Obviously, the recall size in case

Obviously, the recall size in case of contamination of
a processing batch (R2) strongly depends on the processing batch
size. Thus, this variable unutilized fraction of the process results in
a variable difference in recall size between PS1 and PS2, which is
reflected in the peaks present in the R2 results curve in Fig. 4. The
recall size in case of contamination of the raw material (R1) is also
affected by this variable difference, but since this type of recall
consists of multiple batches of finished product, this has a relatively
small influence on the results, and cannot be seen in Fig. 4.
Overall, Fig. 4 shows that for all processing batch sizes the
relative difference in production efficiency ranges between 7 and
22%, the relative difference in recall size for R1 between 0.5 and 4%,
and for R2 between 6 and 16%. In general, when comparing PS1 to
PS2, the difference of both production efficiency and recall sizes
between PS1 and PS2 increases when increasing processing batch
sizes, due to the decreasing utilization and the increased size of
mixed batches.
When comparing the benefits in increased production efficiency
of PS1 to the benefits in reduced recall sizes of PS2, we did not
include the probability that a product recall occurs. A product recall
obviously does not occur for every batch produced, while benefits
in production efficiency influence all produced batches. Therefore,
when deciding whether adopting a new production strategy would
be beneficial, the probability that a product recall occurs must be
assessed and included in the decision-making process. However,
even without this probability, the results obtained in this paper
show that the effects in terms of production efficiency in PS1 are
significant. This suggests that adopting a production strategy
focused on reduced batch dispersion (such as our PS2 strategy)
might often not be economically feasible. However, it should also
be taken into account that the efficiency losses and the amount of
lost product are only part of the economical effects. It is well known
that in case a product recall occurs, food companies also incur costs
related to organizing the recall, bad publicity and damage to the
reputation of the brand (Onyango et al. 2007). If these aspects were
quantified and taken into account, reducing recall sizes would have
a bigger impact and PS2 might also become economically feasible.
Quantifying the losses caused by bad publicity and damage to the
reputation of a brand in case of product recall is outside the scope of
our paper.
4.2. Influence of different traceability systems on product recall size
When constructing the case study it was seen that currently no
full traceability is in use in the supply chain of chocolate. This is
probably due to the fact that cocoa is farmed in non-European
countries, where there is no legislation obliging the actors
involved in a food supply chain to have traceability systems. The
European legislation states that within Europe, food companies
must be able to identify immediate suppliers and customers of
a specific product (European Commission, 2002). This legislation
guarantees full traceability within the European borders, but
traceability is lost when part of the supply chain goes outside these
borders. In the example case study constructed in this paper it was
seen that cocoa could be traced up to the local exporter, which,
since located outside Europe, is not obligated to comply with the
European law on traceability and therefore does not record any
further traceability information.
In our study the benefits of having a traceability system that is
more accurate than what the lawrequires are quantified in terms of
reduction of product recalls. In order to propose realistic
improvements of the traceability system of the case study supply
chain, TSþ and TSþþ were designed taking into consideration that
most of the cocoa farmers would not have the resources to invest in
technologies, therefore these improvements imply very low
investments. In order to analyze the influence of these different
traceability systems (TS0, TSþ and TSþþ) on the recall size of the
finished product in case of contamination of the raw materials, we
again used the simulation tool. In order to have an accurate result
the simulation model has been run 100 times for each processing
batch size and data are shown in Fig. 6 as the average of the averages
of the results of all roasting equipment sizes (kg).
The results show an average recall size of 1,608,719 units of
product for TS0, an average recall size of 714,584 units of product
for TSþ and an average recall size of 55,789 units of product for
TSþþ. No standard deviation is shown in the Fig. 6 because it would
show a variation which is mainly caused by the different roasting
equipment sizes, thus giving a false sense of variability.