Improve capacity utilization planning for Maruti by forecasting future demand

Project Details




Kumar Mukul, Anuj Kumar, Ritesh S Rao, Mihirr Sose, Liza Mohanty, Hridhay Rajkumar





Problem description
Improve capacity utilization of Maruti’s Manesar and Gurgaon plants by forecasting
future demand of Maruti cars and hence scheduling production.
Maruti has 2 manufacturing plants at Manesar and Gurgaon. Manesar plant has a capacity of
550k and Gurgaon plant has a capacity of 900k as of 2016. The production numbers for 2016
shows that Manesar plant produced 630k cars suggesting overtime at the plant, whereas
Gurgaon plant produced 678k cars only suggesting underutilization.
The problem that we are trying to solve is to figure out where capacity augmentation can be
done to optimize the capacity utilization based on forecasts for various models.

Data Description
Source: Sales Data obtained from annual report
We collected data for 56 months starting from May 2012 for 18 models of Maruti cars and
aggregated them based on the manufacturing plant they are being produced at, to get Manesar
and Gurgaon plant data. Some models were discontinued in between and some were launched
in between this period.

Key Characteristics of the Data
Trend – Linear trend can be observed in the production at Manesar plant, whereas Gurgaon
has a quadratic trend and the models (Dzire, Swift and Celerio) have varying trends. In case of
Dzire and Swift, a downward trend can be observed in the last 12months of the data.
Seasonality – No observable seasonality in all the data that we have.
Level and noise are always there.
Other – The month of August 2012, has a significant dip across both the plants because of a
agitation in the plants. While considering data for forecasting we have used the average of the
other August months for August 2012 since this data was an anomaly. This period is circled in
the graph above and going forward.
• Forecast demand for next 12 months at Gurgaon and Manesar plant
• Forecast individual demand of Dzire (Monthly and Quarterly) and Swift models which
contribute to 80% of cars produced in Manesar plant
• Forecast demand of a 3rd model which can be shifted to Gurgaon plant to accommodate
Manesar’s production plant for next 12 months
• Models used
● Naïve forecasts (Lag 1, Lag 3 and Lag 12)
● Multiple Linear Regression
● Holts Winter
● Double exponential smoothing, data deseasonalized using CMA of 12 and
seasonality indices
Conclusion and Recommendations
Based on our forecasts, capacity at Manesar plant will be extremely strained over the next year
● While Swift demand is 185,936 over the next year, Dzire demand is 225,124. These
two combined capture 74% of regular capacity at Manesar plant
● Celerio demand is 90,481 cars out of which 54,368 can be accommodated in regular
capacity of Gurgaon plant while rest can be managed at OT. This would enable Maruti to
maintain OT level in 2017 at Manesar plant similar to 2016 level
1. Shift complete production of Celerio to Gurgaon plant since shifting production of Swift
or Dzire (both produced in high volume) partially will have impact on economies of scale
achieved in sourcing
2. Initiate capacity augmentation at both plants, at Manesar to deal with demand in Swift,
Dzire and in Gurgaon for other models.

Application Area: