was
Ahmedabad). The considerations for a candidate location were (i) the availability of
secondary transportation and (ii) proximity to dealers so that service levels to the dealers
could be maintained. The stockyard management would be outsourced to a third party, as
was the practice in the industry. Further, the operating costs were governed mostly by the
extent of land required for parking the tractors and the real estate costs. The volume of

throughput did not influence the cost structure (although such cost structures can be
negotiated), since volumes were expected to be at levels where the minimum manning at
the stockyards would suffice. Data are provided for the five potential stockyard locations,
monthly operating costs (as specified by the third party) and distance from factory
(Exhibit TN-2, also refer Exhibit 6 of case), and location of the 19 dealers of the
company in Gujarat, along with the expected monthly demand and distances from the
potential stockyards (Exhibit TN-3, also refer Exhibit 7 of case). The total Gujarat
demand was expected to be 500 tractors per month.
The mathematical programming model for Gujarat had five zero-one variables to decide
on the stockyard locations and 95 zero-one variables to decide on dealer stockyard
assignment. The objective function optimized the total relevant cost consisting of the
primary and secondary transportation costs and the stockyard operating costs. There were
19 constraints to ensure that each dealer was assigned to a stockyard, and five constraints
to ensure that a stockyard was open, if required for being assigned to a stockyard. There
could be a few additional constraints, depending on stockyard capacities, minimum
throughput volumes (for outsourcing) and limitations for control, etc. Since this model
facilitated a tactical decision, it would be run whenever there were significant changes in
(i) demand patterns within a state, (ii) stockyard location costs or (iii) ability to service
the dealers with appropriate service levels. In general, this was not expected to occur
within say two years.
Mathematical programming model
Variables
i = 1, 2, . . . ,
j = 1, 2, . . . ,
Yi
=1
=0
Yij
=1
=0

s potential stockyard locations (s is typically 45 in a state),

n dealer locations ( j is typically 1520 in a state).
if stockyard i is selected
otherwise
if dealer j is served via stockyard i
otherwise

Inputs
pi = primary transport cost per tractor km from factory to stockyard i
sij = secondary transport cost per tractor km from stockyard i to dealer j
di = distance from factory to stockyard i
dij = distance from stockyard i to dealer j
Dj = demand at dealer j per month
Ci = cost of operating stockyard i per month
M = large number
mi = minimum number of dealers required to be handled by stockyard i
Mini = minimum volume required to be handled at stockyard i
Maxi = maximum volume that can be handled at stockyard i

Formulation of zero-one IP
Min (pidi + sijdij) Dj Yij + CiYi
Yij = 1

for every j

(Every dealer must be assigned one stockyard)

Yij < M Yi

for every i

(A stockyard must be opened to be assigned)

If no more than 2 stockyards should be selected, we can limit

Y1 + Y2 + ... Ys < 2
For minimum number of dealers per stockyard, say mi,
Yi1 + Yi2 + ....... + Yin > mi
For minimum volume by a stockyard, say vi
D1Yi1 + D2Yi2 + ...... + DnYin > vi
We can also do maximum dealers, volume, with <

Exhibit TN-4 (also refer Exhibit 10 of case) gives the results (optimal stockyard locations
with total costs) of the analysis based on the mathematical programming model for
various scenarios affecting the stockyard location. Ten scenarios were considered on the
dimensions of cost structure (current and future), secondary distance limit (none, 350 and
500 km), and the minimum number of tractors to be handled at a stockyard (0 and 200).
The minimum number of tractors was taken at zero for the case of the most restrictive
secondary distance limit of 350 km. For each of the cost structures, the existing location
scenario was also analysed. The model also gave the allocation of dealers to stockyards
for these scenarios.

Implication of the Recommendations

The model recommended relocation of the stockyard from Ahmedabad to Valsad, for
example for the state of Gujrat. The total relevant cost for the existing location was about
Rs 1,120,000 per month. A saving of nearly Rs 300,000 per month was possible by
relocating the stockyard to Valsad. With a secondary distance limit of 500 km, the
possible saving was about Rs 230,000 per month and with a limit of 350 km, about Rs
180,000 per month. It should be noted that for the current cost structure, the selection of
Valsad by the model as the location or one of the locations (depending on the restriction
on the secondary distance) is trivial because of the triangular inequality, since Valsad is at
the entry point to the state on the highway from the factory. With the expected future cost
structure, the total relevant cost for the existing stockyard location was about Rs
1,030,000 per month (an automatic saving of Rs 90,000 per month because of cost
structure alone). A further saving of about Rs 150,000 per month was possible if the
stockyards could be located elsewhere, even with the additional constraints for service

level. Of the scenarios considered, the locations at Valsad and Ahmedabad were
preferred. This was also driven by (i) the convenience of retaining the existing location
and (ii) expected opportunities for growth in the markets near Valsad.
When similar models were run for other states, the recommendations yielded a total
saving, across the states where stockyard locations were revised, of about Rs 1 million
per month, i.e. Rs 12 million per year. The final recommendations for stockyard locations
of the major states, based on the model output and implications in terms of the criteria
considered, are given in Exhibit 5. The model indicates a shift in the location of the
stockyards and the number stockyards. The locations are to be shifted towards Mumbai
and the number of stockyards should also be increased. This indicates that there is greater
emphasis on transportation costs rather than warehousing costs.
Apart from the specific recommendations, one of the greatest benefits of the modeling
exercise was in convincing the organization that a variety of factors can be considered for
analysis, often leading to counterintuitive solutions. Also, the scenario analysis
demonstration prompted the in-company logistics team to carry out a sensitivity analysis
by examining marginal violations of desirable parameter values by considering more
scenarios by varying parameter values.

Exhibit TN 1: Inventory because of seasonality, with uniform production policy

Month

Demand

Production

January
February
March
April
May
June
July
August
September
October
November
December
Total
Average
per month

5,000
4,000
4,500
6,000
5,900
5,700
4,500
4,000
4,500
5,500
5,400
5,000
60,000
5000

5,000
5,000
5,000
5,000
5,000
5,000
5,000
5,000
5,000
5,000
5,000
5,000
60,000
5000

Inventory due to
uniform production
1,100
2,100
2,600
1,600
700
0
500
1,500
2,000
1,500
1,100
1,100
15,800
1317

Inventory cost (at Rs 200,000 per tractor and 18% per annum inventory carrying cost)
1317 x 200,000 x 0.185 = Rs 47.4 million per annum

Exhibit TN 2: Inventory Because of Seasonality, with Uniform Production policy

Potential stockyard locations (i), operating cost (Ci) and distance from factory (di)
Sr No Stockyard
Location
1
2
3
4
5

Valsad
Surat
Vadodara
Ahmedabad
Rajkot

Operating
Cost
per
Month (Rs)
25,000
20,000
30,000
30,000
25,000

Distance from
Thane (Kms)
136
263
448
545
761

Exhibit TN-3: Dealer Location, Demand and Distance

30

25

40

25

20

20

20

35

No of Tractors/
Month

35

20

1 Valsad

633 272 62

163 514 32

2 Surat

566 205 31

96

3 Vadodara

399 38

125 71

4 Ahmedabad

258 73

225 182 200 377 40

5 Rajkot

105 255 492 365 175 560 162 321 304 88

30

45

19

Surendrana
gar

Rajpipla

Porbandar

18

25

20

385 315 424 616 630 293 419 647 491

456

715

204

431

447 109 318 248 357 549 563 226 352 570 424

389

648

141

364

280 266 151 81

185 403 257

238

481

82

200

74

125

412

195

116

255

187

255

111

313 327 52

20

17

Patan

Palanpur

16

20

136 79

20

15

30

190 382 396 59

20

14

Morbi

13

Mehsana

12

Nadiad

11

Junagadh

10

Jamnagar

Himmatnag
ar

Godhara

Dholka

Dharampur

Bhavnagar

Bharuch

Bardoli

Anand

Dealer Location

Amreli

292 146

102 234 299 67

371

* All distances are in Kilometers.

Exhibit TN 4: Scenario Analysis for Gujarat: Total Relevant Cost and Stockyard Sites

(Rs)

Cost/tractor/km

Primary: 2.5
Secondary: 3.5

Primary: 3.0
Secondary: 3.0

Current

Secondary
Distance
limit:
None

10,28,999

8,73,533

8,78,209

8,75,454

Secondary
distance limit:
None
Minimum no of tractors to be
serviced by a stockyard:
200/month
8,75,484

Ahmedabad

Valsad
Rajkot

Valsad
Ahmedabad

Valsad
Ahmedabad

Valsad
Ahmedabad

11,19,855

8,22,880

Valsad
Ahmedabad
Rajkot
9,43,085

8,87,380

8,22,800

8,99,080

Ahmedabad

Valsad

Valsad
Ahmedabad
Rajkot

Valsad
Vadodara

Valsad

Valsad
Vadodara

Secondary
Secondary
distance limit: distance limit:
350 kms
500 kms

Secondary
distance limit:
500 kms
Min. no of tractors to be serviced by
a stockyard: 200/month
875,484

Exhibit TN 5: Recommendations for Stockyard Locations

State

Existing Yard(s)

Andhra Pradesh

Hyderabad

Tamil Nadu

Chennai

Karnataka

Bangalore

Gujarat

Ahmedabad

Madhya Pradesh

Bhopal

Rajasthan

Jaipur
SriGanganagar

Punjab
Haryana

Jalandhar
Karnal

Optimal Locations
Hyderabad
Vijaywada
Hosur
Trichy
Belgaum
Davangere
Valsad
Ahmedabad
Indore
Raipur
Kota
Jodhpur
SriGanganagar
Patiala
Gurgaon

10