Options for reducing food waste by quality-controlled logistics using intelligent packaging along the supply chain

ABSTRACT

Optimising supply chain management can help to reduce food waste. This paper describes how intelligent packaging can be used to reduce food waste when used in supply chain management based on quality-controlled logistics (QCL). Intelligent packaging senses compounds in the package that correlate with the critical quality attribute of a food product. The information on the quality of each individual packaged food item that is provided by the intelligent packaging can be used for QCL. In a conceptual approach it is explained that monitoring food quality by intelligent packaging sensors makes it possible to obtain information about the variation in the quality of foods and to use a dynamic expiration date (IP-DED) on a food package. The conceptual approach is supported by quantitative data from simulations on the effect of using the information of intelligent packaging in supply chain management with the goal to reduce food waste. This simulation shows that by using the information on the quality of products that is provided by intelligent packaging, QCL can substantially reduce food waste. When QCL is combined with dynamic pricing based on the predicted expiry dates, a further waste reduction is envisaged.

KEYWORDS:

• Intelligent packaging
• waste reduction
• quality-controlled logistics
• computer simulations
• fresh food chain

Introduction

One-third of the food produced worldwide is lost or wasted according to the FAO (Gustavsson et al. 2011; FAO 2013) According to Verghese et al. (2013), even 40% of all food intended for human consumption in developed countries ends up as waste; and Stuart (2009) estimates that in North America and Europe 30–50% of the food supply is discarded. Beretta et al. (2013) showed that 48% of the total calories produced by primary production are lost across the whole supply chain. Quantification of food waste per food category by households has been studied by Williams et al. (2012), who showed that fruit and vegetables, prepared foods and dairy products contribute most when expressed on a weight basis. ‘Food waste’ can be defined as food that is appropriate for human consumption but which is being discarded either before or after it spoils (FAO 2013), and that results from the negligence or a conscious decision to throw food away (Lipinski et al. 2013). The later a product is lost or wasted along the supply chain, the higher the environmental cost, as impacts arising, for instance, during processing, transport or cooking will be added to the initial production impact (FAO 2013). Preventing avoidable food waste generation throughout the supply chain features the most advantageous option within the food-waste hierarchy (Papargyropoulou et al. 2014). Spoilage and reaching the end of the printed shelf life date on a package is an important reason for food waste. But also a lot of food is thrown away when a sample from a batch does not fulfil the specified quality levels, leading to destruction of a whole batch. These food losses can be reduced by monitoring the quality of foods throughout the whole supply chain and using this information efficiently (Verghese et al. 2013; Jedermann et al. 2014). Especially in medium- and high-income countries is food wasted primarily at the retailer and consumer levels (Gustavsson et al. 2011). Unfortunately, it is during these steps of the supply chain (consumer and retailer) where quality monitoring is difficult to achieve. No appropriate non-destructive methods for quality monitoring of individual products are available, since most methods are too complicated or require expensive and complicated equipment or materials.

Intelligent packaging (IP) offers the opportunity to monitor the quality of foods throughout the whole chain, also in those parts that could not be monitored previously. IPs are packaging systems that monitor the condition of packaged food during its life cycle and communicate (i.e., indicate) this information related to the quality or safety of the packaged product. An intelligent package can contain sensors and/or indicators that monitor quality indicator compounds of the product or quality indicator environmental conditions (Heising et al. 2014b). The information from these sensors needs to be translated into a meaningful quality indication that is communicated to some or all the actors of the chain (Jedermann et al. 2014). For this purpose, the measured signals can be converted by mathematical models into values indicating the quality condition of the food product. Wireless sensor networks offer the most advantages for automatically monitoring each food product or its conditions in the whole supply chain and for optimal supply chain management (SCM) (Aung et al. 2012; Hammervoll et al. 2012).

This paper describes how IP can be used to reduce food waste when used in SCM based on quality-controlled logistics (QCL). A conceptual approach is explained and supported by quantitative data from simulations on the effect of using the information of IP in SCM with the goal to reduce food waste.

Why can IP be used to reduce food waste when used in SCM?

The causes of food losses and waste in medium- and high-income countries mainly relate to consumer behaviour as well as to a lack of coordination between different actors in the supply chain (Gustavsson et al. 2011). For optimal food SCM, perishable food must be sold to consumers before the product reaches its printed expiry date, while maximising profit. The current practices in the management of perishable products is far from satisfactory and perishable food loss at grocery retailers can be as high as 15% due to damage and spoilage (Ferguson & Ketzenberg 2006). The main cause of this is that the expiry date on the package of the food product is fixed (Tromp et al. 2012). The expiry date is fixed on the date by which, in the worst case, the quality or safety of the food is no longer acceptable. However, if the initial condition of a product and/or the conditions in the supply chain are more optimal, the quality of a food product is still acceptable after the expiry date on the package.

Typical for fresh food supply chain networks (FSCN) is the heterogeneity in and between batches of foods. For example, a high variation in the acceptance period has already been found between tomato growers growing the same cultivar and harvesting at similar maturity stages (Schouten et al. 2006). The quality and shelf life of foods in consumer markets are influenced by the product quality at the origin when harvested (in this paper named the ‘initial quality’) and the conditions to which the product has been exposed during its passage through the supply chain. This leads to a high variation in the shelf life of products, making it very difficult to print an accurate ‘best before’ or ‘use by’ date on the package to indicate the remaining shelf life.

Another downside of a fixed expiry date (FED) is that the willingness to pay of a consumer for a food product generally decreases if there are fewer days left before the expiry date (Tsiros & Heilman 2005). This means that as soon as a new batch with products with a longer shelf life enters the supermarket, consumers will more likely buy these new products and the older ones will not be bought, and after the expiry date is reached the supermarket will have to discard those products.

Therefore, a dynamic expiry date (DED), potentially with dynamic pricing, i.e., lowering the price as the expiry date approaches, is proposed by Tromp et al. (2012). Implementing a DED is possible with IP, since IP can monitor food quality continuously and, thereby, update the remaining shelf life. With IP the quality of foods can be monitored during phases in the supply chain that could not be monitored before (Figure 1).

Figure 1. Extended possibility for the monitoring of food quality during various stages of the supply chain of a product.

Intelligent packages can monitor environmental conditions that influence the kinetics of changes in the quality attributes of the food, e.g., temperature by a time–temperature indicator, or monitor quality attributes or quality indicator compounds directly, e.g., trimethyl amine (TMA) to indicate fish spoilage (Heising et al. 2014a) (Figure 2). The deterioration in the intrinsic quality attributes of food products should be considered to determine which IP concept is the most useful for monitoring or assuring good product quality (Figure 3). An advantage of monitoring a quality attribute directly is that differences in the initial quality of foods are directly taken into account in the measurement of product quality (Heising et al. 2014b). The time–temperature indicator’s ability to give meaningful information on product quality is limited to products with a constant initial quality, e.g., pasteurised milk or processed meat products.

Figure 2. Schematic illustration of intelligent packaging (IP) with sensors that monitor environmental conditions, or quality attributes of the product related to overall food quality change. The environmental condition sensor can also be placed inside the package depending on the measured condition. Source: Adapted from Heising et al. (2014b).

Figure 3. Overview of the use of intelligent packaging (IP) to communicate about its quality attributes. Although there are always reactions going on in foods, some quality attributes remain relatively constant for a long time. Whether or not a quality attribute changes after processing depends on both the food type and shelf life. For example, the quality attribute ‘nutritional value’ of fresh fish will not change much during its life cycle since the sensory changes happen much faster and are rate limiting for the shelf life of the product. Therefore, nutritional value is considered stable in fresh fish, although lipid oxidation can become important (rate limiting) in fish with a longer shelf life. Some quality attributes can be estimated well by consumers, e.g., the colour of bananas indicates their ripeness. In this case there is no need to monitor this quality attribute by an intelligent package. To monitor a quality attribute, product properties that can be measured in a non-destructive way in the package must give a good indication of the quality attribute. Source: Adapted from Heising et al. (2014b).

Van der Vorst et al. (2011) described how real-time information on actual product quality can be combined with logistics decision-support models to improve the performance of FSCNs. The additional information gained from sensors from IP can be incorporated in quality-change models during the complete distribution process, leading to knowledge about product quality status at its finally destination (Hertog et al. 2014). This advanced logistics decision-making is called ‘quality-controlled logistics’ (QCL) (Van der Vorst et al. 2011). QCL is based on the variation in product quality, developments in technology, the heterogeneous needs of customers and the possibilities to manage product quality changes in the distribution chain. The concept matches consumer demands for a specific product quality and/or price with the available variation in the supply of products with a certain quality. Van der Vorst et al. (2011) identified the basic elements of QCL, two of them being ‘product quality measurement and prediction’ and ‘logging and exchange of information’, for which sensors combined with radiofrequency identification (Zou et al. 2014) in an IP would be the perfect solution.

Single food products from a batch are packaged by the producer. From this stage on, IP associated with DED (i.e., IP-DED) information can be used for decisions on the further distribution through the supply chain. Since no longer will all products in a batch have the same expiry date, retailers can apply a first-expired-first-out (FEFO) approach instead of a first-in-first-out (FIFO) approach, which will reduce food waste according to Dada and Thiesse (2008). As IP-DED not only provides information on the current quality of the product but also predicts future quality based on the current status, logistic decisions can be made based on quality via QCL (Kaipia et al. 2013). As conditions vary along different routes in the supply chain, products with the lowest quality, e.g., with the earliest expiry date, can be sent to a distribution centre or a supermarket that is nearest, has the largest turnover rate or is on a route where the conditions are most favourable. In this way products with lower quality have a higher chance of being sold before they expire, whereas the products with a higher initial quality can be sent on less favourable routes through the supply chain. A dynamic product quality evaluation-based price can influence consumers’ buying decision and help to reduce food waste in a perishable food supply chain (Wang & Li 2012). The use of IP-DED also makes it possible to implement a dynamic pricing system in which the price of the food is automatically adapted from the (electronic) signal of the quality sensor, dependent on the predicted remaining shelf life. This should result in fewer products expiring before sale and consumption, thereby reducing avoidable food waste.

Overall, implementing QCL and FEFO by using the information from IP-DED can help reduce food waste. Figure 4 shows this conceptual approach for using IP-DED in QCL to reduce food waste. In the next section this concept is illustrated quantitatively with the help of a mathematical model and computer simulations.

Figure 4. Conceptual model for using IP-DED in QCL to reduce waste.

Materials and methods

Modelling quality in the supply chain

The modelling approach is based on a simplified generic food-supply chain. The model concerns the distribution over the route between the producer and the retailer, as displayed in Figure 5.

Figure 5. Simplified general supply chain used in this modelling approach.

Models for food-quality degradation were integrated in a mixed-integer linear programming model for planning distribution (Rong et al. 2011) and to make simulations on food waste in a simple food-supply chain.

The raw material supply is divided over the producers. At the producer the products are packed in an IP-DED system. Therefore, it is known how many products of which quality level are in the batch of supply. The quality of the products decays according to a function F that depends on the temperature during transportation, the time and the initial quality. The producer can control which products are sent at which transportation temperature to which retailer, depending on the quality of the products and the demands of the retailer. This resembles the choice a producer can make between routes on which the quality decays differently. If the total amount of supply is bigger than the total amount of demand in a certain period, there will be an amount of products left over at the retailer. If the quality of these products is below an acceptable quality level, they are no longer suitable for consumption and, thus, will be considered as waste. This waste needs to be minimised. Products of higher quality levels can be kept in storage by the retailer and sold at a later date. The quality of these products will be continuously decreasing; the likelihood that they are sold before they reach the acceptable quality level is higher if their quality at reception by the retailer is higher. The retailer can lower the demand of new products depending on the amount and quality level of the products at his location.

Lowering the temperature during transportation will reduce the rate of quality decay, but costs extra money. Therefore, the end quality of still unsold products can be increased by increasing the cooling costs. Minimising the costs requires making a trade-off between costs and waste-minimisation objectives. Another way would be to allocate costs to the amount of products that are wasted since they require an increase in production to meet the total demand and in addition will have disposal costs.

Notation

Table 1 shows the notation introduced for the purpose of modelling.

Table 1. Notation used in the model.

Indices
i	Producers
j	Retailers
q	Quality levels, where 1 is the highest quality level and Q is the lowest quality level
t	Temperature levels during transportation, where 1 is the coolest temperature level and T is the non-cooled temperature level
Decision variables
	Amount (kg) of product that leaves producer i with quality level q, to go to retailer j, being transported at temperature level t
X_wastej,q	Amount (kg) of products with quality level q that arrived at retailer j but which are not needed to satisfy the demand
Parameters
Ckt	Cooling costs during transportation per temperature level t (€ kg^–1)
Cti,j	Transportation costs per route (€ kg^–1)
Quaq	Part of a batch that has an initial quality level q
Qstart	Quality level at the beginning of a distribution segment (%)
Qenf	Quality level at the end of a distribution segment (%)
kT_ref	Reaction rate constant at the reference temperature (h^−1)
Tref	Reference temperature (°C)
T	Temperature (°C)
Ea	Activation energy (J mol^−1)
R	Universal gas constant (8.314 J K^−1 mol^−1)
time	Time (h)

Model formulation

The developed model belongs to one of the most widely used mathematical programming models in operations research, which are also referred to as constrained optimisation models (Hillier & Lieberman 2005; Claassen et al. 2007; Williams 2013). This class of models aims to give an abstract description of a decision problem from practice, e.g., the described multi-objective quality-based distribution planning problem. The essential feature of this class of models is that they all share the same general structure of optimising an objective function subjected to at least one constraint. Decisions are represented by decision variables that are restricted to taking only allowable values, expressed by constraints. The constraints restrict the set of solutions to so-called feasible solutions. An optimal solution for the considered problem is obtained by maximising or minimising one (or more) objective function(s):

(1)

(2)

In the above formulation there are two objective functions. Objective function (1) aims at minimising transportation costs and cooling costs. Product waste is minimised by objective function (2).

These objective functions are subjected to several constraints:

• The total supply equals the supplies to the producers. The maximum capacity at each producer should not be exceeded. As the supply is being packaged in an IP-DED system at the producer, it is known from this stage on how many products are at each quality level. The maximum transport capacity should not be exceeded. The retailer’s demands and quality requirements are based on the quality and amount of unsold products.

• The quality decay of the products is dependent on temperature, time and the initial quality. For example, in the case of fresh fish, the quality levels are inversely related to the concentration of TMA that can be measured by IP (Heising et al. 2015). TMA is produced by spoilage bacteria. The activity of the microorganisms on the fish depends on the temperature and time conditions of the storage of the fish. Although the formation of TMA (Heising et al. 2014c) and the degradation kinetics of fresh fish can be simulated by a complex model (Heising et al. 2015), in this simulation product quality decay is simplified to a temperature-dependent first-order reaction.

For each segment of the distribution route, the quality-decay factor X for a first-order reaction can be described as:

(3)

where Q_start and Q_end are quality levels related to a quality attribute monitored by IP-DED. The reaction rate constant k_T,_ref and the activation energy E_a depend on the indicator compound measured by IP-DED, thus on which quality attribute is critical for determining the expiry date. The final quality level is dependent on the X-values that occur along the route that a product has passed before reaching the consumer. Therefore, the implementation of QCL can influence the quality of food products when they reach the supermarket.

Results and discussion

Model simulations

Different scenarios were simulated with the model to obtain quantitative insight into how IP-DED can be used in SCM to reduce food waste. Three different scenarios (with IP-DED and QCL) are optimised using the mathematical programming package Xpress-IVE (v. 7.2.1) and compared with a standard situation. Xpress-IVE is a commercial state-of-the-art software package distributed by FICO (see http://www.fico.com/en/products/fico-xpress-optimisation-suite). The FICO-xpress optimisation suite is widely used by academics and in practice for modelling and solving large-scale-constrained optimisation problems (Xpress FICO 2011). After successfully developing and implementing IP-DED the main advantage is that ultimately the quality of the products is known at all steps in the supply chain; therefore, it is known how many products have a certain initial quality level. The distribution of this initial quality level of the products was varied in the different scenarios. For the different scenarios, the Qua_q parameter was changed according to Table 2. A simplified supply chain (Figure 5) was used and only three temperature levels (2, 4 and 7°C) and three quality distributions were used to demonstrate the QCL concept for waste reduction. In scenario 1, the majority of the supplied products have the highest initial quality level. In scenario 2, the average initial quality is lower and the deviation of the quality of the products is larger. In scenario 3, most supplied products have an average initial quality level.

Table 2. Distribution of the quality levels of the initial food product supply for the different scenarios in fraction of the total supply.

Scenario\Quaq	q = 1	q = 2	q = 3
0 (standard)	0.00	1.00	0.00
1	0.85	0.15	0.00
2	0.60	0.35	0.05
3	0.15	0.80	0.05

The standard situation represents the current situation without IP-DED. Although the initial quality of the supplied products will in reality be variable, this is unknown due to the absence of IP, so all products will have the same expiry date. In the current situation this is represented by the expiry date of quality level 2. In the standard situation there is only one temperature level used for transportation. The expiry date is, therefore, based on the worst-case scenario with respect to the assumed initial quality and the conditions in the supply chain. There are no products that arrive at the retailer with the expiry date of the highest quality level that can be kept longer at the retailer. Therefore, all the products that are leftover at the retailer are considered as waste. The total waste is the amount of products that is supplied in excess of the sellable products, which has been estimated by this simulation to be 42.5% in the standard situation. In scenarios where IP-DED is used, this amount can be minimised by using the model equations.

Whereas only one temperature level was used in the standard situation, in the scenarios all three temperature levels are used. Therefore, the total supply is divided over routes with different temperature levels. Additionally, the total demand is divided over the available distribution of the quality levels in each scenario. The resulting waste reduction of all scenarios compared with the standard situation is shown in Figure 6.

Figure 6. Waste percentages in the standard situation and after optimising the different scenarios.

In all scenarios the waste could be reduced considerably compared with the standard situation when QCL was applied. The higher was the initial quality level, the more waste could be prevented; in scenario 1 even zero waste was simulated. Even if transport and cooling cost minimisation were the objective, waste was reduced for two out of three scenarios, but not as drastically as for the waste-reduction optimisation. It has to be taken into account, however, that applying IP-DED costs extra money. Looking at it from an economic perspective only, the extra cost of implementing IP-DED needs to be less than the reduction of the total costs (transport, cooling and wasted products).

Conclusions

Without the implementation of IP-DED and QCL the expiry date is based on a worst-case scenario. As a result, the products with high initial qualities or those that have taken a route with a low decay factor will not yet be expired on the expiry date shown on the package. These products will not be bought and consumed after the expiry date, although they are still suitable for consumption, and will have to be thrown away, resulting in food waste.

However, when packed in an IP system that displays a DED, products can be sorted based on their expiry date and, thus, on their initial quality. This allows the implementation of QCL. As conditions vary along different routes in the supply chain, products with the lowest quality, e.g., with the earliest expiry date, can be sent to a distribution centre or a supermarket that is nearest, has the largest turnover rate or is on a route where the conditions are most favourable. In this way, products with lower quality have a higher chance of being sold before they expire, whereas products with a higher initial quality can be sent on less favourable routes through the supply chain. This should result in fewer products expiring before sale and consumption, thereby reducing food waste.

Because IP-DED is more expensive than regular packaging, it is only economically profitable when used for food products that cause a high-value loss when they are wasted. These are usually perishable, fresh products that are not or minimally processed (Heising et al. 2014b) and which are transported in batches with a heterogeneous quality distribution (Van der Vorst et al. 2007). Examples of such products are fresh fish, shellfish, meat, exclusive fruits, and vegetables and salads.

When IP can be combined with active intelligent packaging in SCM (AIP-DED), even more possibilities for food-waste reduction will arise (Heising et al. 2014b). In this way the quality of a food not only can be monitored and controlled but also even improved or the spoilage rate can be slowed down, e.g., by activating a volatile antimicrobial emitter.

Before these highly sophisticated (active) IP will become mainstream in the supply chains of high-value perishable foods, some hurdles must be overcome. The sensitivity and reliability of the sensor signals have to be quantified and validated. They should be accepted by all stakeholders in the supply chain, including consumers, and comply with legislation. Furthermore, consumer studies need to be conducted to study whether IP-DED can also help consumers with planning their purchases at the retailer and with consumption at home to minimise food losses in households, and how to communicate the information of IP-DED to consumers. These aspects are important areas for future research in this field which according to the presented simulations has great potential, amongst others, in food-waste reduction.

Acknowledgments

The authors thank Nathalie Kersten and Anne van den Hurk for their contributions through their BSc thesis research.

Disclosure statement

No potential conflict of interest was reported by the authors.