Review: Consumption-stage food waste reduction interventions – What works and how to design better interventions

with aiming to halve per global food waste at the retail and consumer levels by 2030. However there is no review that has considered the effectiveness of interventions aimed at preventing food waste in the consumption stages of the food system. This significant gap, if filled, could help support those working to reduce food waste in the developed world, providing knowledge of what interventions are specifically effective at preventing food waste. This paper fills this gap, identifying and summarizing food-waste prevention interventions at the consump-tion/consumer stage of the supply chain via a rapid review of global academic literature from 2006 to 2017. We identify 17 applied interventions that claim to have achieved food waste reductions. Of these, 13 quantified food waste reductions. Interventions that changed the size or type of plates were shown to be effective (up to 57% food waste reduction) in hospitality environments. Changing nutritional guidelines in schools were reported to reduce vegetable waste by up to 28%, indicating that healthy diets can be part of food waste re- duction strategies. Information campaigns were also shown to be effective with up to 28% food waste reduction in a small sample size intervention. Cooking classes, fridge food sharing apps, advertising and information sharing were all reported as being effective but with little or no robust evidence provided. This is worrying as all these methods are now being proposed as approaches to reduce food waste and, except for a few studies, there is no reproducible quantified evidence to assure credibility or success. To strengthen current results, a greater number of long- itudinal and larger sample size intervention studies are required. To inform future intervention studies, this paper proposes a standardised guideline, which consists of: (1) intervention design; (2) monitoring and mea- surement; (3) moderation and mediation; (4) reporting; (5) systemic effects. Given the importance of food-waste reduction, the findings of this review highlight a significant evidence gap, meaning that it is difficult to make evidence-based decisions to prevent or reduce consumption-stage food waste in a cost-effective manner.


Introduction
Within the last decade, food waste has become an issue of international concern to policy makers, practitioners, and researchers across a range of academic disciplines. Recent estimates suggest that globally one third of food never reaches a human stomach (FAO, 2011), and global food waste is associated with large amounts of greenhouse gas emissions (FAO, 2013). Growing political and public consensus around the urgency of these challenges has provided the impetus for govern-wide range of food-waste-related issues. However, to the best of our knowledge, there is no review of the effectiveness of downstream foodwaste interventions. 4 Four intervention studies were reviewed by WRAP (see Appendix F of WRAP, 2014a). These were all from the grey literature and UK-based. Since then a number of further studies have emerged, the most important of which are mentioned in the discussion section below.
In summary, there is no peer-reviewed study that has considered the effectiveness of interventions aimed at preventing food waste in the consumption stages of the food system. This represents a significant gap, which, if filled, could help support those working to reduce food waste in the developed world, providing knowledge of what interventions are specifically effective at preventing food waste. This paper fills this gap, reporting a rapid review of the food-waste literature from 2006 to 2017 focussing on downstream food-waste reduction interventions. 5 Based on the findings, the paper then categorises the successful interventions and discusses the components of a successful food waste reduction intervention.

Methods
The methodology for rapid reviews has emerged as a streamlined approach to synthesizing evidence in a timely manner -rather than using a more in-depth and time-consuming systematic review (Khangura et al., 2012;Tricco et al., 2015). As discussed by Tricco et al., there is no set method for a rapid review; however, there are several common approaches. For this study, a rapid review was undertaken to provide fast and up-to-date information, responding to demand from the policy and academic community (c.f. Lazell and Soma, 2014;Porpino, 2016).
We used Google Scholar to identify relevant papers using combinations of the following terms: 'Food waste', 'household', 'quantification', 'behaviour change', 'consumer', and 'downstream'. The time period was restricted to January 2006 until January 2017. This was a result of discussion with expert advisors and evidence from other bibliometric studies that food waste studies only began to be published from 2006/7 onwards (Chen et al. (2015), Hebrok and Boks (2017), Carlsson Kanyama, Katzeff, and Svenfelt (2017), and Schanes, Doberning, and Gӧzet (2018). This search enabled the inclusion of online first/only preprints of 2017 journal articles. The search was restricted to English-language publications. Each paper was then mined using the Google Scholar "citation" function to explore the network of papers that have cited each paper. Each of these papers was then captured and explored via the process described above. Fig. 1 outlines our rapid review method, with 454 items narrowed down to 17 peer reviewed journal articles focussing on downstream food-waste reduction interventions.
Though it is common in rapid reviews to use scoring criteria to sort and exclude papers on the basis of method or data quality, no such scoring method was used in this paper. This is due to the small number of studies found, and wishing to provide the food waste community 1 The Sustainable Development Goals are a collection of 17 global goals set by the United Nations General Assembly in 2015. The SDGs cover social and economic development issues including poverty, hunger, health, education, global warming, gender equality, water, sanitation, energy, urbanization, environment and social justice.
2 Grey literature refers to non-peer reviewed literature such as reports, conference proceedings, doctoral theses/dissertations, newsletters, technical notes, working papers, and white papers. 3 I.e. where food is consumed such as in the household, and in hospitality and food service sectors. 4 While this manuscript was in final stages of peer review, a review of downstream food waste interventions between 2012 and 2018 was published by Stöckli et al. (2018b). It identified the same papers as identified by this manuscript (with addition of 2017-2018 peer reviewed papers: (Qi and Roe, 2017;Romani et al., 2018;Stöckli et al., 2018a)), and came to similar conclusions regarding the need for systematic evaluation of interventions between. The additional novelty of our paper is (1) situating a broader range of peer reviewed intervention papers (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) within the broader food waste literature (see , and (2) our in-depth discussion and proposal of standardised guidelines for intervention development. 5 "Downstream" being a wide definition, but meaning the consumer side of the food system. Downstream interventions could include interventions in supermarkets, hospitality and food service sectors (including food served in education and healthcare, government, etc.), and household consumption.

Table 1
A summary of the nine bibliometric studies and meta-analyses that review food waste literature. 2000-2015 "Food waste" in combination with the words "household", "packaging", "consumer", "behaviour" and "design".
Results must be written in English, the resultant were from Western  Countries  2017 Reviews aspects of consumer food  waste (consumer behaviour,  attitudes, beliefs and values,  quantifications and compositional analyses, waste prevention, and (continued on next page) C. Reynolds et al. Food Policy 83 (2019) 7-27  1987-2017 "Food waste" AND "behavior change", "food waste" AND "intervention", "food waste" AND "sustainable consumption", "food waste" AND "nudging". C. Reynolds et al. Food Policy 83 (2019) 7-27 with as comprehensive as possible assessment of recent intervention studies. It should also be noted that the waste reduction percentages reported here have been calculated from all studies that reported weights and changes to waste generation. The waste reduction percentages are not directly comparable with each other as they have differing functional units, i.e. per plate, per person (participating or general population), per organisation (kitchen and front of house), per total weight of waste, etc.), or differing time scales (for data collection or experiment duration).

Broad rapid review
The rapid review identified 292 downstream food waste articles that were published in 39 journals between 2006 and 2017.
From 2006, the number of downstream food waste articles published yearly increased rapidly as greater attention was given to the challenge of food waste, with the largest spike in articles that quantify food waste (Fig. 2) occurring in 2013 after the publication of reports highlighting the global issue (Institution of Mechanical Engineers, 2013; Lipinski et al., 2013). Out of the articles surveyed, only 17 (5%) feature applied downstream food waste reduction interventions. The most popular methodologies (Fig. 3) used in the rest of the downstream food waste studies include surveys (n = 80, 27%), reviews (n = 77, 26%) and Life Cycle Assessment (LCA) modelling (n = 50, 14%). Journal articles featuring qualitative, observational and ethnographic methods (following Evans (2014)) are consistently published throughout the time period (n = 18, 5%).
48 countries or geographic areas were identified within in the broader downstream food waste literature ( Fig. 4) with 8 articles not identifying their geographic location, and 53 global studies. The next most studied areas were the USA (n = 42), the UK (n = 34), Sweden (n = 21) and Italy (n = 20). China (n = 13) is the only developing country in the top 10 countries/regions studied. Our results show that global studies emerge after 2010 -as data quality and accessibility increases. Countries that had an early identification of food waste as a social problem (including USA, UK and, Sweden) continue to publish prolifically.

Intervention studies
The seventeen journal articles focussing on downstream food-waste reduction interventions were first categorised by the main intervention types that were applied: information based, technological solutions, and policy/system/practice change. Journal articles can be in more than one category if multiple interventions were used (either applied separately or together). Table 2 provides a detailed summary of each intervention and paper.
The seventeen articles with applied interventions were found in sixteen journals covering nutrition and health (5 journals), psychology and consumer behaviour (5), environmental (3), human computer interactions (2), food (1) and economics (1). The majority of these articles were published in relatively 'low' impact factor journals (under impact factor 3). 6 Within the applied downstream food waste reduction interventions ten countries feature, with the USA being the site for 6 articles, 3 in the UK (one of which is a cross country comparison with Austria), and 2 in the Netherlands. The geographic spread of these 17 articles is focused on the global north, with Thailand the notable exception.
The areas of study for the seventeen applied downstream food waste reduction interventions are focused on households and the community (n = 6), hospitality and hotels (n = 5), and educational establishments (n = 6). This is a much narrower field of study than what is found across the rest of the downstream food waste literature with 8 categories of intervention area identified in Fig. 4.
The success of these interventions varied. A student-focused education campaign (Martins et al., 2016) resulted in a 33% waste reduction in main dishes, while the Home Labs intervention (a collaborative experiment with householders) led to an overall reduction in food waste generation of 28% . New hotel signage reduced food waste by 20% (Kallbekken and Saelen, 2013). Enewsletter use resulted in 19% reduction in self-reported food waste in the home (Young et al., 2017). Schmidt's information campaign resulted in a 12% perceived (self-reported) improvement in food waste reduction in the home (Schmidt, 2016). Whitehair et al.'s information prompt resulted in a measured 15% food waste reduction in a university cafeteria, while portion advertising information also resulted in greater uptake of smaller portions (up to 6% from 3.5%) (Jagau and Vyrastekova, 2017).
Technological solutions Ganglbauer et al., 2013;Lazell, 2016;Lim et al., 2017;Wansink and van Ittersum, 2013;Williamson et al., 2016a;Young et al., 2017) involve the introduction or modification of technologies and/or objects that seek to alter the behaviours around food (waste). These included changes to plate or portion sizes (Williamson et al., 2016b) or the introduction of fridge cameras or food sharing apps (Ganglbauer et al., 2013). Only plate and portion size studies have quantified waste reduction. The largest reported waste reduction (57%) was due to shifting to smaller plate sizes, although in this study there was also a 31% decrease in the amount of food consumed via the plate size shift (Wansink and van Ittersum, 2013). 7 Other studies have reported a 19% reduction in food waste due to reduction in plate size (Kallbekken and Saelen, 2013), and a 51% reduction in food waste was achieved by using permanent rather than disposable plates (Williamson et al., 2016a). A 31% reduction in french fries waste was enabled by moving to smaller portion sizes (Freedman and Brochado, 2010).
Policy/system/practice change (Cohen et al., 2014;Dyen and Sirieix, 2016;Freedman and Brochado, 2010;Kallbekken and Saelen, 2013;Martins et al., 2016;Schwartz et al., 2015) is where polices or systems are altered and the population changes food waste behaviours (or practices). Two articles involved changing school dietary guidelines, which resulted in a 28% (Schwartz et al., 2015) and 14.5% (Cohen et al., 2014) vegetable waste reduction, while changing how schools and students were taught about food waste resulted in a 33% waste reduction from main dishes (Martins et al., 2016). These results indicate that diet reformulation and healthy eating can be part of food-waste reduction strategies.
In the seventeen journal articles with interventions, five relied on self-reported (usually survey-based) measurements of food waste (a method that is relatively low-cost but suffers from substantial biases (World Resources Institute, 2016)). One paper did not disclose any waste weights, while another two estimated food waste via visual analysis or pictures. The remaining nine used weight-based waste measurement. It is a challenge to accurately quantify food waste prevented, largely due to the costs of waste measurement (especially in the home). The cost of waste measurement could explain why only 123 of the 292 journal articles (42%) identified by the broader rapid review include some quantification of food waste generation/ diversion/ reduction. Due to this reliance on self-reporting, only the accuracy of the three plate-change/size-reduction interventions can be assessed with any certainty (Kallbekken and Saelen, 2013;Wansink and van Ittersum, 2013;Williamson et al., 2016a). The comparative measurement of these studies is also not directly comparable as the methods of weight measurement and the unit of measurement vary (i.e. per plate or aggregated total waste), and time intervals (study duration, number of observations, etc.) differ between each study as reported in Table 2.
Around a third of these studies (5 articles) do not integrate any theoretical framework or disciplinary orientation into their experimental design. Those that do are typically single theory in nature, and do not interact with the broader food waste literature. Theoretical frameworks and disciplinary orientations in the downstream intervention articles include Social Practice Theory; Behavioural Economics (nudge-approaches such as visual prompts), Transformative Consumer Research, pro-environmental behaviour change, behaviour change determinants, and the integrative influence model of pro-environmental behaviour.

Discussion of themes and policy implications
In light of the above results, in this section we provide an overview of the methodologies, theoretical lenses and types of interventions C. Reynolds et al. Food Policy 83 (2019) 7-27 employed in both the academic and grey literatures, and then recommended a series of recommendations -or principles -for organisations undertaking intervention studies relating to food waste prevention related to the consumption stages of the supply chain.

Methodologies
Although there has been a rapid increase in articles that quantify or investigate downstream food waste since 2006, there have been only 17 peer-reviewed journal articles that feature downstream interventions that resulted in a food waste reduction. Of these, nearly 30% (5 articles) used self-reported methods to measure food-waste reductions, while another two estimated food waste via visual analysis or pictures. Due to the methods used, the results from these studies should be interpreted with caution (as indeed many of their authors note); in these cases, a claimed reduction in food waste should not be read as an actual reduction. Furthermore, 16 of the 17 interventions occurred in developed countries and most interventions have focused on small groups with time-limited evaluations.
Part of this limited methodological development may be due to previous food waste research having had limited cross-pollination between disciplines, both in terms of substantive questions as well as in theoretical development. Many researchers tend to rely on the theories they are comfortable with, resulting in a "silo"-ing not only of theories that could be useful in explaining food waste, but regrettably also a "silo"-ing of substantive findings related to actually reducing such waste. Further research is required to map the literature (and food waste's theoretical developments further) to understand if this is the case.

Theoretical lenses
The absence of explicit reference to theory means that readers are left to infer connections between cause and effect in food waste  behaviours or that connections are imputed without explicit justification. Nearly 30% (5 articles) of the downstream intervention studies did not mention a theoretical framework. Of those that did, this was often not a key part of the paper or research design. This is an interesting finding: on the one hand, it could imply that those working on foodwaste interventions are not aware of theoretical frameworks developed for interventions in other domains; on the other hand, it could implyas discussed by Quested et al. (2013) -that food-waste prevention in consumption settings is very different from other areas of behaviour change (see also Evans et al. (2017)) and that many of the theories developed elsewhere are of limited value without further development. The lack of theoretical integration into food waste intervention design may also imply that theoretically rich accounts of household food waste (for example Waitt and Phillips (2016)) have yet to fully consider the implications of their analysis for interventions. We suggest that there is a need for greater integration of theory and previous research findings into the design of interventions. We also suggest that there is need to discuss how different theoretical frameworks, disciplinary perspectives and methodological techniques could combine to contribute to the reduction of food waste. Would it, for instance, be possible to combine a qualitative account of the social practices that generate food waste with quantitative tools that model the effects of different interventions?

Intervention types
Reduction methods such as improved information (Manomaivibool et al., 2016) or changes to plate type and size (Lazell, 2016;Wansink and van Ittersum, 2013;Williamson et al., 2016a), portion size (Freedman and Brochado, 2010), or menu composition (Cohen et al., 2014;Martins et al., 2016;Schwartz et al., 2015), all accept that their effectiveness may be due to greater consumption of the food, or shifts in the types of foods consumed and wasted. That is, as has been observed in other interventions studies, there may be unintended consequences (Peattie et al., 2016) that need further investigation. If this unintended shift is towards the overconsumption of unhealthy foods or at the expense of healthy foods, this could lead to negative health outcomes. For

USA
C. Reynolds et al. Food Policy 83 (2019) 7-27 this reason, attention must be given to communicating and encouraging people to monitor portion size rather than reducing food waste at the expense of public health. However some of the reviewed studies, indicate that some interventions result in a reduction in consumption alongside waste prevention (Kallbekken and Saelen, 2013;Wansink and van Ittersum, 2013 8 ;Williamson et al., 2016a). Further research is needed to understand which (healthy or unhealthy foods) are involved in this consumption shift and waste reduction. Moreover, it could be the case that many of the unintended consequences could be due to a lack of understanding around causal mechanisms and supporting theoretical frameworks. If this is the case, further engagement with theory-based evaluations would be an obvious solution.
Cooking classes (Dyen and Sirieix, 2016), additional technologies such as fridge cameras (Ganglbauer et al., 2013) or apps (Lazell, 2016;Lim et al., 2017), and advertising and information campaigns (Young et al., 2017) were all reported as being effective but with no accurate quantification provided. This is worrying as all these methods are now being proposed by peer reviewed studies as options to reduce food waste with no reproducible quantified evidence to assure credibility or long-term effectiveness. Future research and resources are needed to test these interventions with accurate measurement methods. 9 For many organisations working on food-waste prevention, they would like to affect change across relatively large populations (e.g. a country, city or state/province/county). Therefore, to assess the appropriateness of interventions, these organisations require information on their cost effectiveness, how easy they are to scale up and whether they can be tailored to different 'audiences' within the population. However, this additional information is currently non-existent in the literature.
In addition, many of interventions that feature advertising or an information campaign did not provide enough detail to analyse and correlate the content type, and tone (positive, negative, shocking, etc.), with the effectiveness of the campaign. This is an avenue for future research.

Links to other literature
As noted above, academic literature is not the only source of research and evidence relating to downstream food waste. Although not a primary focus of this review, the authors are aware of a small number of intervention studies in the practitioner/policy-focused 'grey' literature. For example, during 2016, the UK supermarket chain Sainsbury's undertook a year-long trial using a range of methods to prevent or reduce food waste in the home (Waste less, 2016). These interventions were a mix of information (via Food Saver Champions), technology (fridge thermometers, smart fridges and cameras, apps, etc.) and policy/ system/practice change (introducing tenant welcome packs, new food waste events and school programmes). Some of these interventions included actual measurement of food waste (via audits or Winnow/ Leanpath systems 10 ) -resulted in between 18% and 24% food waste reductions. Other interventions relying on self-reported measures, resulted in between 43% and 98% food waste reductions for the homes that took part.
In the USA, a partnership called Food: Too Good To Waste reported the findings of seventeen community-based social marketing (CBSM) campaigns aimed at reducing wasted food from households (U.S. EPA Region 10, 2016). These interventions were mainly information interventions, which introduced new information and tools into households. Measurement of food waste was conducted before and after the campaigns using a mixture of self-reported audits (participants weighing their own waste) and photo diaries. The results showed measured decreases between 10% and 66% in average household food waste (7-48% per capita) for fifteen of the seventeen campaigns. The successful interventions were between 4 and 6 weeks long, with samples of between 12 and 53 households.
The EU project FUSIONS reported several waste prevention strategies focused on social innovation (Bromley et al., 2016). Though most interventions involved food redistribution, the Cr-EAT-ive intervention worked with school children (n = 480) and their parents (n = 207) to reduce food waste in the home and promote key food waste prevention behaviours. The results from 18 households (of 29 households) that completed the kitchen diary activity managed to reduce their food waste by nearly half -if scaled (with the intervention effects kept constant) to a yearly quantity, this would equal a reduction of 80 kg per household per year. However, it is not known how long the intervention effects would last for, the longer term engagement/attrition rates of children and households, and if some of this reduction was caused by the effect of measurement itself (rather than the intervention).
During 2012/13, WRAP ran a food-waste prevention campaign aimed at London households (WRAP, 2013a). These interventions were mainly information interventions. This was evaluated via waste compositional analysis and reported a 15% reduction in household food waste. However, as noted by the authors, some of this reduction could have been the result of the research itself (i.e. households being influenced by participating in a detailed survey).
Between 2007 and 2012, household food waste in the UK reduced by 15% (WRAP, 2013b). However, it is not possible to isolate the effect of different interventions that were running over this period. In addition, economic factors -increasing food prices and falling incomes in real terms -are likely to have contributed to this reduction (WRAP, 2014b).
These examples from the grey literature do not alter the main conclusions of this review: that there is a lack of research surrounding interventions designed to reduce the amount of food waste generated, and a lack of evidence of the ease with which it is possible to scale up previous smaller interventions.
It is important for researchers, policy makers and practitioners working to prevent food waste that this evidence gap is filled with research of suitable quality. Below, we offer guidance and general principles that, if followed, will improve the quality of this emerging field of study, and allow the effectiveness of interventions to be compared and fully understood. Building on the shortcomings of previous studies and improvement suggestions as outlined by Porpino, (2016), we categorise these recommendations into 5 strands: intervention design; monitoring and measurement; moderation and mediation; reporting; and consideration of systemic effects. These recommendations are based on our review of the literature and the authors' prior knowledge and experience regarding food waste intervention design and application.

Recommended principles for effective interventions
This section presents a series of recommendations -principles -for organisations undertaking intervention studies relating to food waste prevention related to the consumption stages of the supply chain. We then discuss interventions with potential with reference to our results. 8 The impact of Wansink and van Ittersum's research may have been affected by recent allegations of poor academic practices, with two other publications by Wansink and van Ittersum having had corrections published since the allegations were made (Etchells and Chambers, 2018;van der Zee, 2017). 9 It is worth noting that preventing food becoming wasted (e.g. via preventing food waste at source, feeding to other people, etc.) may be more effective than diverting food that has already been categorised as waste away from landfill and incineration to other waste destinations higher up the food waste hierarchy (e.g. composting, anaerobic digestion). This is because, for a given weight of food waste, preventing it being wasted usually has a much larger positive impact -socially, environmentally and economically -than diverting it from (Blatt, 2017;Garrone et al., 2014;Moult et al., 2018;Quested et al., 2011). 10 Winnow and Leanpath offer in-kitchen 'smart' food waste weighing services for the hospitality sector. Winnow was trailed in home as part of the Sainsbury's intervention.

Design of intervention
We recommend that an initial decision should be made about whether the study is focusing on an 'applied' intervention and/or one used to develop understanding of the intervention process. This should be explicitly stated in the methods and (experimental or intervention) design.
An applied intervention aims to reduce food waste across a given population or sub-population (i.e. it is scalable, with a clear target audience). For the interventions reviewed this was not always the case. For a communications-based intervention, this would need to be similar to the type and tone of material that could be used by a campaign group or similar organisation. If it were a change to food packaging, for example, it would need to be a change that could be adopted by a wide range of food retailers (e.g. it would have to ensure food safety and other packaging attributes whilst still being cost-competitive). To ensure that the 'quality' of such interventions is sufficient for the study, researchers should consider partnering with appropriate organisations with expertise in, for the above examples, developing communications materials or packaging technology. Partnerships also ensure that work is not being carried out in this area by organisations at cross purposes. In addition, applying techniques such as logic mapping (based on theory of change -see The Travistock Institute, 2010) can aid the design process to ensure that the intervention has the best possible chance of meeting its stated aims (i.e. preventing food waste in the home or other downstream settings). In addition, logic mapping and theory of change can enable the research to investigate how change occurs, as well as quantifying the degree of change. Much of this research and methods development has already been carried out on general behaviour intervention strategies within the field of environmental psychology, see Vlek (2009), or Abrahamse et al. (2005).
In contrast to 'applied' interventions, some research of interventions is designed to understand and evaluate how different elements of an applied intervention work. For these interventions the criteria discussed above are not strictly applicable. These types of studies may aim to understand which element of a larger intervention is responsible for the change -e.g. it may compare a range of campaign messages drawn from different disciplines and theories under controlled conditions. In such cases, it is not necessary that this module is scalable, although it would help future application of the research if the intervention studies needed only small modification to be deployed on a larger scale.
We also note that many studies use convenience sampling, which is likely to result in a group of study participants who are not representative of the wider population (or target populations within it). It will often include a sample with higher than average levels of education and income (Schmidt, 2016). Therefore, where possible, the design of the study should be considered to ensure that the sample is as representative of the population of interest as possible, ideally through random selection or, failing that, some form of quota sampling.
Previous discussion has indicated a lack of theory involved in the development of interventions; we feel that this stage is a key part of the intervention design process where theoretical understanding could be used to help develop more effective interventions.

Monitoring and measurement methods
Measurement of outcomes and impact of the interventions is challenging. Objective measures of food waste -such as through waste compositional analysis of household waste -are relatively expensive and are more easily deployed in geographically clustered samples (World Resources Institute, 2016). In addition, these methods only cover some of the routes by which wasted food can leave the study area, and so food and drink exiting the study area via the drain, or food that members of a household/school etc. waste in locations outside of the study area are not covered by such measurement methods (Reynolds et al., 2014). However, where there is an opportunity to deploy methods involving direct measurement, it is beneficial as these are generally more accurate and also minimise the amount of interaction with the household, reducing the impact of the measurement itself on behaviour.
Most of the other methods rely on some form of self-reporting -e.g. diaries, surveys, self-measurement of food-waste caddies, taking photographs. All of these methods generally give lower estimates of food waste in the home compared to methods involving direct measurement (e.g. waste compositional analysis) when comparison is made for a given waste stream. For diaries -one of the more accurate methods -around 40% less food waste is reported compared to waste compositional analysis (Høj, 2012). More recent analysis has shown that measuring food waste via caddies or photos gives similar results to diaries (Van Herpen et al., 2016). This lower estimate is likely due to a range of factors: people changing their behaviour as a result of keeping the diary (or other method), some items not being reported, and people with -on average -lower levels of waste completing the diary exercise (or similar measurement method).
Few studies discussed the problems presented by self-reported data. However, issues relating to self-report are discussed more extensively in the environmental (in particular recycling) and social marketing literature where self-reported measures of perceptions and behaviours are often considered unreliable (Prothero et al., 2011) and a gap is expected between self-reported and actual behaviour (Barker et al., 1994;Chao and Lam, 2011;Huffman et al., 2014). This should be discussed with reference to each intervention to understand the scale of uncertainty present in the results.
This means that those monitoring interventions have some difficult decisions to make: methods that are accurate may be unaffordable while methods that are affordable may be subject to biases that can compromise the reliability of the results. For instance, a communication-based intervention monitored using diaries may increase the level of underreporting of waste in the diaries, which could be erroneously interpreted as decreasing levels of food waste. This could have substantial -and costly -implications for those deploying the (potentially ineffective) food waste intervention in the future.
To address these issues, studies should try to obtain the requisite funding to be able to measure food waste directly (e.g. by waste compositional analysis). This may mean fewer studies, or studies comprising a panel of households, in which food waste is regularly monitored (with the householders' consent), creating the possibility of longitudinal studies. To make such an approach cost effective, this would likely require a consortium of partners, who could explore the emerging data to answer multiple research questions.
For studies using self-reported methods, these should carefully consider the design of the monitoring to ensure that reporting is as accurate as possible. The smaller the gap between actual and measured behaviour arising, the less measurement artefacts can influence the results and the ensuing conclusions. Recent work calibrating these self-reported methods has been undertaken (Van Herpen et al., 2016) and this type of information should be used in the measurement design. Further advances in calibration, especially in the context of intervention studies (i.e. is the level of underreporting stable during typical interventions?) would also help to improve monitoring and measurement.
In some circumstances, effects relating to self-reported measurement methods can be mitigated by the careful use of control groups. Where possible these should be used, as levels of food waste may change over time, influenced by food prices, income levels and other initiatives aimed at preventing food waste. However, adding a control to the research will increase costs and there can be practical difficulties in creating equivalent (e.g. matched) control groups, especially where samples are geographically clustered.
This discussion raises wider questions about the most appropriate evaluation approach and method, where different research designs may be fit for different intervention purposes. For example, where the priority is to measure an impact or effect, an experimental or quasiexperimental method should be considered, while assessing multiple outcomes and causal mechanisms may require a non-experimental research design (e.g. including qualitative methods). If the purpose is to decrease food waste by X percent, then the level of food waste should be measured over the course of the intervention (and beyond, to understand the longevity of the effect). In some contexts however, the purpose is to achieve a precursor to food-waste prevention (e.g. increased reflection on food waste, or to improve cooking skills), which may eventually lead to decreased food waste. In the latter cases, evaluation may want to focus on measuring the level of reflection, cooking skills, etc. to assess the effectiveness of the intervention.
We acknowledge that research on food waste is an interdisciplinary field. This can be a virtue, with many perspectives tackling this 'wicked problem'. However, it also means that different disciplines have different conventions and priorities, e.g. over the experimental scale or duration, and measurement of uncertainty vis-à-vis determining how much food is actually wasted. These differences should be acknowledged in order that more accurate and consistent measurement takes place.

Moderation and mediation
In addition to changes in the level of food waste, intervention studies may benefit from measuring changes in other quantities. This may help understand whether the intervention is effective, especially in situations where measurement of food waste is imperfect. Additional dietary (purchase and consumption) data can be collected and would provide greater certainty regarding food waste generation statistics. Additional waste generation data (beyond just food waste) could also be useful to help understand wider waste generation issues and drivers.
Examples of other measurements may include 'intermediate outcomes': depending on the intervention and how it operates, there may be intermediate steps that would need to occur for the intervention to operate as envisioned (as articulated in the intervention's logic map -see stage 1). This is an approach often used in social marketing where changes in behaviour that are difficult to measure might instead track changes in knowledge, beliefs and/or perceptions (Lee and Kotler, 2015). For instance, an educational campaign aimed at increasing the level of meal planning prior to people going shopping could monitor the change in people's awareness of educational material and their (self-reported) level of meal planning. These types of learning processes are slower, and are more difficult to assess in the short term, but they might still be successful and might achieve more long-term effects. Triangulation data is not sufficient in itself to state whether an intervention was successful, but can provide supporting evidence. Such analysis of moderating or mediating effects is useful and often uncovers interesting insights that would not be highlighted if this analysis were not conducted.
Observational analysis and measurement can provide insight into why the intervention works. By observing the intervention in action, this allows insight into the intervention itself, in addition to the effects of the intervention. This expands upon the intervention proposals of Porpino et al. (2016) by not only measuring the main objective, but also the intervention process, reflecting recent studies that highlight the importance of both process and outcome evaluation in interventions (Gregory-Smith et al., 2017).

Reporting
In order to make any study replicable and repeatable, there should be sufficient information provided about the intervention and the measurement methods to be able to replicate both elements.
The reporting of food waste has become standardised with the publication of the Food Loss and Waste Accounting and Reporting Standard (World Resources Institute, 2016). This standard was designed for countries, businesses and other organisations to quantify and report their food waste; it was not developed with intervention studies in mind. However, many of the principles it describes are useful in this context: studies should clearly describe the types of food waste measured (e.g. just the wasted food (i.e. edible parts) or including the inedible parts associated with food such as banana skins; the destinations included (e.g. only material bound for landfill, or also food waste collected for composting); the stages included (e.g. in a restaurant, only plate waste, or also kitchen waste).
A description of the details of how the quantification method (e.g. for waste compositional analysis) was undertaken is crucial, alongside what the study classified as food waste and which waste destinations were included. Details of the sample sizes and how they were drawn should also be covered. Data reporting should include the average weight, alongside appropriate measures of the spread of the data (e.g. standard deviation, standard error, interquartile ranges). Detailed waste composition data, where available, should also be provided. Changes of food waste between time periods should be reported as both weights and percentages, with significance and p values clearly stated. This minimum level of comparable data was lacking in many of the papers reviewed, with only 12 (70%) of the papers providing some statistics or statistical analysis, 2 (11%) providing waste composition analysis, and 5 (29%) providing results or analysis of food waste reduction from multiple time periods post intervention.
To allow for the actual measurement of food waste rather than participants' perceptions, several methods of disruptive thinking and scaling innovations could be considered. One such innovation is smart bins (Lim et al., 2017). This allows automatic recognition of food waste type and their weighting which can help remove uncertainty in selfreporting of food waste. Such data from smart bins (and also smart fridges and online shopping devices) could be shared with local authorities, policy organisations, community groups and industry, enabling planning and optimisation of food waste management locally. Smart bins are already being used in the hospitality industry to track food waste (e.g. products such as Winnow or Leanpath).

Considering systemic effects
None of the intervention studies in the review considered systemic effects. Systemic effects, like the rebound effect (i.e. improved technology to reduced environmental impacts may, due to behavior and other system effects, result in no change, or increased environmental impacts. See Khazzoom (1987) or Sorrell and Dimitropoulos (2008) for further discussion), are relevant and vital to consider for measures that are saving money or time for the consumer. Several of the measures presented above are not only measures that can lead to reduced food waste, and thus reduced environmental impact, but also measures that could lead to reduced costs, both for consumers and for other actors in the food chain. Since less food needs to be wasted, less food needs to be bought. Reduced costs can be an advantage from a private economic point of view, but it can also in the worst case, lead to further negative environmental effects. The money saved can be used for other types of consumption and perhaps increased environmental impact. These type of system effects, are sometimes called second order effects or rebound effects (Arvesen et al., 2011;Börjesson Rivera et al., 2014). How consumers choose to spend the money saved determines what the overall environmental impact will be. If the money or time is used for something more environmentally friendly, then the effect will be positive, and the environmental potential will be realised. But if instead the money is used for activities with more environmental impact, such as a food with higher environmental impact or, taking a trip with a fossil fuel driven car or even a flight, then the environmental impact is negative. Sometimes the second order effect exceeds the environmental benefits of the intervention, and the situation becomes worse than it was from the outset (known as the Jevons paradox (Alcott, 2005)). This means that measures for reduced food waste do not always only produce the desired results with regard to environmental impact, but also more unintended side effects.
This does not mean that measures to reduce food waste are ineffective, but that second order effects need to be taken into account. Otherwise, there is a risk that interventions might not be efficient in a systems perspective. Due to the complexities involved in considering full systemic effects, the practicality of detailed analysis must be weighed up for each intervention. The use of theory-based interventions, with extended logic mapping (e.g. with systems mapping as discussed above) will be useful in enabling this detailed analysis, as the theoretical background and logic mapping may be able to acknowledge cross-boundary input and outcomes (but not necessarily assist with measuring them).
Ideally, Intervention studies, where possible, should collect data to monitor these second-order effects, in addition to monitoring the direct impact on food waste. However, as this may involve recording household spending (on food as well as other expenditure) and food consumption, it will greatly inflate the cost of studies and may not be possible. Another option is to, at least, identify risks for second order effects, look for ways to minimize negative second-order effects and maximize any potential positive effects of this nature.

Policy implications
According to our review, in spite of the shortage of downstream intervention studies, there are still several evaluated interventions that have good potential for use in a wider context. These include so-called "low hanging fruits" which might not have a huge impact but also do not imply high cost, high maintenance or side effects, or interventions that have been assessed and have produced good results. One example of the former kind is to encourage guests at restaurants and in largescale households to adjust the portions to how hungry they are (Jagau and Vyrastekova, 2017), or to take smaller portions at a buffet and come back if you want more (Kallbekken and Saelen, 2013). This kind of measure is relatively simple and inexpensive and could be combined with other measures, such as for example a lower price for a smaller portion. Examples of the latter kind, assessed with good results but with an economic cost, are the interventions with smaller plates (Kallbekken and Saelen, 2013;Wansink and van Ittersum, 2013).
A number of interventions use social media (e.g. Lim et al., 2017) and the evaluated studies indicate that there is potential for this in particular as a way of spreading knowledge and creating discussion and reflection. However, caution must be taken as using social media to message the correct audience with content that resonates has its own challenges due to audience segmentation. Another intervention that is quite simple and can be done without major investment in apps, is colour coding of shelving or sections in the refrigerator (Farr-Wharton et al 2012). Similar initiatives have been tested in "Food: Too good to waste" where the solution was even easier -with just a note in the fridge on food to be eaten soon (U.S. EPA Region 10, 2016). More extensive campaigns (e.g. U.S. EPA Region 10, 2016; WRAP, 2013b) have also had good effects, although it is difficult to estimate the impact of individual components of the overall campaign. With a mix of complementary interventions and actors at local level, this type of measure should have good potential given that the necessary resources and commitment, which seems to have been the case in both the UK and the United States.

Conclusion
This paper has summarised 17 applied food-waste prevention interventions at the consumption/consumer stage of the supply chain via a rapid review of academic literature from 2006 to 2017. This led to the identification of interventions that could be deployed effectively at scale in the home (e.g. fridge colour coding, product labelling, and information provision), and out of the home (e.g. plate and portion size adjustment, changes to menus and nutritional guidelines, and redesign of class room syllabus).
Our discussion has identified the weaknesses of the current literature; proposed guidelines for the development of further food waste interventions, and set out an agenda for further research: • Well-designed interventions covering a range of types (including longer interventions and those exploring a raft of measurers), • Tested using carefully selected methods to understand the outcome of the intervention and how it works (or not), • Adoption of higher sample sizes and representative sampling for quantitative elements, • Replication studies in different countries • Consideration of systemic effects • Improved, more consistent reporting. This is a novel and important addition to the researchers', policymakers' and practitioners' tool kit. Our review found that the majority of current interventions achieve only a 5-20% reduction in food waste. To achieve Sustainable Development Goal 12.3 by 2030, (halve per capita global food waste at the retail and consumer levels) these interventions (and others) need to be combined, refined, tested further at different scales and geographies, and adopted on a global scale.