论文信息

Gabriel Vieria Mariano / Aldara Da Silva César / Marco Antonio Conejero 等 6 人

Frontiers in Sustainable Food Systems

论文解读

AI 辅助整理
方法
单案例研究,2025年称重记录餐盘浪费与剩余,SIPOC与泳道图映射流程。
发现
预测准确未阻止浪费;指标量化损失量,流程映射定位损失来源。
机会点
可将该指标-流程耦合方法推广至其他地区公共餐饮,并实证检验六个机制假设。

论文摘要

引言 餐饮服务产生了全球约28%的食物浪费,但有关公共资助餐饮中预防措施的证据主要集中在欧洲情境,拉丁美洲的相关研究十分匮乏。浪费指标量化了损失的数量,却未揭示损失产生的环节。方法:对巴西一家公共补贴社区餐厅进行单案例研究,该餐厅每日供应2,000份餐食,其中午餐1,200份、晚餐300份,就餐人群约80%为老年人,每份餐食的各个组成部分均由工作人员分装。2025年全年使用校准秤对每日餐盘浪费(plate waste,PW)和清洁剩餐(clean leftover,CL)进行记录,按餐次分层,与菜单及预测供餐量与实际供餐量进行交叉核对,并筛查是否同时超出该单位的内部参考值(PW 3%,CL 2%)。定性证据来自非参与式观察及对该单位管理者和生产营养师的集体访谈,访谈内容经转录并进行主题编码(69个单元,十二个类别)。通过SIPOC和泳道图将损失归因于流程阶段,并依据设计科学研究方法与从业人员共同评估所得制品。结果 该单位在374,604份供餐中记录了7.68吨PW和CL,即每餐20.49克。午餐占5.901吨,晚餐占1.776吨,但晚餐强度更高,每餐23.55克对19.72克,且PW峰值超过内部参考值三倍以上,而CL峰值仅为偶发。在九个供餐日观察中,两项指标均被超出,尽管预测偏差范围为零至五份餐食。流程映射将这些情况归因于批次规模、供餐末补充、分餐及厨房与配送沟通。讨论 准确的预测并未防止损失升高:指标确定了损失的数量,其与流程映射的结合定位了损失产生的环节。由于上述机制均未被测量,故以六项假设的形式提出,每项假设均附有检验所需的工具。作为单案例研究,其推广属于分析性而非统计性主张,以六项操作前提为条件(包括标准化菜单和成文的标准操作程序、批次生产、使用校准秤进行常规称重,以及PW和CL分开记录),所提出的措施按分阶段路线图排序,待与该单位管理层重新评分。

全文

Frontiers in Sustainable Food Systems

Frontiers in Sustainable Food Systems 01 frontiersin.org
Coupling food-waste indicators
with SIPOC and swimlane
process mapping to locate
avoidable waste in large-scale
public catering: an exploratory
single-case study in a Brazilian
community restaurant
Gabriel Vieria Mariano 1, Aldara Da Silva César 1*, Marco
Antonio Conejero 1, Nathan Ribeiro Messias 1, Lucas
Rodrigues Deliberador 2 and Ricardo César da Silva Guabiroba 1
1Departamento de Engenharia de Agronegócios, Universidade Federal Fluminense, Niterói, Brazil,
2Faculdade de Engenharia – FAEN, Universidade Federal da Grande Dourados, Dourados, Brazil
Introduction: Food service generates around 28% of global food waste, yet evi-
dence on prevention in publicly funded catering is concentrated in European
settings and scarce for Latin America. Waste indicators quantify how much is
lost without revealing where the loss originates. Methods: Single-case study of
a publicly subsidised community restaurant in Brazil, serving 2,000 meals daily,
amongst them 1,200 lunches and 300 dinners, to a clientele approximately 80%
elderly, with every component plated by staff. Daily plate waste (PW) and clean
leftover (CL) records, weighed on calibrated scales throughout 2025, were strati-
fied by meal service, cross-checked against menu and forecast-versus-served
counts, and screened for simultaneous exceedance of the unit’s internal refer -
ence values (PW 3%, CL 2%). Qualitative evidence came from non-participant
observation and a collective interview with the unit’s manager and production
nutritionist, transcribed and thematically coded (69 units, twelve categories).
Losses were referred to process stages through SIPOC and swimlane mapping,
and the resulting artefacts were evaluated with the practitioners following design
science research.
Results: The unit recorded 7.68 t of PW and CL across 374,604 meals served,
or 20.49 g per meal. Lunch accounted for 5.901 t and dinner 1.776 t, yet dinner
showed the higher intensity, 23.55 against 19.72 g per meal, and PW peaks
exceeding three times the internal reference value, whereas CL peaks were spo-
radic. On nine service-date observations both indicators were exceeded although
forecast deviation ranged from zero to five meals. Process mapping referred
these occasions to batch sizing, end-of-service replenishment, portioning and
kitchen-distribution communication.
Discussion: Accurate forecasting did not prevent elevated losses: the indica -
tors established how much was lost; their coupling with process mapping
located where it arose. Because none of the mechanisms was measured, they
are stated as six hypotheses, each with the instrument required to test it. As a
single case, transfer is an analytical and not a statistical claim, conditional on six
operational prerequisites (including standardised menus and documented SOPs,
TYPE Original Research
PUBLISHED 16 November 2026
DOI 10.3389/fsufs.2026.1981661
OPEN ACCESS
EDITED BY
Lilia Arely Prado Barragan,
Metropolitan Autonomous University,
Mexico
REVIEWED BY
Mohammad Heydari,
Unitec Institute of Technology,
New Zealand
Bosompem Ahunoabobirim Agya,
Hochschule Nordhausen, Germany
*CORRESPONDENCE
Aldara Da Silva César
aldaracesar@id.uff.br
RECEIVED 28 August 2026
REVISED 14 October 2026
ACCEPTED 19 October 2026
PUBLISHED 16 November 2026
CITATION
Mariano GV, César ADS, Conejero MA,
Messias NR, Deliberador LR and
Guabiroba RCdS (2026) Coupling
food-waste indicators with SIPOC and
swimlane process mapping to locate
avoidable waste in large-scale public
catering: an exploratory single-case
study in a Brazilian community
restaurant.
Front. Sustain. Food Syst. 10:1981661.
doi: 10.3389/fsufs.2026.1981661
COPYRIGHT
© 2026 Mariano, César, Conejero,
Messias, Deliberador and Guabiroba.
This is an open-access article distributed
under the terms of the Creative
Commons Attribution License (CC BY).
The use, distribution or reproduction in
other forums is permitted, provided the
original author(s) and the copyright
owner(s) are credited and that the
original publication in this journal is
cited, in accordance with accepted
academic practice. No use, distribution
or reproduction is permitted which does
not comply with these terms.

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 02 frontiersin.org
batch production, routine weighing on calibrated scales, and separate PW and
CL recording), and the proposed actions are ordered in a phased roadmap to be
re-scored with the unit’s management.
KEYWORDS
clean leftovers, design science research, food waste prevention, plate waste, process
mapping, public catering, single-case study, SIPOC
1 Introduction
Food waste (FW) is a resource-efficiency problem before it is a
disposal problem: every wasted meal embodies the raw materials,
energy, water and labour invested in producing it, so waste preven -
tion is inseparable from the efficient use of resources along the
food production chain ( Çelekli et al., 2024 ; United Nations
Environment Programme (UNEP), 2024 ). Approximately 1.05 bil -
lion tonnes of food were wasted worldwide in 2022, and the food
service sector accounted for around 28% of this total ( United
Nations Environment Programme (UNEP), 2024 ). Within the
waste management hierarchy, prevention should be prioritised for
avoidable food waste, whereas food waste that remains after pre -
vention efforts may subsequently be directed towards recovery and
valorisation pathways, including conversion into useful products
or energy ( Heydari et al., 2025 ). Target 12.3 of the Sustainable
Development Goals (SDGs) commits countries to halving per
capita food waste at the retail and consumer levels by 2030 ( United
Nations, 2015 ). Food service is therefore a priority setting for pre -
vention because most of the waste it generates is avoidable rather
than intrinsic to food preparation, and because avoidable waste
corresponds directly to resources that were processed, transformed
and distributed without ever generating value for the end
consumer.
Recent research has broadened the food-waste mitigation
agenda beyond conventional disposal-oriented approaches, high -
lighting complementary strategies across the food supply chain.
These include integrated prevention and approaches aimed at
improving measurement and resource efficiency (Heydari, 2024), the
valorisation of unavoidable food-waste streams through energy-
recovery pathways (Heydari et al., 2025 ), and digital technologies
such as blockchain, intelligent packaging, biosensors and Internet of
Things systems to enhance traceability, monitoring and real-time
decision-making ( Li et al., 2026 ; Rahman et al., 2026 ). Together,
these developments illustrate a shift towards more integrated food-
waste management systems that combine prevention, circular
resource use and data-enabled process control. Within this broader
agenda, prevention at source remains a foundational principle across
multiple stages of the food supply chain, including production, pro-
cessing, retail and consumption, where avoidable waste can be
addressed before downstream recovery or valorisation becomes
necessary.
Sector evidence supports this priority. A study covering 1,189
kitchens in four European countries estimated that approximately 20%
of the food served in the catering sector was wasted, with substantial
variation across establishment types and operational contexts
(Malefors, 2021), whilst Waste and Resources Action Programme
(WRAP) (2013) estimated that around 75% of the waste generated by
hospitality and food service is preventable. Public catering has
attracted particular attention because it is financed from public bud-
gets and operates under standardised procedures that make waste
both measurable and manageable ( Betz et al., 2015 ; Malefors et
al., 2022).
In Brazil, large-scale Food and Nutrition Units (FNUs), particu-
larly publicly subsidised community restaurants known locally as
‘restaurantes populares, ’ provide nutritionally balanced meals at low
cost to socially vulnerable populations (Canonico et al., 2014; Brasil,
2023). Waste in these facilities is simultaneously an environmental
burden, an operating cost and a misallocation of public funds
intended to promote food and nutrition security; its reduction has
been incorporated into Brazil’s national policy framework, initially
through the Intersectoral Strategy for the Reduction of Food Loss and
Waste and, more recently, through the II Intersectoral Strategy and
the National Policy to Combat Food Loss and Waste (Brasil, 2018;
Brasil, 2025a; Brasil, 2025b). Systematic monitoring of plate waste
(PW) and clean leftovers (CL) is therefore both a waste management
task and a matter of public accountability ( Grandhi and Appaiah
Singh, 2015).
FW indicators, however, quantify how much is lost without
revealing where in the production system the loss originates. Process
mapping addresses this gap: it is a process-engineering diagnostic
tool, developed and used across manufacturing and service indus -
tries to represent the sequence of activities, the actors responsible
for each task and the interfaces between operational stages, so that
waste ‘hotspots’ in a production chain can be identified and correc -
tive actions targeted rather than generic (Harrington, 1991; Steur et
al., 2016; Wesana et al., 2019; Martins et al., 2022). Applied to a food
production system, this diagnostic logic treats the kitchen as what it
is from an engineering standpoint: a batch production chain in
which inputs are transformed, moved and allocated through a
sequence of operational stages, each a potential point of
resource loss.
Previous research has increasingly examined FW from a pro -
cess-oriented perspective, identifying waste-generating stages and
operational drivers across food-service and hospitality systems
(Derqui et al., 2016; Filimonau and De Coteau, 2019). These studies
have demonstrated the importance of examining food waste across
interconnected operational stages rather than treating it solely as an
aggregate outcome. Building on this literature, the present study
focuses on a more specific integration: routinely collected PW and
CL indicators are combined with formal SIPOC and swimlane pro -
cess mapping to relate observed waste patterns to operational stages
and interfaces in a high-capacity public catering facility. The
intended contribution is therefore not the introduction of process-
oriented food-waste analysis itself, but the application of these com-
plementary process-engineering tools as a structured diagnostic
protocol using data already generated through routine operational
monitoring.

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 03 frontiersin.org
This study is designed as a single-case study and analyses 1 year
of PW and CL records from a high-capacity community restaurant in
the countryside of Rio de Janeiro State, Brazil. The current study inte-
grates them with SIPOC and swimlane mapping, which is a process-
engineering technique for tracing material and information flow
through a production chain, to identify the operational stages that
generate avoidable waste and derive resource-efficient prevention
actions.
The study’s intended contribution is twofold: (i) a diagnostic pro-
tocol that pairs process mapping with waste indicators to locate loss-
generating activities in a batch food production chain, and (ii) a
waste-minimisation indicator framework—the joint stratification of
PW and CL by service, cross-checked against forecast accuracy and
combined into a ‘fully critical day’ criterion based on simultaneous
exceedance of site-specific internal reference values that was com -
puted entirely from data already generated by the facility’s routine
operational monitoring.
2 Background
2.1 Food waste in food service: definitions
and categories
Food loss and food waste are distinct phenomena that reduce
food availability along the supply chain. Food loss occurs during
production, postharvest handling, storage, transport and process -
ing, before food reaches the retail sector, whereas food waste
occurs at the retail, food service and household stages, after food
has become available for human consumption ( FAO—Food and
Agriculture Organization of the United Nations, 2011 ; FAO—Food
and Agriculture Organization of the United Nations, 2019 ; United
Nations Environment Programme (UNEP), 2024 ). This distinction
underpins the separate targets defined under SDG 12.3 and deter -
mines which prevention strategies are applicable at each stage.
Food waste is conventionally classified into three categories:
avoidable waste, comprising food that could have been consumed
had planning, storage or handling been better; potentially avoid -
able waste, comprising edible fractions whose consumption
depends on individual habits and cultural preferences, such as
peels, stalks and bread crusts; and unavoidable waste, comprising
inedible fractions such as bones and eggshells ( Richter and
Bokelmann, 2016 ).
This distinction matters for waste management because only the
first two categories respond to operational and behavioural interven-
tion. Since approximately 75% of the waste generated by the food ser-
vice sector is preventable (Waste and Resources Action Programme
(WRAP), 2013 ), prevention at source offers substantially greater
potential for resource efficiency in this sector than downstream treat-
ment or recovery.
In the present study, this classification requires consideration of
the measurement boundaries of the two waste streams analysed. CL
consists of prepared food remaining unserved at the end of service
and is therefore treated as avoidable waste. PW , however, is recorded
as an aggregate mass without compositional sorting and may include
avoidable, potentially avoidable and unavoidable fractions.
Consequently, the avoidable proportion of recorded PW cannot be
determined from the available data.
2.2 Drivers of food waste in food service
units
FW in food service is a multifactorial phenomenon. It arises from
operational and managerial deficiencies, individual behaviour, and
communication failures amongst actors, and occurs throughout stor-
age, preparation, distribution, and consumption, with distribution
losses and plate waste consistently identified as the two dominant
sources (Betz et al., 2015; Richter and Bokelmann, 2016).
In FNUs specifically, waste has been associated with inadequate
meal planning, inappropriate portioning practises, insufficient staff
training, poorly planned procurement and weak operational control,
indicating that it cannot be attributed to consumer acceptance alone
(Canonico et al., 2014 ; Silvério and Oltramari, 2014 ; Martins et
al., 2022).
2.3 Food waste indicators in food and
nutrition units
Two indicators are routinely used in FNUs, and together they
form a simple waste-minimisation indicator set for food production
chains. Clean leftovers refer to prepared food that remains unserved
at the end of the service period, whereas plate waste refers to food
served to consumers but returned uneaten on trays or plates (Betz et
al., 2015; Grandhi and Appaiah Singh, 2015; Parisoto et al., 2013).
From a process perspective, PW is located at the consumption inter-
face because it occurs after food has been served. CL, however, should
not be attributed exclusively to production or batch-sizing decisions.
Depending on the operational system and measurement boundary,
CL may result from several interconnected mechanisms, including
overproduction, inappropriate batch sizing, excessive or late serving-
line replenishment, and mismatches between the quantity transferred
to the serving line and residual demand. Accordingly, CL is inter -
preted here as an indicator associated with the production—distribu-
tion interface rather than as evidence of a single upstream cause. These
operational definitions should be interpreted in conjunction with the
avoidability classification established in Section 2.1, particularly the
distinction between avoidable CL and the compositionally unclassified
PW stream.
Joint monitoring of both indicators has been recommended as a
management tool in Brazilian community restaurants, since each cap-
tures losses generated at a different stage of the process, a property that
makes the pair useful as a diagnostic indicator set, and not only as a
compliance metric (Messias et al., 2025).
PW is calculated as the percentage of food returned by consumers
relative to the total weight of food served ( Martins et al., 2022), as
shown in Equation (1), where PR is the weight of food returned by
consumers and PA is the total weight of food served.

( )= ×% 100PRPW PA (1)
Recent systematic evidence further indicates that PW varies sub-
stantially according to the institutional and operational characteristics
of food services. A systematic review and meta-analysis by Guimarães
et al. (2024a) identified food-service type and distribution system as
important factors influencing both percentage and per capita plate
food waste. A complementary systematic review and meta-analysis
focusing on waste-reduction strategies showed that intervention

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 04 frontiersin.org
effectiveness also varies according to food-service type, management
modality, distribution system and the target of the intervention
(Guimarães et al., 2024b). These findings reinforce the importance of
interpreting food-waste indicators within their specific operational
and institutional contexts rather than assuming direct comparability
across food-service settings.
Although percentage-based classifications for PW have been pro-
posed in the food-service literature (Aragão, 2005), such classifica-
tions are not treated in this study as validated or universally applicable
diagnostic thresholds. No Brazilian regulatory authority establishes
official percentage thresholds for PW or CL. Therefore, the present
analysis relies on the facility’s pre-established internal reference values
of 3% for PW and 2% for CL. These values are used exclusively as site-
specific managerial conventions for operational monitoring and pro-
cess analysis and should not be interpreted as regulatory standards,
validated diagnostic cut-offs, or benchmarks transferable to other
food-service settings.
Evidence from Brazilian community restaurants also illustrates
the use of locally defined operational reference values. Messias et al.
(2025), in a case study conducted at a community restaurant in the
state of Rio de Janeiro, reported that the management contract also
established internal limits of 3% for PW and 2% for CL. These values
were used by the facility for routine operational monitoring rather
than being derived from a national regulatory standard. This case sup-
ports treating such percentages as context-specific managerial refer-
ence values, whose interpretation depends on the operational
characteristics and monitoring practises of each establishment.
Beyond the mass diverted to disposal, PW and CL are themselves
resource-efficiency indicators: both quantify raw materials, labour,
water and energy that were processed and transformed but never con-
verted into consumed output, so reducing either indicator delivers
simultaneous economic and environmental resource savings (FAO—
Food and Agriculture Organization of the United Nations, 2013 ;
FAO—Food and Agriculture Organization of the United Nations,
2014; Malefors et al., 2022).
2.4 Process mapping in food service
operations
Process mapping is a core process-engineering diagnostic tool for
any production system, food-related or not: it provides a systematic
representation of how raw materials and other inputs flow through a
sequence of operational stages until they become finished output; in
this case, ready-to-serve meals. By making explicit the sequence of
activities, the actors responsible for each task and the interfaces
between stages, it supports the identification of process failures, inef-
ficiencies and waste ‘hotspots, ’ independently of the specific industry
in which it is applied (Harrington, 1991; Wesana et al., 2019; Abreu et
al., 2023).
Process-oriented perspectives have also become increasingly rel-
evant in food-service research. Derqui et al. (2016) examined food
waste across the food-service supply chain and highlighted that waste
arises from multiple interconnected stages and operational drivers,
including storage, preparation, demand forecasting, portioning, and
the handling of food that is prepared but not served, or served but not
consumed. Similarly, Filimonau and De Coteau (2019) proposed a
managerial framework for food-waste mitigation across different areas
of hospitality operations, emphasising the need to understand both
the scale of waste and the operational causes underlying its generation.
Together, these studies demonstrate that process-oriented analysis is
already an established perspective in food-waste research. The present
study builds on this body of work by combining routinely recorded
PW and CL indicators with formal SIPOC and swimlane mapping to
structure the diagnosis of waste-generating stages and operational
interfaces in public catering.
The SIPOC framework (Supplier, Input, Process, Output,
Customer) is a generic process-engineering tool, originating in Lean
Six Sigma and quality-management practise, that defines process
boundaries and provides a high-level view of the actors and material
flows involved, prior to any waste-specific analysis ( Antony et al.,
2012), consistent with the process management principles of cross-
functional coordination and value creation (Silva et al., 2012). Cross-
functional flowcharts, or swimlane diagrams, complement SIPOC by
assigning each activity to the sector that performs it, translating the
high-level SIPOC boundary into an operational-level map of who
does what, where, and with what handoffs—the same diagnostic logic
used to locate hotspots and bottlenecks in any multi-stage produc-
tion chain.
When process maps are combined with waste indicators, quanti-
fied losses can be attributed to specific activities, such as communica-
tion failures between the kitchen and the serving line, delayed or
excessive container replenishment, and overproduction of specific
batches, allowing corrective action to be targeted at the resource-inef-
ficient step rather than generic (Betz et al., 2015; Wesana et al., 2019;
Martins et al., 2022). This is the same attribution logic applied in pro-
cess audits of production or processing lines, where a yield-loss indi-
cator only becomes actionable once it is mapped onto the specific
operational stage responsible for the loss.
3 Methodology
The study was designed as an applied, descriptive and proposi-
tional single-case study with a mixed-methods approach, conducted
in three sequential phases: (Phase I) process analysis and mapping;
(Phase II) quantification and analysis of waste indicators; and (Phase
III) development of prevention proposals. An exploratory literature
review of peer-reviewed publications and technical reports from FAO,
UNEP , WRAP and Brazilian government agencies preceded the field-
work and informed the selection of indicators, the classification of
waste categories and the mapping techniques applied.
3.1 Research design, case selection and
basis for generalisation
The research design is a single-case study, adopted because the
phenomenon of interest, how waste indicators recorded as part of a
routine can be converted into targeted prevention actions, is insepa-
rable from the operational context in which those indicators are pro-
duced (Yin, 2018).
The facility was selected following a replication rather than a sam-
pling logic (Conejero et al., 2024): it is an exemplary instance of a
high-capacity public catering unit that weighs PW and CL daily on
calibrated scales, records the two fractions separately in a manage-
ment system, retains a full year of historical records, operates with
standardised menus and documented standard operating procedures,
and produces meals in batches. These features make the case

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 05 frontiersin.org
information-rich for the research problem; they also delimit the set of
units to which the findings may be extended.
Accordingly, the study makes no claim to statistical generalisation.
A case study is generalisable to theoretical propositions rather than to
populations: the inference runs from the case to a conceptual claim at
a higher level of abstraction, not from a sample to a universe, and Yin
(2018) stresses that such an analytic generalisation must be advanced
as an argued claim rather than obtained by a formula. Eisenhardt
(1989) grounds its warrant in the systematic comparison of the emer-
gent concepts with prior literature, which sharpens generalisability
and the definition of constructs.
Flyvbjerg (2006) argues that the limitation of the single case con-
cerns formal generalisation specifically, which is itself overrated as the
main source of scientific progress, whilst the ‘force of example’ is
underestimated. Siggelkow (2007) adds that a paper based on a single
case cannot stand on descriptive grounds alone and must deliver a
conceptual insight whose internal logic is convincing independently
of the site.
For applied and evaluative research, Simons (2015) proposes a
complementary criterion: generalisation from a single case is natural-
istic and situated, arising when readers recognise similarities and dif-
ferences between the reported case and their own setting, which
requires that the case be described in sufficient contextual detail for
that judgement to be possible. This article therefore reports the opera-
tional conditions of the case in detail, restricts its claims to the diag-
nostic logic and the proposed mechanisms, and states the scope
conditions for application elsewhere only after presenting the results.
Although the single-case study is the overarching research strat-
egy, the components produced by the study—the SIPOC representa-
tion, the cross-functional process map, the indicator protocol
including the ‘fully critical day’ screening criterion, and the action
plan—are prescriptive artefacts rather than descriptive findings, and
their assessment followed the logic of Design Science Research (DSR).
In design science, the object of research is a purposeful artefact whose
utility, quality and efficacy must be demonstrated through explicit
evaluation, which Hevner et al. (2004) state as Guideline 3 of design-
science research and locate in the requirements of the business envi-
ronment in which the artefact is to be used.
Van Aken (2004) characterises the corresponding output of man-
agement research as a field-tested and grounded technological rule,
that is, a general prescription that is grounded in evidence and tested
in a real organisational setting rather than deduced from description
alone. The DSR methodology of Peffers et al. (2007) organises this into
six activities: problem identification and motivation, definition of
solution objectives, design and development, demonstration, evalua-
tion, and communication.
The phases of the present study map onto them as follows: the
literature review and the initial diagnosis correspond to problem iden-
tification and to the definition of objectives; the SIPOC, the process
map and the indicator protocol correspond to design and develop-
ment; their application to the 2025 series corresponds to demonstra-
tion; the feedback meeting with the facility staff corresponds to
evaluation; and this article corresponds to communication.
The evaluation conducted here is observational and descriptive in
the terms of Hevner et al. (2004): the artefacts were applied in the case
setting and then submitted to the practitioners who would use them,
so that their perceived usefulness and the adjustments required for use
could be established, rather than being tested for effectiveness against
a counterfactual. Following Holmström et al. (2009), this places the
study in the exploration part of the design-science cycle—solution
incubation followed by a first round of solution refinement—and not
in the subsequent explanatory phases, in which a solution design is
tested repeatedly and theorised. One site and one evaluation round
constitute an initial field test rather than a repeated one, which is the
principal reason why the prescriptive claims made here are bounded
by explicit scope conditions and why the proposed causal mechanisms
are stated only as hypotheses.
3.2 Phase I: process analysis and mapping
The facility was selected purposively because it is a high-capacity
unit that already monitors waste indicators daily with calibrated
scales, retains a full year of historical records and granted access to its
operational procedures. These characteristics make it an information-
rich case for examining how routine waste data can be converted into
prevention actions, a condition that is increasingly common in public
catering systems and is therefore relevant beyond this single site.
The unit distributes approximately 2,000 meals per day, compris-
ing 500 breakfasts, 1,200 lunches and 300 dinners. Breakfast consists
of coffee and bread only and generates no measurable organic waste,
so the indicator analysis reported here covers the lunch and dinner
services. Meals are plated by the serving staff: no component of the
main meal is self-served, and the dessert, although collected by the
consumer, is pre-portioned at one unit per person under staff supervi-
sion. According to the demographic study carried out by the state
government before the unit opened and to the management’s own
characterisation of its clientele, approximately 80% of users are elderly
and the remaining 20% are commerce workers and people in situa-
tions of homelessness. These features—the fourfold volume asymme-
try between lunch and dinner, the staff-plated service model and the
predominantly elderly user profile—form the operational context in
which the indicator values must be read, and they are amongst the
conditions that delimit transfer to other settings.
More detailed contextual data on staffing levels and their distribu-
tion across meal services, the facility’s operating budget, and service-
specific demographic composition were not available to the research
team. Although menu information was available for the service dates
examined in the operational records, the dataset did not provide a
structured characterisation of the menu cycle suitable for systematic
comparison between lunch and dinner. These contextual dimensions
were therefore not included as explanatory variables in the analysis.
Two qualitative data-collection procedures were carried out
during a technical visit to the facility in March 2025. They are reported
separately because they differ in participants, instrument, form of
record and analytical use, and because they produce evidence of dif-
ferent kinds: observation documents what is done, whereas the inter-
view documents how decisions are made and justified. A review of the
unit’s Standard Operating Procedures (SOPs) was conducted in the
same visit as a third, documentary source.
The first procedure was non-participant direct observation of the
operational routine. The researcher followed the production flow
through its successive stages—receiving, storage, pre-preparation,
cooking, meal distribution, and waste segregation, weighing and dis-
posal—observing the activities performed by the staff of the storage,
pre-preparation, cooking and distribution sectors, the sequence in
which tasks were executed, the handoffs between sectors, the portion-
ing practise at the serving line, the criteria applied when replenishing
food containers, and the procedure for weighing and recording PW

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 06 frontiersin.org
and CL. Seven employees from these sectors were observed during
their normal work. No questionnaire was administered to them and
no task was interrupted; observation was structured by process stage
and the record consisted of field notes taken during and immediately
after each observed stage.
The second procedure was a single collective semi-structured
interview with the unit’s manager and the production nutritionist, the
two professionals responsible for demand planning and the definition
of the forecast number of meals per service, for menu planning, and
for supervising the recording of the waste indicators and compliance
with the internal reference values. The interview applied the twelve-
topic guide reproduced in Supplementary Material S1, which covers
the characterisation of the service and its planned capacity per meal;
the operational flow from receiving to disposal and the responsibilities
of each sector; demand planning and the definition of the forecast
number of meals; the weighing, recording and reporting routines for
PW and for CL; the internal reference values and how compliance is
monitored; container replenishment at the serving line and the criteria
for ending production and service; perceived operational difficulties;
the documentation and SOPs related to waste reduction; waste segre-
gation and final disposal; and the waste-reduction actions already
practised by the team. The interview was conducted by two members
of the research team, lasted 27 min, was audio-recorded with the par-
ticipants’ consent and transcribed in full with Transkriptor, yielding
261 conversational turns.
The qualitative material—the interview transcript, the observation
field notes and the SOPs—was analysed by thematic content analysis
in four steps. First, the transcript and the field notes were read repeat-
edly to establish familiarity with the material. Second, the transcript
was segmented into thematic units, each corresponding to one opera-
tional decision, routine, responsibility, quantity or difficulty reported;
the 261 turns yielded 69 substantive units attributable to the two inter-
viewees. Third, these units were coded into ten categories defined a
priori—the eight stages of the production process (demand planning
and forecasting; receiving and storage; pre-preparation; cooking and
batch sizing; serving-line replenishment; portioning at the consump-
tion interface; indicator weighing and recording; and waste segrega-
tion and disposal), the internal reference values and their compliance
monitoring, and the characterisation of the service, staffing and
SOPs—to which two categories were added inductively during coding,
namely kitchen—distribution communication and perceived con -
sumer behaviour.
For consistency, the following operational terms are used through-
out the study. “Batch sizing” refers to the determination of the quantity
of food prepared or released in each production batch. “Portioning”
refers to the amount of each menu component placed on the con -
sumer’s plate during service. “Replenishment” refers to the transfer of
additional food containers from the kitchen to the serving line during
service. “Distribution interface” refers to the operational boundary
between kitchen production and the serving line, where batch release,
replenishment, portioning and communication between teams occur.
Each unit received up to three codes, the first being the dominant
category, and was additionally flagged according to whether it stated
a concrete operational fact, routine, quantity or constraint; 47 of the
69 units were so flagged. The categories most frequently coded were
the characterisation of the service and SOPs, demand planning and
forecasting, serving-line replenishment, and perceived consumer
behaviour. Fourth, the coded categories were consolidated into the
process representations, so that every activity, responsibility and
interface shown in the SIPOC framework and in the swimlane map is
traceable to at least one coded unit from the transcript, the field notes
or the SOPs, maintaining a chain of evidence between the raw material
and the representations derived from it (Yin, 2018).
The complete coding matrix is provided as Supplementary Material S2
and the codebook, with the definition and the decision rule of each cat-
egory, as Supplementary Material S3, so that the derivation of every cat-
egory can be inspected. The initial coding matrix was produced with the
assistance of a large language model and was subsequently verified, cor-
rected and approved unit by unit by the authors, who take the responsi-
bility for the final coding; this use is declared in the statement on
generative artificial intelligence. Coding was performed by one analyst
and reviewed in research-team meetings; no independent second coder
was used and no inter-coder agreement statistic was computed, a limita-
tion stated in the conclusions.
The information collected was organised using two complemen-
tary process-engineering techniques. SIPOC defined the boundaries
of the production process by identifying suppliers, inputs, key activi-
ties, outputs and customers and highlighted the stages with the great-
est potential for generating losses.
A cross-functional process map (swimlane diagram) was then
constructed based on the field notes, the semi-structured interview,
the SIPOC analysis and the SOPs, representing the activities per -
formed by the storage, pre-preparation, cooking, distribution and
administrative sectors and the interfaces between them. The map was
presented to the facility staff at a feedback meeting held in June 2026
to verify its consistency with actual operational routines.
The study addressed organisational processes and operational per-
formance indicators only. No personal, sensitive or personally identifi-
able information was collected, no intervention involving participants
was performed, and the anonymity of both the participants and the
institution was fully preserved.
3.3 Phase II: quantification and analysis of
waste indicators
Daily PW and CL data from January to December 2025 were
obtained from the facility’s operational records rather than measured by
the researchers. The unit weighs both fractions daily on calibrated scales
and records them in the Teknisa food service management system,
together with the number of meals forecast and the number actually
served. PW and CL percentages were calculated as 100 × recorded waste
mass (kg)/[number of meals actually served × 0.70 kg per meal], where
0.70 kg is the nominal portion weight specified in the contract between
the Municipality and the contracted food-service provider, reflecting the
client’s requirement. This contractual standard was used as the reference
denominator and does not imply that the actual mass served at every
meal was independently measured or consistently equal to 0.70 kg. Waste
mass came from the unit’s weighing records and was not inferred from
indicator percentages. Three CL records reporting zero meals served
were excluded from percentage-based summaries because their denomi-
nator was zero; one dinner PW record reporting 6.15 kg for one meal
(878.57%) was also excluded as implausible under the nominal portion
convention. No percentage was imputed. The recorded masses were
retained in annual mass totals, and zero CL values in December were
treated as measured zeros.
The supplied dataset contained 498 service-date observations cov-
ering 249 dates for each meal service. No blank PW or CL mass fields

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 07 frontiersin.org
were identified within these records. However, the number of operat-
ing days omitted from the supplied series could not be established
without reconciliation against the facility’s operating calendar. Cross-
checking the PW and CL reports identified nine discrepancies in
served meal counts across eight dates. These discrepancies were pre-
served as recorded and were not independently confirmed as record-
ing errors. The PW and CL percentages were evaluated using the meal
count recorded in their respective source reports; discrepant counts
were not automatically substituted. Zero values were retained as
recorded rather than imputed, including the PW zeros recorded at
both services from 24 to 28 November 2025. Their status as genuine
zero measurements or recording omissions could not be indepen -
dently verified. These checks assess the completeness of fields and
consistency between the supplied reports, rather than validating the
original measurements or proving complete coverage of operat -
ing days.
PW and CL were interpreted against the restaurant’s pre-estab-
lished internal reference values of 3 and 2%, respectively. In this study,
they were retained solely as internal reference points to identify devi-
ations from the facility’s expected operating conditions. Days on
which an indicator exceeded its respective internal reference value
were therefore classified as site-specific operational deviations for
analytical purposes. For exploratory analytical purposes, days on
which both indicators simultaneously exceeded their respective site-
specific internal reference values within the same meal service were
classified as “fully critical days” . This conjunctive criterion was used
as a simple screening tool to identify occasions when elevated losses
occurred simultaneously at both the production and consumption
stages. It should not be interpreted as a validated diagnostic metric or
as evidence that simultaneous threshold exceedance is inherently
more meaningful than alternative approaches, such as combined
waste mass, percentile-based thresholds, or other multivariate
criteria.
Two features of the unit’s recording practise delimit what the
indicators measure and are stated here because they bound the
interpretation of the results. First, the CL figure recorded by the
unit corresponds to the fraction remaining in the serving-line con -
tainers at the end of service, which, in this facility, is discarded
rather than reused; food that was produced but retained in the
kitchen and never transferred to the line is not recorded. The
recorded CL therefore measures a loss at the production-to-line
transfer, and not the full quantity of food prepared but not distrib -
uted. Second, waste generated during pre-preparation, in particular
in the preparation of salads, is not weighed, so it lies outside the
measurement boundary, and the indicator percentages are pro -
duced automatically by the management system, which applies
stored average values to the mass entered by the staff; the computa -
tion routine of the proprietary system was not made available to the
research team. Both features were established during the interview
with the unit’s manager and production nutritionist and are
reflected in the study’s limitations. In addition, PW is recorded as
an aggregate mass without compositional sorting; therefore, its
avoidable, potentially avoidable and unavoidable fractions could
not be separately quantified.
A supplementary descriptive Pareto analysis was performed
using the itemised CL reports for the nine service-date observa -
tions meeting the “fully critical day” . Preparations were grouped
according to the categories used in the source records: main
dishes, beans, side dishes and rice. Two distributions were
constructed: (i) the frequency of services with positive recorded
CL in each category, counting each category once per service, and
(ii) the recorded CL mass summed across the selected services for
each category. Categories were ranked in descending order sepa -
rately for each distribution, and cumulative percentages were cal -
culated using the total number of category—service occurrences
and the total recorded CL mass, respectively. A service could con -
tribute to several categories. The analysis describes the composi -
tion of recorded CL in the selected services; operational causes
were not quantified, and PW was not disaggregated by food
category.
3.4 Phase III: development of prevention
proposals
Prevention proposals were derived by triangulating the process
representations, the qualitative material and the quantitative analy -
sis of the indicators and were consolidated into a structured action
plan covering logistics, portion standardisation, serving-line con -
trol, managerial use of the indicators and consumer
communication.
Where the three sources converged, the representations record
the converging account; where they diverged, the divergence was
resolved in favour of the practise reported by the practitioners and
the operational definition in use was recorded. This occurred once.
The literature definition of CL treats it as prepared food that was not
distributed and is in principle suitable for reuse, whereas the inter-
view and the evaluation meeting established that in this unit food
transferred to the serving line cannot be returned to production
under any circumstances. What the unit records as CL is therefore
stated in Subsection 3.3, amongst the features that bound what the
indicators measure. Where one source was silent rather than contra-
dictory, as on the composition of plate waste, the silence is stated and
not filled.
In line with the evaluation activity of the design-science cycle
(Hevner et al., 2004; Peffers et al., 2007), every artefact produced in
the study—the SIPOC framework, the cross-functional process map
(Figure 1), the indicator protocol including the ‘fully critical day’
screening criterion, and the action plan—was submitted to the prac-
titioners of the unit at a feedback meeting held at the restaurant on
3 June 2026, that is, after the analysis of the full 2025 series and after
the construction of the process representations and of the
action plan.
The meeting was conducted by the research team with two
participants from the facility: the unit’s manager and the produc -
tion nutritionist, that is, the same two professionals who had
taken part in the diagnostic interview of March 2025. They were
selected because they are the professionals who would use the
artefacts in practise, holding responsibility for demand planning,
menu planning, the supervision of indicator recording and the
management of the service; the implication of their having also
been the informants of the diagnosis is stated amongst the
limitations.
The meeting followed the agenda reproduced in
Supplementary Material S4 stage-by-stage verification of the SIPOC
and of the swimlane map against the actual routine, with an explicit
request for corrections and omissions; presentation of the indicator
results by meal service, critical months, outliers and occasions of
simultaneous exceedance; presentation of each proposed action with

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 08 frontiersin.org
two explicit questions to the participants, on whether they would use
the artefact and what would have to change for them to do so; and
open discussion of the constraints on implementation. The evalua-
tion was therefore formative and use-oriented: it assessed the per -
ceived usefulness and operational fidelity of the artefacts, not their
effect on waste, since no action was implemented during the
study period.
The meeting was not audio-recorded. The feedback was registered
as field notes taken by the research team during the session and con-
solidated immediately afterwards into a written record organised arte-
fact by artefact, distinguishing confirmations, corrections and
additions. This record was analysed by the same thematic procedure
applied to the March 2025 material and yielded three categories of
feedback.
The first was confirmation of fidelity: the sequence of activities,
the allocation of responsibilities by sector and the interfaces repre -
sented in the SIPOC framework and in the swimlane map were rec -
ognised by both participants as corresponding to the actual
operating routine, and no correction was requested to either repre -
sentation. Both participants also agreed with the indicator results
and with the proposed actions, and no divergence was recorded
between them or in relation to the results presented. Because the
session produced no corrections, it provides confirmation of the
fidelity of the representations but little discriminating evidence
about them, the substantive yield of the meeting lay in two
additions.
The second category of feedback was an operational constraint
that had not been documented in the SOPs: food already transferred
to the serving line cannot be returned to production for reuse in this
unit, which establishes that replenishment beyond residual demand
near the end of service converts directly into clean leftovers, and
which is the reason why serving-line monitoring was retained as the
action with the highest expected leverage.
The third comprised complementary staff observations concern-
ing consumer behaviour. Because these observations were not
obtained through a systematic assessment of consumers, they were
retained only as contextual information and were not used as evi -
dence of specific consumer attitudes or behavioural mechanisms.
Because this evaluation relied on field notes rather than on a ver-
batim record, the qualitative claims derived from it are reported as
complementary evidence, and the absence of a recording is stated
amongst the limitations of the study.
4 Results
4.1 Mapping of the food production
process through SIPOC and swimlane
analysis
Table 1 summarises the SIPOC framework developed to define
the boundaries of the food production process, from demand plan -
ning and input receiving to waste weighing and final disposal, iden-
tifying the suppliers, inputs, activities, outputs and internal and
external customers involved. To enhance its analytical usefulness, the
table also indicates the potential relevance of each process stage to
PW and CL. This classification is intended to support the diagnostic
interpretation of the process and does not imply a demonstrated
FIGURE 1
Simplified cross-functional swimlane process map of the community restaurant.

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 09 frontiersin.org
causal relationship between a specific process stage and waste
generation.
A cross-functional process map (Figure 1) represents the activi-
ties performed by the storage, pre-preparation, cooking and distribu-
tion sectors and highlights the critical interfaces in the system,
particularly communication between the kitchen and the serving
line, the timing of food container replenishment and portioning
practises during meal service.
Taken together, the SIPOC and swimlane representations provide
the process-engineering framework for interpreting the waste indica-
tors. This process structure was subsequently integrated with the
quantitative PW and CL results to identify process hotspots associated
with observed waste deviations, without assuming demonstrated
causal relationships.
4.2 PW analysis: comparing lunch and
dinner, identifying critical months, outliers,
and menu impacts
Over the study period, the unit recorded 7.68 t of food waste
from PW and CL combined, comprising 6.875 t of PW and 0.802 t of
CL. Across the 374,604 meals served during the year, this corre -
sponded to 20.49 g of recorded PW and CL per meal and 2.93% of
the nominal meal mass, calculated based on the unit’s standard por-
tion weight of 0.70 kg. The apparent discrepancy between the 2.93%
of nominal meal mass recorded in the present study and the approxi-
mately 20% of food served reported as waste by Malefors (2021)
should be interpreted in light of important differences in measure -
ment boundaries and denominator conventions.
Malefors (2021) adopts a process-based framework in which
food waste can arise as receiving waste, storage waste, preparation
waste, safety-margin waste, serving waste, and plate waste; receiving,
storage, preparation, and safety-margin waste may also be aggregated
as kitchen waste. Preparation waste specifically includes losses gener-
ated during food preparation and trimming, such as peels, bones, and
fat. By contrast, the present study was restricted to the two waste
streams available in the facility’s routine operational records, PW and
CL, and therefore did not quantify waste generated during receiving,
storage, or food preparation.
The denominators also differ: Malefors (2021) defines waste as a
percentage of the actual mass of food served, whereas the 2.93%
reported here was calculated by relating the combined recorded mass
TABLE 1 SIPOC framework for the food production process of the community restaurant.
Suppliers (S) Inputs (I) Process (P) Outputs (O) Customers (C) Potential waste
relevance
Administration Demand forecast;
Historical data
Calculation of the
required number of
meals
Planned number of
meals; List of required
inputs; Inventory
requirements
Storage sector; Kitchen CL
Food suppliers Vegetables, meat, dairy
products; Invoices
Receiving and
inspection; Weighing
of incoming items
Received and verified
inputs
Storage sector; Pre-
preparation sector
No direct association
identified
Storage team Perishable and
nonperishable food
products
Food storage (cold
storage and dry storage
areas)
Properly stored
ingredients
Pre-preparation sector;
Kitchen
CL
Pre-preparation team Vegetables, meat
products
Food sanitation;
Cutting; Portioning
Pre-prepared food;
Food trimmings and
residues
Kitchen CL
Kitchen team Pre-prepared ingredients Cooking; Frying;
General food
preparation; Pass-
through operation
Ready-to-serve meals Meal distribution sector CL
Meal distribution team Ready-to-serve meals Plate assembly; Meal
service
Served meals Restaurant users/
customers
PW and CL
Restaurant users/
customers
Partially consumed
meals
Return of plates and
trays
PW; Used dishes Kitchen PW
Dishwashing team Plates containing food
residues
Waste separation;
Washing; Drying
Clean dishes; Separated
residues
Waste control sector;
Disposal process
No direct association
identified
Waste control team Separated PW Waste weighing and
data recording
Food waste indicators Administration;
Nutritionist
PW and CL —
monitoring stage
All operational stages Organic and inorganic
waste
Final waste disposal Properly separated and
stored waste
Waste collection company PW and CL — disposal
stage
PW = plate waste; CL = clean leftovers. “Potential waste relevance” indicates the process stage in which each waste indicator may be operationally relevant within the process framework. This
classification is intended for diagnostic purposes and does not imply a demonstrated causal relationship between the process stage and waste generation. “Monitoring stage” refers to the
measurement and recording of waste indicators rather than waste generation.

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 10 frontiersin.org
of PW and CL to the facility’s nominal standard portion weight of
0.70 kg per meal. Consequently, the two percentages do not represent
equivalent measures of food waste and should not be directly
benchmarked.
The lower percentage observed in the present study should there-
fore be interpreted as a site-specific measure of the PW and CL streams
captured by the facility’s monitoring system, rather than as evidence
that the restaurant generates less total food waste than the catering
establishments examined by Malefors (2021). Thus, the discrepancy is
primarily attributable to differences in measurement scope, waste-
category coverage and denominator conventions, rather than necessar-
ily reflecting superior waste performance at the studied facility.
Across daily service records with usable denominators, mean PW
(standard deviation) was 2.56% (0.63 percentage points; n = 249) at
lunch and 2.96% (1.09 percentage points; n = 248) at dinner. The corre-
sponding CL values were 0.30% (0.53 percentage points; n = 248) at
lunch and 0.46% (0.82 percentage points; n = 247) at dinner. These
descriptive summaries retain the measured zero CL values in December
and exclude the four undefined or implausible percentage records
described in Section 3.3. Tables 2–5 display selected critical service dates;
these descriptive statistics do not establish menu-related effects.
When the results were stratified by meal service, lunch accounted
for 5.901 t of the annual total and dinner for 1.776 t. However, dinner
showed a higher waste intensity, at 23.55 g per meal compared with
19.72 g at lunch. This difference was also reflected in the PW indica-
tors. Dinner displayed greater dispersion and a higher frequency of
days above the facility’s site-specific internal reference value of 3% in
the historical averages (Figure 2), whilst monthly maximum values
reached substantially higher levels, including peaks exceeding three
times this internal reference value (Figure 3).
At lunch, PW remained relatively stable throughout the year,
although occasional peaks were observed and warrant attention. The
main outliers identified for this service were analysed alongside
demand and menu information (Table 2). Tables 2–5 are intended as
descriptive screening tables of selected service-date observations
rather than as inferential comparisons. Accordingly, the reported
values are observed daily measurements and are not accompanied by
confidence intervals or significance tests. Statistical comparison of
meal services and menu-related patterns would require analysis of the
full longitudinal dataset and an inferential modelling framework
beyond the exploratory scope of these tables.
Forecast deviations on the selected high-PW lunch dates were
small, ranging from one to five meals out of approximately 1,200
meals served ( Table 2 ). These observations show that elevated
recorded PW can coexist with close agreement between forecast and
served meal counts. However, they do not establish the relative con-
tribution of forecasting error after accounting for other covariates.
Portioning and menu acceptability remain hypotheses for further
investigation, since neither was directly measured.
In contrast, dinner exhibited the highest PW levels of the study
period. Table 3 summarises the most critical occurrences, including
peaks exceeding three times the facility’s site-specific internal refer-
ence value.
On the most critical dinner days (22 December and 2 December
2025), there was a substantial gap between forecast and served meals
(Table 3). Because PW is expressed as a proportion of the food actually
served, however, these high values indicate that the consumers who
did attend rejected a considerable share of their meals, so overproduc-
tion alone does not explain the result.
Certain menu components recur on high-PW days across Tables 2,
3: meat dishes served in sauce, processed products and soft-textured
side dishes such as mashed potatoes and creamy polenta. In the absence
of sensory evaluation, these observations do not support causal infer-
ence, but their recurrence indicates that recipe standardisation, cook-
ing methods and portion sizes for these items deserve targeted
assessment. These recurrences are therefore stated as hypotheses H1 to
H3 in the discussion, to be tested by sensory, portioning and plating-
interaction studies, and not as findings of the present analysis.
Formal statistical testing of dish recurrence, such as chi-square
analysis, was not performed because the available operational records
were not structured at the menu-component level required for such
inference. Likewise, multilevel modelling of daily PW and CL would
require a broader set of systematically recorded operational covariates.
These approaches represent relevant extensions for future research.
4.3 CL analysis
CL followed a different pattern from PW . Whereas PW showed
persistent periods of elevated values, CL remained at or close to the
facility’s site-specific internal reference value of 2% for most of the
study period, with only sporadic exceedances on specific dates
(Figures 4, 5).
TABLE 2 Main PW values exceeding the target during lunch service in 2025.
Date Day of the
week
Near a
public
holiday
PW (%) Forecast
meals
Served
meals
Main dish Side dish
22 May 2025 Thursday No 4.94 1,195 1,200 Roasted chicken
nuggets
Garlic and olive
oil fusilli
18 June 2025 Wednesday Ye s 4.48 1,195 1,199 Beef and smoked
sausage stew
Braised Swiss
chard
12 June 2025 Thursday No 4.46 1,195 1,200 Chicken fricassée Packaged potato
sticks
20 February
2025
Thursday No 3.40 1,195 1,200 Chicken
stroganoff
Packaged potato
sticks
10 March 2025 Monday No 3.39 1,195 1,200 Chicken sausage
with onions
Garlic and olive
oil fusilli

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 11 frontiersin.org
TABLE 3 Main PW values exceeding the target during dinner service in 2025.
Date Day of the
week
Near a public
holiday
PW (%) Forecast
meals
Served
meals
Main dish Side dish
22 Dec 2025 Monday Ye s 9.46 300 141 Baked chicken
meatballs with
tomato sauce
Instant mashed
potatoes
2 Dec 2025 Tuesday No 7.59 300 158 Baked beef
meatballs with
tomato sauce
Instant mashed
potatoes
29 Jul 2025 Tuesday No 5.86 300 300 Tripe stew Creamy polenta
7 Feb 2025 Friday No 5.79 300 300 Chicken
stroganoff
Shoestring
potatoes
22 May 2025 Thursday No 5.62 300 300 Baked chicken
nuggets
Shoestring
potatoes
3 Jul 2025 Thursday No 4.81 300 300 Roasted chicken
drumstick
Cassava farofa with
carrots
TABLE 4 Main CL values exceeding the target during dinner service.
Date CL (kg) CL (%) Forecast meals Meals served Main dish Side dish
24 Jun 2025 10.80 5.14 300 300 Baked chicken
nuggets
Creamed corn
6 Jun 2025 8.00 3.81 300 300 Braised beef chuck
with calabrese sausage
Sautéed carrots
4 Jul 2025 7.40 3.52 300 300 Braised beef chuck
with calabrese sausage
Sautéed carrots
23 Jul 2025 6.15 2.93 300 300 Baked chicken
meatballs with tomato
sauce
Fusilli with tomato
sauce
19 Mar 2025 6.05 2.88 300 300 Sautéed chicken strips
with onions
Quibebe (Brazilian
mashed pumpkin)
12 May 2025 6.00 2.86 300 300 Baked chicken steak Creamed corn
21 Feb 2025 5.70 2.71 300 300 Braised beef chuck
with calabrese sausage
Cassava flour with
banana
7 Feb 2025 5.60 2.67 300 300 Chicken stroganoff Shoestring potatoes
25 Mar 2025 5.59 2.66 300 300 Spring-style ground
beef
Mashed potatoes
8 Aug 2025 4.70 2.25 300 298 Sautéed calabrese
sausage with onions
Creamy polenta
9 Jun 2025 4.67 2.22 300 300 Baked chicken
nuggets
Creamed corn
9 Jul 2025 4.60 2.19 300 300 Sautéed beef liver with
onions
Mashed potatoes
24 Jul 2025 4.40 2.10 300 300 Pork and chicken
cassoulet
Quibebe (Brazilian
mashed pumpkin)
22 Apr 2025 4.30 2.05 300 300 Spring-style ground
beef
Mashed potatoes
5 Mar 2025 3.50 1.29 300 274 Baked beef meatballs
with tomato sauce
Mashed potatoes

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 12 frontiersin.org
When stratified by meal service, lunch exceeded the facility’s
site-specific reference value of 2% only occasionally. The main outli -
ers occurred on 16 September (6.05% CL, 5.17 kg), when roasted
pork leg and canjiquinha were served, and on 22 May (2.67% CL,
22.40 kg), with baked chicken nuggets and fusilli. Dinner showed a
higher frequency of days above this internal reference value,
although occurrences remained sporadic ( Table 4). For the two
highest CL peaks, recorded on 24 June (5.14%) and 6 June (3.81%),
the forecasting error was zero: 300 dinners were forecast and exactly
300 were served.
High CL was recorded on dates when forecast and served meal
counts coincided. For those dates, the recorded meal counts show no
attendance shortfall relative to the forecast. However, they do not
establish why prepared food remained undistributed. Batch sizing,
demand for individual preparations and replenishment practises are
plausible mechanisms that require direct investigation; their contribu-
tions cannot be determined from agreement between total forecast
and served meal counts alone.
Several days exceeding the facility’s internal CL reference value also
exceeded its internal PW reference value, thereby meeting the
TABLE 5 Occurrences of fully critical days (PW > 3% and CL > 2%).
Date Meal service PW (%) CL (%) Forecast meals Meals
served
Main dish Side dish
22 May 2025 Lunch 4.94 2.67 1,195 1,200 Baked chicken
nuggets
Fusilli with garlic
and oil
4 Jul 2025 Dinner 3.17 3.52 300 300 Braised beef chuck
with calabrese
sausage
Sautéed carrots
12 May 2025 Dinner 3.48 2.86 300 300 Baked chicken
steak
Creamed corn
7 Feb 2025 Dinner 5.79 2.67 300 300 Chicken stroganoff Shoestring potatoes
8 Aug 2025 Dinner 3.93 2.25 300 298 Sautéed calabrese
sausage with
onions
Creamy polenta
9 Jun 2025 Dinner 3.14 2.22 300 300 Baked chicken
nuggets
Creamed corn
9 Jul 2025 Dinner 3.81 2.19 300 300 Sautéed beef liver
with onions
Flaked mashed
potatoes
24 Jul 2025 Dinner 3.52 2.10 300 300 Pork and chicken
cassoulet
Quibebe (Brazilian
mashed pumpkin)
22 Apr 2025 Dinner 4.10 2.05 300 300 Spring-style
ground beef
Flaked mashed
potatoes
FIGURE 2
Historical average PW by food service in 2025.

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 13 frontiersin.org
site-specific ‘fully critical day’ criterion adopted in this study (4 Jul, 12
May, 7 Feb, 8 Aug, 9 Jun, 9 Jul, 24 Jul and 22 Apr 2025). Conversely,
dates with CL above its internal reference value and PW within its ref-
erence value indicate that only CL exceeded the adopted threshold.
This pattern does not, by itself, identify the mechanisms responsible for
the deviation or rule out factors related to menu acceptability or
portioning.
4.4 Integrated analysis of PW, CL, menu,
and forecast versus actual meals served
PW primarily reflects losses at the consumption interface,
whereas recorded CL reflects losses emerging across the
production—distribution interface and may be associated with batch
sizing, the quantity transferred to the serving line, replenishment
practises and residual demand at the end of service. High values
therefore need not occur simultaneously and, importantly, neither
indicator alone identifies the specific operational mechanism respon-
sible for the observed waste. Integrating both datasets provided an
exploratory means of identifying days on which elevated losses
occurred concurrently across these operational interfaces. For this
site-specific analysis, a “fully critical day” was operationally defined
as a day on which PW exceeded the facility’s internal reference value
of 3% and CL exceeded its internal reference value of 2% within the
same meal service (Table 5). This classification was used as an explor-
atory screening criterion rather than as a validated diagnostic metric.
FIGURE 3
Historical maximum PW by food service in 2025.
FIGURE 4
Historical average CL by food service in 2025.

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 14 frontiersin.org
Accordingly, the identified days should be interpreted as cases
selected for closer operational examination, rather than as objectively
more critical than days identified through alternative criteria.
The criterion’s value lies not in establishing causality, but in direct-
ing managerial attention to service episodes that warrant joint exami-
nation of menu characteristics, portioning, batch sizing, replenishment
and kitchen—distribution interfaces.
Across the nine service-date observations classified as fully critical
(Table 5), the absolute difference between forecast and served meal
counts ranged from zero to five meals. These selected cases show that
close agreement between forecast and recorded attendance can coexist
with PW and CL exceeding their respective internal reference values.
However, this descriptive comparison does not establish the association
between forecast deviation and waste across the full year. The associ-
ated menus provide context for further investigation, but the available
records do not establish whether menu acceptability, portioning, batch
sizing or replenishment practises contributed to these deviations.
A multivariate analysis controlling for factors such as day of the
week, month, proximity to public holidays, seasonal menu variation,
staffing patterns and serving-line supervision could provide a stronger
test of the relative contribution of these factors. However, such explan-
atory modelling falls outside the diagnostic and process-mapping
scope of the present study and represents an important direction for
future research.
The monthly summaries in Figures 2–5 describe variation within
2025. No formal test of month effects, seasonal decomposition or
time-series model was performed. Consequently, these summaries do
not establish statistically significant monthly differences or a recurring
seasonal pattern. The available records cover one annual cycle; addi-
tional years would be needed to assess whether an apparent annual
pattern repeats. Exploratory comparisons within 2025 remain possi-
ble, but would need to account for meal service, calendar effects and
dependence between daily observations, whilst recognising the lim-
ited availability of systematically recorded operational covariates.
The supplementary Pareto analysis of the nine fully critical meal
services identified 22 category—service occurrences of positive
recorded CL (Figure 6A). Main dishes accounted for eight occur -
rences, beans for six, side dishes for five and rice for three. The same
categories contributed 30.57, 16.40, 10.50 and 6.60 kg, respectively,
totalling 64.07 kg ( Figure 6B ). Main dishes and beans together
accounted for 63.64% of occurrences and 73.31% of recorded CL
mass. These distributions identify food categories for closer investiga-
tion within the selected services; they do not establish operational
causes or represent the annual distribution of waste.
To further integrate the quantitative findings with the process-
engineering analysis, the main waste deviations were cross-referenced
with the process stages and operational interfaces identified through
the SIPOC and swimlane mapping. This synthesis highlights the inter-
faces at which PW and CL become operationally relevant and provides
specific quantitative examples from the 2025 records. The resulting
hotspots are presented in Table 6.
The purpose of this integration is not to attribute waste to specific
activities as demonstrated causes, but to narrow the diagnostic focus
and identify process conditions that warrant targeted investigation. In
this sense, the process map provides the operational context in which
the quantitative deviations can be interpreted, whilst the waste indica-
tors provide empirical evidence for prioritising the interfaces requir-
ing attention. Each association in Table 6 is therefore a proposed
mechanism that remains a candidate explanation until it is examined
in a study designed for that purpose, typically an intervention study
that modifies the suspected mechanism and monitors both indicators
before and after.
5 Discussion
5.1 Interpretation of the observed waste
patterns
Food waste in the studied unit is distributed across several stages
of the production process and cannot be attributed to a single cause.
FIGURE 5
Historical maximum CL by food service in 2025.

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 15 frontiersin.org
The observed exceedances are interpreted here as deviations from the
facility’s own monitoring reference rather than as evidence of objec-
tively or universally high PW or of statistically defined process insta-
bility. The recurrence of meat dishes in sauce, processed products and
soft-textured side dishes on high-PW days suggests a relationship
between acceptability, portioning and PW that, in the absence of sen-
sory evaluation or structured consumer surveys, remains an analytical
hypothesis rather than a demonstrated causal link.
The higher PW observed at dinner than at lunch should not be
interpreted as an effect of meal service itself. The two services differ
substantially in scale, with approximately 1,200 lunches and 300 din-
ners planned per day, and other service-specific factors that were not
systematically recorded, including the demographic composition of
users attending each service and staffing levels and workload, may also
contribute to the observed difference. Although menu information
was available for individual service dates, the dataset did not support
a systematic comparison of menu-cycle characteristics between lunch
and dinner. Consequently, the present study cannot determine
whether the higher dinner PW reflects differences in menu composi-
tion, consumer profile, staffing conditions or other time-of-service
characteristics. These factors should be treated as potential covariates
in future comparisons between meal services.
CL was sporadic, with peaks concentrated on specific days.
Notably, high CL occurred even when forecast and served meal counts
coincided exactly, indicating that attendance shortfalls alone cannot
explain these observations. Given the facility’s recording boundary,
recorded CL represents food remaining in serving-line containers at
the end of service and may arise from interconnected mechanisms at
the production—distribution interface, including batch-sizing deci-
sions, the quantity transferred to the serving line, kitchen—distribu-
tion communication and late or excessive replenishment. The available
data do not permit the relative contribution of these mechanisms to
be determined.
The days identified through the exploratory ‘fully critical day’ cri-
terion provide additional support for this interpretation. On those
dates, food was left uneaten by consumers whilst recorded CL simul-
taneously remained at the end of service. This pattern suggests that
factors related to both the consumption interface and the produc -
tion—distribution interface may have contributed to the observed
waste; however, the available data do not allow specific mechanisms,
such as menu acceptability, portioning, batch sizing or serving-line
replenishment, to be identified as causal factors.
These results are not specific to the case. Distribution losses and
PW have been identified as the dominant waste sources in the Swiss
food service industry (Betz et al., 2015), and Swedish public catering
has shown a comparable dependence on portioning and batch-sizing
decisions (Malefors et al., 2022).
The unit’s management attributed the low recorded values to the
service model itself, since every component of the main meal is plated
by the serving staff rather than self-served, which they considered to
afford closer control over the quantity leaving the line. This is a prac-
titioner explanation rather than a tested one, but it is a plausible rival
explanation for part of the difference between the values recorded in
this unit and those reported for self-service settings and one that a
comparative study across service models could examine.
From a process-engineering standpoint, the mapping identifies
distribution as the control point with the greatest leverage over
resource efficiency. This stage concentrates decisions on portion size,
container replenishment and communication with the kitchen, so
minor failures at this interface propagate into both indicators, much
as a poorly controlled interface between two operational stages propa-
gates yield loss downstream in any production chain. It also follows
that the waste records already collected by the unit have substantial
untapped diagnostic value: analysed alongside menu, meal service,
critical dates and observed operating conditions through the SIPOC/
swimlane lens, they become a resource-efficiency monitoring instru-
ment rather than merely a compliance record.
A particularly relevant comparison is provided by previous evi-
dence from a Brazilian community restaurant in the state of Rio de
Janeiro. Messias et al. (2025) reported mean monthly PW rates of
2.47%, 2.57%, and 1.71% from November , 2023 to January 2024,
respectively, whilst mean CL rates were 1.49%, 1.50%, and 0.77% over
the same period. The study also documented day-to-day variability,
including seven days in December on which PW exceeded the facili-
ty’s internal reference value of 3%. These findings provide a contextu-
ally closer benchmark for interpreting the present results because they
derive from the same type of Brazilian public food-service setting and
distinguish between waste generated at the consumption interface and
CL from production. Brazilian evidence from other institutional food-
service settings provides useful context for interpreting the present
FIGURE 6
Pareto analysis of recorded clean leftovers across nine fully critical meal services in 2025. (A) Frequency of food categories with positive recorded clean
leftovers, comprising 22 category service occurrences. A service may contribute to more than one category. (B) Recorded clean-leftover mass by food
category, totalling 64.07 kg. Blue bars represent occurrences in panel A and mass in panel B; orange lines indicate cumulative percentages. These
distributions describe recorded leftovers and do not establish their operational causes.

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 16 frontiersin.org
TABLE 6 Integration of the waste indicators with the process map: observed deviations, process stages and interfaces involved, and proposed mechanisms to be tested.
Observed
deviation
(empirical
evidence)
Process stage/
interface (Table 1;
Figure 1)
Proposed
mechanism (not
tested)
Corresponding
hypothesis
Evidence required to
test the mechanism
PW systematically higher at
dinner than at lunch
throughout 2025 (monthly
means 2.08–3.58% vs. 1.90–
3.15%; Figure 2), with
monthly maxima up to
9.46% (Figure 3)
5. Meal distribution (5.1
meal portioning; 5.2 plate
assembly; 5.3 meal service)
and 6. Consumption (6.1–
6.3)
The portion standard applied
at the serving line does not
distinguish the dinner service,
whose population is fourfold
smaller (300 forecast meals vs.
approx. 1,200 at lunch) and
may differ in composition and
appetite, so served portions
exceed demand
H2 Measurement of served and
returned mass per menu
component, by meal service
Elevated PW at lunch on
dates whose forecast
deviation was 1–5 meals out
of approx. 1,200 served
(Table 2)
6. Consumption (6.3
uneaten food), downstream
of 1. Planning (1.1–1.3)
The loss is generated after
demand planning;
acceptability of the
preparation, portion size or
the plating interaction, rather
than forecasting error, account
for the deviation
H1, H2, H3 Sensory evaluation and
structured acceptability
assessment with users,
combined with component-
level portion measurement
Highest dinner PW on 22
Dec (9.46%) and 2 Dec
(7.59%), when attendance
was far below forecast (141
and 158 of 300 meals;
Table 3)

  1. Planning (1.1 demand
    forecast) and 5. Meal
    distribution (5.4 container
    replenishment)
    Because PW is expressed as a
    proportion of food actually
    served, low attendance does
    not by itself produce a high
    value; the consumers who did
    attend rejected a substantial
    share of the meal served to
    them
    H1 Plate-level analysis of returned
    food by component on low-
    attendance dates
    Recurrence of specific menu
    components on high-PW
    and high-CL dates: soft-
    textured starch-based side
    dishes (instant mashed
    potato, creamy polenta,
    creamed corn) and
    reformed or processed
    protein items (baked
    nuggets, meatballs)
    (Tables 2–5). In the
    interview, the unit’s
    management independently
    identified low acceptance of
    particular dishes as the
    main circumstance in which
    waste arises, citing a
    preparation withdrawn
    from the menu after poor
    acceptance
  2. Pre-preparation (3.4–3.5)
    and 4. Cooking (4.1–4.2);
    effect observed at 6.
    Consumption (6.3)
    Low sensory acceptability
    associated with the texture,
    cooking method or
    presentation of these specific
    preparations
    H1 Sensory evaluation of the
    recurring items and structured
    consumer acceptability survey
    (Continued)

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 17 frontiersin.org
TABLE 6 (Continued)
Observed
deviation
(empirical
evidence)
Process stage/
interface (Table 1;
Figure 1)
Proposed
mechanism (not
tested)
Corresponding
hypothesis
Evidence required to
test the mechanism
Management reported that,
given the pace of the serving
line, components are placed
on the plate before the user
can decline them, and that
staff do not always check
whether the user wants a
given component (interview
with the manager)
5. Meal distribution (5.1
meal portioning; 5.2 plate
assembly)
Absence of a preference check
at the moment of plating
transfers unwanted
components onto the plate,
generating PW independently
of both the acceptability of the
preparation and the portion
standard
H3 Structured observation of the
plating interaction, combined
with component-level
measurement of returned food
Staff reported at the artefact
evaluation meeting that
some users return to the
serving line for an
additional portion in
periods of shorter queues,
with consumption
concentrating on specific
components, notably the
meat item
5. Meal distribution (5.3
meal service) and 6.
Consumption (6.1–6.3)
The remaining components of
the additional portion are
returned uneaten, so repeat
servings raise PW without a
corresponding increase in
food actually consumed
H6 Records of repeat servings
during service, combined with
component-level measurement
of returned food
CL peaks on dates with zero
forecast error (24 Jun,
5.14%; 6 Jun, 3.81%;
Table 4)
4. Cooking (4.1–4.4) and 5.1
holding of prepared meals;
5.4 container replenishment
Batch sizes and end-of-service
replenishment exceed the
residual demand of the
service; food already
transferred to the serving-line
container is discarded rather
than reused in this unit, so the
excess converts directly into
recorded CL
H4 Records of batch size per
preparation, replenishment
times and residual quantity
per container at the end of
service
CL at or close to the internal
reference value for most of
the year, with sporadic
peaks on specific dates
(Figures 4 and 5)

  1. Planning (1.2–1.3) and 4.
    Cooking (batch sizing)
    The CL loss is event-driven,
    associated with particular
    preparations and services,
    rather than a systemic feature
    of production planning
    H4 Time-series analysis of CL
    against operational covariates
    (preparation, batch size,
    attendance, day of week)
    Nine service-date
    observations on which PW
    and CL exceeded their
    internal reference values
    within the same service
    (Table 5)
    Interface between 4.
    Cooking/Kitchen and 5.
    Meal distribution
    Absence of a formalised
    routine for reporting low
    uptake of a preparation, so
    replenishment continues
    whilst the item is being
    rejected at the consumption
    interface
    H5 Before-and-after intervention
    study introducing the
    communication routine, with
    both indicators monitored
    (Continued)

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 18 frontiersin.org
findings. In university food services, Zotesso et al. (2016) identified
substantial organic waste generation across meal-production pro -
cesses, whereas Borges et al. (2019) reported significant reductions in
food CL and per capita PW following staff training and a consumer-
awareness campaign.
Hospital food services present a different operational and con-
sumer context. Lucena et al. (2024) reported both CL and PW in a
hospital FNU in northeastern Brazil, whilst Godoy et al. (2024) found
marked differences in PW amongst hospital staff, patients and com-
panions. More recently, Targino et al. (2025) quantified PW in a hos-
pital FNU in Rio de Janeiro. Taken together, these Brazilian studies
illustrate that food-waste patterns vary across institutional settings
according to service characteristics, consumer populations and opera-
tional arrangements.
What the present case adds is evidence that forecast accuracy, fre-
quently treated as the main lever for prevention, is not sufficient: on
the days identified through the exploratory ‘fully critical day’ criterion,
the forecast was essentially exact, yet both PW and CL exceeded their
respective site-specific internal reference values.
In a publicly subsidised facility, the recorded losses also carry
social and economic weight because discarded food represents an
inefficient use of public resources, including raw materials, labour,
water, energy and public funds allocated to food and nutrition secu-
rity. In this context, CL represents an avoidable loss, whereas the
resource-efficiency implications of PW should be interpreted with
caution because its avoidable, potentially avoidable and unavoidable
fractions were not separately quantified.
These findings also position process-based diagnosis within the
broader range of contemporary food-waste mitigation strategies.
Recent literature has explored approaches ranging from digital trace-
ability and real-time monitoring to intelligent packaging and down-
stream waste valorisation (Heydari, 2024; Heydari et al., 2025; Li et al.,
2026; Rahman et al., 2026 ). These approaches operate at different
stages of the food-waste hierarchy and supply chain.
The protocol proposed in the present study addresses an earlier
diagnostic and preventive stage: routinely generated waste data are
linked to specific operational stages and interfaces to identify where
avoidable losses emerge before selecting corrective actions. In this
sense, SIPOC and swimlane mapping can support the identification
of operational points at which targeted interventions or enhanced
monitoring may be considered. Consumer behaviour represents a
plausible but unmeasured dimension of PW generation in the studied
facility. Because consumer attitudes, preferences and motivations were
not systematically assessed, the present data do not support conclu-
sions regarding specific behavioural mechanisms. Previous research
nevertheless indicates that food-waste behaviour may be influenced
by awareness, cultural norms and socioeconomic context, and that
entrenched behaviours and limited awareness can represent barriers
to waste reduction (Agya, 2025). Behavioural interventions should
therefore be adapted to the socioeconomic and cultural characteristics
of the population concerned (Agya, 2025). In the context of a publicly
subsidised community restaurant serving socially vulnerable popula-
tions, these factors should be investigated directly through structured
consumer surveys, interviews or systematic observational protocols
before specific behavioural mechanisms are incorporated into targeted
interventions.
5.2 Conditions of applicability of the
protocol in other catering contexts
The protocol reported above was developed and evaluated in one
facility. Any extension to other settings is therefore an analytic and not
a statistical claim (Yin, 2018), and it is stated here as a set of scope
conditions derived from those features of the case on which the pro-
tocol actually depends, so that readers can judge the correspondence
with their own setting (Simons, 2015).
Six operational prerequisites are necessary for the protocol to be
applied as described. The unit must (i) operate as public catering, or
as an equivalent institutional food service, in which the cost of waste
falls on a public or institutional budget and the efficient use of those
funds is an explicit management objective; (ii) work with standardised
menus planned in advance, so that recorded losses can be related to
identifiable preparations; (iii) operate under documented standard
operating procedures, so that a process map represents a stable flow
rather than an improvised one; (iv) produce meals in batches rather
than to individual order, so that CL exist as a distinct loss category and
batch sizing is a controllable decision; (v) weigh its waste routinely, on
TABLE 6 (Continued)
Observed
deviation
(empirical
evidence)
Process stage/
interface (Table 1;
Figure 1)
Proposed
mechanism (not
tested)
Corresponding
hypothesis
Evidence required to
test the mechanism
Both indicators weighed
and recorded daily, but used
mainly to demonstrate
compliance with the
internal reference values of
the management contract;
no waste-reduction
campaign had been
undertaken because
recorded levels remained
within target (interview)
7. Waste control (7.2–7.4
weighing and
recording) → 1.
Administration
The recorded data are not fed
back into planning, batch-
sizing and replenishment
decisions, so deviations recur
without a documented
managerial response
Managerial diagnosis (no causal
claim)
Documented monthly review
of flagged occasions with the
decision taken recorded,
enabling before-and-after
evaluation

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 19 frontiersin.org
calibrated scales and with a recording system that retains a historical
series long enough to distinguish recurring patterns from isolated
events; and (vi) record PW and CL separately, which is the condition
that makes it possible to locate a loss upstream, at production and
batch sizing, or downstream, at the consumption interface.
Amongst the units that satisfy these prerequisites, three institu-
tional contexts are the closest candidates for application, each requir-
ing a specific adaptation. The closest is the school food service,
because it shares the feature that governs the consumption interface
in the present case: meals are plated by staff rather than self-served, so
portion size and the decision to place a component on the plate are
operational variables under the unit’s control. The adaptation required
their concerns stratification, since acceptance varies systematically
with age group and with position in the menu cycle, and the indicators
should be stratified accordingly rather than by meal service.
In hospital food services the protocol applies to the production
chain, but the consumption interface is governed by prescribed thera-
peutic diets and by the clinical condition of patients; PW is then only
partly an operational variable and must be interpreted together with
nutritional intake, so a critical occurrence cannot be defined by
exceedance of an internal reference value alone.
Corporate and university catering requires the greatest adaptation,
despite being superficially similar in scale, because a self-service line
moves the portioning decision from the server to the consumer: the
mechanisms proposed here for the plating interaction do not apply,
the control point shifts to container sizing and replenishment, and
demand is more volatile than in a subsidised community restaurant
with a stable attendance profile, so forecast accuracy must itself be
treated as a monitored variable rather than as a stable baseline.
What does not transfer, in any of these contexts, are the quantita-
tive results of this case. The internal reference values of 3% for PW and
2% for CL are managerial conventions of the studied unit and not
validated diagnostic cut-offs; the observed indicator levels depend on
menu composition, consumer profile, portion standard and measure-
ment boundary; and the ‘fully critical day’ criterion is an exploratory
screening device whose diagnostic performance has not been com-
pared with alternatives such as combined waste mass or percentile-
based thresholds.
What is offered for transfer is the diagnostic sequence—stratify
both indicators by service, cross-check them against forecast accuracy,
screen for simultaneous exceedance of the unit’s own reference values,
and attribute the screened occasions to stages and interfaces of a pro-
cess map—together with the process-engineering representations that
make such attribution possible. Whether the pattern observed here
recurs in units that satisfy the six prerequisites is an empirical question
requiring replication, for which additional cases should be selected
because they predict either similar results or contrasting results for
predictable reasons (Conejero et al., 2024).
Figure 7 represents the protocol as a sequence of diagnostic steps
with the scope conditions attached to each, distinguishing what is pro-
posed for transfer—the diagnostic sequence and the process-engineer-
ing representations—from what is specific to this case.
5.3 Hypotheses for future testing
Because the mechanisms discussed in this study were not directly
measured, they are explicitly formulated as hypotheses for future test-
ing rather than presented as established findings.
H1: low sensory acceptability of specific preparations, in particu-
lar soft-textured starch-based side dishes such as instant mashed
potato, creamy polenta and creamed corn, and reformed or pro-
cessed protein items such as baked nuggets and meatballs,
increases PW at the consumption interface.
H2: served portion sizes exceeding consumer demand for spe -
cific menu components increase PW independently of the
acceptability of the preparation.
H3: the absence of a preference check at the moment of plating
increases PW independently of both acceptability and portion
size, because components are placed on the plate before the user
can decline them.
H4: batch sizing and container replenishment in excess of the
residual demand of the service increase CL independently of
forecast accuracy.
H5: the absence of a formalised kitchen—serving-line commu-
nication routine on low uptake of specific preparations increases
the probability of an occasion on which both indicators exceed
their internal reference values.
H6: repeat servings taken when queues are short increase PW ,
because consumption concentrates on specific components,
notably the meat item, whilst the remaining components of the
additional portion are returned uneaten.
H1 and H3 are supported by converging qualitative evidence,
since the unit’s management independently identified both low accep-
tance of particular dishes and the pace of the serving line as the cir-
cumstances in which waste arises, and cited a preparation withdrawn
from the menu after poor acceptance; H6 rests on reports collected at
the artefact evaluation meeting. In both cases this is triangulating
interview evidence, not a measurement, and it establishes none of the
mechanisms.
Testing H1 requires sensory evaluation and a structured accept-
ability assessment with users; H2 requires measurement of served and
returned mass by menu component; H3 requires structured observa-
tion of the plating interaction together with component-level mea-
surement of returned food; H4 requires records of batch sizes,
replenishment times and residual quantity per container; H5 requires
an intervention study introducing the communication routine and
monitoring both indicators before and after; and H6 requires records
of repeat servings during service combined with component-level
measurement of returned food.
None of these instruments was applied in the present study, and
the recorded data cannot substitute for any of them. The unit does
administer an internal satisfaction form with a three-level response
scale, but it is not a validated acceptability instrument, is not linked to
specific preparations and was not used in this analysis; it could, how-
ever, provide a starting point for the assessment that H1 requires.
5.4 Proposed action plan and
implementation roadmap
Building on the process mapping, the indicator analysis and the
fully critical days, resource-efficiency and waste-prevention actions

Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 20 frontiersin.org
were organised into five dimensions: (a) monitoring of distribution
and container replenishment; (b) monitoring of menu acceptability;
(c) portion standardisation; (d) managerial use of waste indicators;
and (e) consumer awareness. The proposals are managerial in nature
and were not implemented during the study period.
The proposals are managerial in nature and were not implemented
during the study period. Amongst these, serving-line monitoring
during distribution, particularly at dinner, carries the greatest
expected leverage, because food already transferred to the serving line
cannot be returned to production for reuse in this unit, so that replen-
ishment beyond residual demand near the end of service converts
directly into clean leftovers. Each action is stated in operational detail
in Table 7.
A consumer awareness campaign may complement the opera -
tional actions because PW originates at the consumption interface
and may be influenced by behavioural and contextual factors.
However, consumer attitudes and motivations were not
systematically assessed in the present study. The proposed com -
munication component should therefore be regarded as a preven -
tive action informed by the broader food-waste literature rather
than as a response to behavioural drivers demonstrated in the stud -
ied facility.
Based on the improvement opportunities identified, a structured
action plan was developed and is presented in Table 7. The plan orga-
nises the main proposed actions by linking the observed problems to
their possible causes, the suggested responsible parties and the
expected outcomes.
The action plan is the prescriptive artefact of the study in the
design-science sense: it translates the diagnosis into a set of design
propositions, each linking an observed problem, a proposed mecha-
nism and an intervention expected to act on that mechanism, which
is the form van Aken (2004) describes as a technological rule—
grounded in evidence from the field, but only initially field-tested until
its effects are measured.
FIGURE 7
Diagnostic protocol.

Vieria Mariano et al. 10.3389/fsufs.2026.1981661
Frontiers in Sustainable Food Systems 21 frontiersin.org
TABLE 7 Proposed action plan for food waste reduction in the community restaurant.
Axis Identified
problem
Possible
cause
Proposed
action
Suggested
responsible
parties
Suggested
timeframe
Expected
outcome
Monitoring of food
distribution and
container
replenishment
Increase in CL on
days with accurate
demand forecasts
Production or
replenishment
exceeding actual
demand during
meal service
Implement visual
checks before
replenishing food
containers, especially
during the second
half of the service
period
Distribution and
kitchen teams
Short term Reduction of
undistributed
leftovers
Monitoring of food
distribution and
container
replenishment
Communication
failures between the
kitchen and the
serving line
Lack of a formal
routine for
reporting low
uptake of specific
preparations
Establish rapid

论文原文与出版信息

Frontiers in Sustainable Food Systems
查看论文原文

继续探索