Best way to improve is often to find what others are doing better. Sometime ago, bench-marking used to be very popular tool for improvement. In bench-marking we often compare performance of different industries or different units to find how the units are performing and identify scope for further improvements. When we do this sort of comparisons common complaint that one gets to hear “our unit is different from the other one.”
There is some truth in these complaint. No two unit or two industry is same. Even for the same unit, environmental factors will be different during different periods. Direct comparison of parameters between two different entities are more like comparing between apples and pineapples. There has to be a way out to solve this.
Data envelopment analysis (DEA), occasionally called frontier analysis, was forwarded by Charnes, Cooper and Rhodes in 1978 to overcome limitation of such benchmark comparisons. DEA is used for evaluating the relative efficiency of decision making units (DMU’s) in organisations. Here a DMU is a distinct unit within an organisation that has flexibility with respect to some of the decisions it makes, but not necessarily complete freedom with respect to these decisions. DEA provide us a powerful tool to find most practical path for improvements. It can also be used to set internal improvement targets in rational manner.
DEA has been used successfully in evaluating efficiency of oil-companies’ branches, estimation of bank branch performances, measuring efficiency of insurance companies, analysis of quality-price ratio of cars sold, rating of fitness centers, rating of fast-food chains, efficiency analysis of retail chain shops, analysis of range of goods in shop, regional performance evaluation.
To understand how DEA works we shall discuss performance of some select companies in Indian automobile sector. The industries include OEM as well as its ancillaries. All the industries perform very similar operation but vary widely in their input condition. To get first level of parity we have considered value addition within the industry as measure of output. As for inputs we consider labor, energy, capital base and land. As direct measure for these are not available in public domain we have used published financial figures, with an assumption that within same industrial sector there is a parity between financial figures and real inputs.
When we use financial figure, there is a possibility of some figures being reported in different accounting periods. These error are due to differences in accounting practices between the companies and also changes in accounting practices due to strategic reasons. Such details do not fall in the preview of stochastic variation. We have taken average of past five year period (Average of 2011-12 to 2015-16) to reduce such random year to year variation in performance reporting. In order to maintain confidentiality, Industry names are replaced with identity code. Average values of different parameters for the industries are given below.
In this mix of different industries there are various levels of processing strategy. Some industry prefer to outsource majority of operations, while some manufacture almost every part in-house. When we compare manufacturing operation we have used value addition in place of turnover as the measure of output. The figure is obtained as: Value Addition = Turnover – Material Consumed
Plotting the data in a graph does give us some idea on how the industries perform. Let us take a look at it for starter.
We can see the ratios differ between the industries as well. While Industry “F” manages its operations with just 9 days of inventory we have industry “S” that keeps a luxurious 121 days of inventory. In terms of operating costs industry “G” stands out as it consumes almost all money it can get as value addition in running its operation. In the other extreme we have industry “C” that runs its operation at 30% of its value addition. Even here we can find industry “A” and “G” that uses comparatively more in terms of administrative cost.
In terms of capital productivity we find again industry “G” to be at worst performer with about 17% value addition on its installed capital. Industry “C” performs excellent on this aspect. It is able to add value to an extent of 251% of its installed capacity value.
Value addition in terms of turnover give us indication on innovation and scale of operation. Industries in niche market will be able to realize more value addition on their raw material. Similarly an industry with more in-house component will add more value in comparison to someone who operates more with outsorced operations. Here we see “S” adding maximum value at 66% and company “G” with minimum value addition at 18%.
While these parametric comparison give us some indication on the good or bad parameter we can better understanding from same data using DEA.
I have taken a 2 parameter example to explain working of DEA.
We have plotted Value Addition / Gross Block vs Inventory Turns in this graph. In the plot, we can see some industries that are clearly on the boundary and a lot of other industries are scattered within the envelop. Basically DEA help us to identify these outliners, which are clearly excellent in some parameter or the other. Apart from identifying the out liners we can also find a comparison on what can be a practical and rational target. For example for Industry D we can say it is clearly has a opportunity to improve its operation to become something similar to Industry C, or rather till the point where the extended line from origin intersects the outline envelop.
Doing this in a two dimensional graph is easy, but when we want to compare more parameters we need to apply some mathematical manipulations to get these out-liners and intersects. I would not bore you with details of such manipulation. You may find them in my book. Let us see result of the DEA :
Here there are two categories – OEM and Ancillaries. BM stands for the best industries, CBM stands for best one within its own category. Industries that are not the benchmark can improve their performance. DEA compares the inputs and indicate their target. Here are the targets:
- DEA help us to compare DMUs that are widely different from its operations
- DEA considers aggregated tendencies of DMUs
- DEA provide us rational targets for improvements
- DEA can handle multiple output-input relationships easily
- DEA does away with personal judgments in terms of weightages
Often we follow-up DEA findings with operation level benchmarking after finding the outliners and gap in performance.