On Cost Reduction in Unilever's Supply Chain and the Interpretation and Implementation thereof

On Cost Reduction in Unilever's Supply Chain and the Interpretation and Implementation thereof

Van Zeeland, J.K.

Veeke, H.P.M. (mentor)
Loonen, M. (mentor)
Van de Loenhorst, M. (mentor)

Mechanical, Maritime and Materials Engineering

Transport Engineering and Logistics

Transport Engineering and Logistics

2015-04-16

Unilever is the market leader in the fast moving consumer goods industry. It is constantlyseeking ways to effectively use its resources. The company approached me and requested a decision tool which could determine the 'switch-over' point between single and double stacking of pallets (the so-called stacking configuration). The model was to present the decision based on the cost. Currently the optimization focus in the logistics chain is on transportation which accounts for e400 of the e640 million in the total logistics spend. Their method: optimizing the transportation by effectively using both weight and volume limits of the truck. This is done by creating either higher pallets, or double stacking the pallets. The choice for the stacking configuration is by and large reliant on the cubic density (the switch-over point is put at 400 kg=m3). However a more holistic approach to finding the switch-over point is suggested, taking into account all the individual cost post making up the total logistics chain spend. This holistic approach solves the problem of possible negative effects that changes in transportation can have elsewhere in the chain. In order to determine which factors influence the cost, a conceptual framework was put in place which addresses all the key cost posts in the logistics chain. It conceives a list of variables which influence the cost and additionally change when changing stacking configuration. These factors are the causal basis for the remainder of the analysis. The amount of influence each of these factors have on the final cost is determined by the method of higher order linear regressions. This gives the correlation or amount of co-occurrence. These regressions were made in Statistics R. The results were tested for satisfying the Gauss-Markov theorem; which tests the results for being the best linear unbiased estimator (BLUE). It was also made sure that all the variables tested were exogenous; meaning that the change in one variable can not be explained by another. At this point we had a model for which the causal basis is supported by the correlation of the variables (some show higher correlations than others, reasons for missing co-occurrence was given everywhere necessary). The model, being quite difficult to interpret, it is not easily used by Unilever employees and therefore a tool was created. The tool is made in Excel for end-user convenience. The tool additionally allows to supplement the cost function by functional constraints; both legislative, and physical. The final tool can be used to project the cost of new (or renewed) products when they are created (or changed) by 'research and development'. The tool gives the design that was used as input ('current design') and the savings that can be made by switching to 'double' or 'single' stacks. It is able to export the answers as a usable PDF format such that it can be easily implemented in Unilevers 'innovation funnel'. The first result are very promising. The model suggests that a savings of 23.9% can be made on the 48 cases under scrutiny. This means that with 95% certainty it can be said that the savings for the complete dataset is somewhere between 21.9% and 26.0%. This would yield annual savings in the logistics chain of between e141 and e160 million. To achieve these savings between 12.8% and 37.5% of the pallets should be changed to double stacking. Validation shows, however, that singular values might differ up to 50% (both positive and negative) from the data; which is attributed to the low explanatory power of the 'primary transport cost' (read: on of the) regression(s). This assumption was also validated and turned out to be true. This is, obviously, a theoretical optimum for savings, it does not account for product integrity (the quality), which was set as a constraint (it was stated that the cost could be minimized under the constraint that the quality remained at least as high as current standards). This is therefore suggested to be a topic for further research. There is some work to be done to the model before it can be taken into use by Unilever employees. It was suggested that the regressions should be re-run with all the indicated variables (which requires some additional data acquisition) and that the gathered data was time-stamped in order to account for money and oil price volatility. These actions would likely yield a tool and model where the answers are centred closer to the true cost value. For all statistical models, and this one in particular, keeping it up to date is of the utmost importance. This keeps the answers up to date and the tool usable for all new and renewed products for Unilever.

supply chain
logistics
unilever
econometric
modelling
stacking configuration

http://resolver.tudelft.nl/uuid:b9d74213-6a46-4cc5-b5ce-c1d5dbc93aee

2018-12-31

Student theses

master thesis

(c) 2015 Van Zeeland, J.K.