In order to analyze the impact of t

In order to analyze the impact of traceability improvements on
the consequences of a safety crisis we simulated recalls in the case
study. To do this, we developed a spreadsheet simulation model
implemented in Visual Basic for Microsoft Excel. These types of
simulation models are often preferable because of the software
usability and availability (Akkerman & van Donk, 2008). The
simulation model is based on a series of parameters that are classified
as constant (C), predetermined (P) or uncertain (U). Constant
parameters refer to parameters that remain unchanged, predetermined
to those that can actively be changed in the model, and
uncertain to unknown parameters (see also Akkerman & van Donk,
2010). For the unknown parameters, a probability distribution is
typically defined and the simulation model randomly chooses
a value for each simulation run. All data used for developing the
simulation model and their classification are shown in Table 1. It
should be noted that wherever a range of values is denoted,
a uniform distribution is used in the simulation.
In the chocolate production process the roasting process
represents a key step for improving the microbiological condition
as well as for defining the aroma profile of the final product (de
Muijnck, 2005). Safety and quality of the finished chocolate
strongly depend on the roasting process. Each batch of cocoa beans
received by the chocolate manufacturer is usually split into several
processing batches, whose dimension depends on the capacity of
the roasting equipment. Thus, each processing batch goes into
a specific roasting process. In this way, if a problem occurs to one
roasting process, only the finished chocolate produced with that
specific roasting process will suffer the consequences of the
problem. The processing batch size is therefore a key (predetermined)
parameter in the simulation model, and an essential
planning decision in practice.
3.3. Experimental design
3.3.1. Different production strategies
The simulation model was designed to simulate the chocolate
production system for two different production strategies, one
based on production efficiency (PS1) and one based on reduced
batch dispersion (PS2). In PS1 the maximum processing batch size
is always used so that the equipment in the production stage is
always used at full capacity. Since the size of the cocoa bean batches
delivered to the chocolate manufacturer is not necessarily
a multiple of the processing batch size, some cocoa beans are mixed
with the next batch of cocoa beans. This results in having some
batches of finished product produced from two different batches of
raw materials. Instead, PS2 focuses on reducing batch dispersion,
where the chocolate manufacturer avoids mixing the different
batches of cocoa beans. Here, some processing batches might be
smaller in size. As batch processes are involved, this results in some
partially unutilized processes in the chocolate production line, with
a corresponding reduction in production efficiency. On the other
hand, if a safety crisis occurs to a batch of raw materials, a PS2
production strategy would lead to smaller recall sizes compared to
PS1. A graphical illustration of both PS1 and PS2 can be seen in
Fig. 2.
In the remainder of this paper, production efficiency is
measured by the number of processing batches because:
 The number of processing batches equals the number of times
a roasting process is performed and the duration of the roasting
process depends on the roasting grade desired, not on the
amount of nibs processed into the equipment (Jinap, Rosli,
Russly, & Nordin, 1998; de Muijnck, 2005). Therefore less
processing batches mean less time needed for roasting, with
a constant number of equipments; or less equipments needed,
with a constant processing time required. Thus, less processing
batches lead to a higher efficiency.
 Smaller batch sizes (also meaning more batches when processing
a constant raw material amount) were found by other
authors to lead to an increase in production setup times and
costs, resulting in losses of production efficiency (Dabbene &
Gay, 2011; Dupuy et al., 2005; Rong & Grunow, 2010; Wang,
Li, & O’Brien, 2009).