An EU-aligned protocol for triangulating food safety culture evidence in manufacturing plants: maturity gaps and confidence ratings

1 Introduction

Food safety management systems (FSMS) are widely implemented in food manufacturing and processing, typically through prerequisite programmes (PRPs), HACCP-based procedures, internal audits, and verification activities. In parallel, widely used FSMS standards and certification schemes (e.g., ISO 22000 and GFSI-recognised schemes) increasingly emphasise people-related enablers such as leadership, competence, and communication, reinforcing the need to assess how systems work in practice, not only how they are documented. FSMS are designed to reduce and control food safety risks rather than eliminate them entirely. Nevertheless, incidents, recalls and non-conformances still occur, showing that documented controls need to be considered alongside how they are implemented, maintained and adapted in daily practice. An implementation gap may arise between documented systems and day-to-day practice: controls may be documented without being consistently applied. This may become evident in routine decisions, particularly under operational pressure (e.g., staff shortages or unexpected disruptions), and in how people recognise, report and manage deviations. Food safety culture (FSC) helps explain these gaps because it reflects the shared values and priorities that shape everyday food safety decisions and practices.

In the European Union, FSC is now an explicit regulatory expectation, not only a ‘best practice’ topic. Commission Regulation (EU) 2021/382 inserted Chapter XIa (Food safety culture) into Annex II of Regulation (EC) No 852/2004. It requires food business operators to establish, maintain and provide evidence of an appropriate FSC through management and employee commitment to safe food production; leadership that engages employees in food safety practices; awareness of food safety hazards and hygiene; open and clear communication; and adequate resources. Commission Notice 2022/C 355/01 (non-binding guidance) provides practical implementation support and in Appendix 3, includes an illustrative example of an FSC questionnaire using five-point Likert-scale items aligned to the same core elements. This example is adapted from the Food Safety Climate self-assessment tool developed by De Boeck et al., (European Commission [EC], 2022; De Boeck et al., 2015). As such, it offers a straightforward, implementable starting point, and recent EU-inspired applications have largely operationalised FSC through questionnaire-based perception measures (e.g., Marzocca et al., 2025). However, questionnaire-based assessments primarily provide evidence of perceptions, and complementary sources are needed when the assessment also seeks to examine how food safety practices are implemented in day-to-day work. Recent conceptual work emphasising the behavioural and dynamic nature of FSC cautions against reducing FSC to a checklist-like compliance component (Da Cunha et al., 2025).

Broader organisational culture theory describes culture as a set of shared assumptions learned and reproduced within a group over time (Schein, 2004). Early work on food safety culture helped shift attention beyond documented FSMS procedures towards the human and organisational conditions that shape food safety in practice (Yiannas, 2009). In the food safety literature, FSC has been described as a relatively stable, learned and shared set of attitudes, values and beliefs that contributes to food safety-related practices within a food-handling environment (Griffith et al., 2010a; Griffith et al., 2010b). It has also been described in terms of the shared values, beliefs and norms that shape food safety mindset and behaviour across an organisation (GFSI, 2018). De Boeck et al. (2015) define food safety climate in terms of employees' shared perception of leadership, communication, commitment, resources and risk awareness concerning food safety and hygiene.

Sharman et al. (2020) further distinguish between food safety culture and food safety climate. They describe FSC as a long-term, organisation-level construct shaped by deeply rooted beliefs, behaviours and shared assumptions learned over time. Food safety climate, in contrast, is commonly treated as a more immediate and time-bound manifestation of culture, typically captured through employees’ shared perceptions and attitudes at a given point in time. Because FSC is a latent construct, it cannot be observed directly, so it is inferred through indicators such as perceptions, attitudes and enacted behaviours in daily work. For this reason, many FSC assessment tools operationalise culture through climate measures. Questionnaire-based tools play an important role in assessing FSC. They can help measure food safety climate, compare perceptions across sites or groups, and identify patterns that may not be apparent from records alone. The issue is not that questionnaires are inherently inadequate, but rather the scope of the conclusions drawn from their findings. When the aim is to understand how FSC is reflected in day-to-day work, questionnaire findings should therefore be considered alongside observations, records, verification activities, and other system- or performance-based evidence. Moreover, a high climate score does not necessarily mean that safe behaviours are consistently enacted. Conversely, a low score may reflect a more critical workforce rather than weak controls. Importantly, score interpretation and benchmarking depend on instrument quality. A recent review found that the reliability and validity of FSC/climate assessment instruments vary and should be considered when interpreting scores and benchmarking results (Nyarugwe and Jespersen, 2024). Pai et al. (2024) suggest that questionnaires may primarily reflect “what people report” rather than directly capturing “what people do”.

Additionally, several biases can affect survey responses. Social desirability bias is especially relevant in hierarchical settings or where staff fear blame. In such contexts, respondents may provide answers they believe are “correct”, even if these do not reflect their actual practice, rather than candid responses. To address this, Jespersen, MacLaurin and Vlerick (2017) developed the Food Safety Desirability Response Scale (FSDRS) to quantify socially desirable responding in FSC surveys. In addition, common method bias (CMB) is another concern. In this instance, when both FSC measures and their supposed outcomes are collected using the same self-report method, observed correlations may be inflated by shared method variance rather than reflecting true relationships. This threat to inference is well described in the organisational and behavioural methods literature (Podsakoff et al., 2003).

For these reasons, FSC assessment benefits from triangulation. Triangulation is commonly defined as the purposeful combination of multiple data sources and/or methods to strengthen inference and reduce single-method bias in organisational assessment (Denzin, 1978; Creswell and Plano Clark, 2011). The U.S. FDA describes triangulation in practice as a package of tools that may include a staff survey, interviews or focus groups, on-site observation, and document/record review (U.S. Food and Drug Administration [FDA], 2022). In this paper, triangulation is understood as the deliberate comparison of evidence from different sources and methods, rather than as the simple accumulation of more data. This interpretation follows Denzin’s distinction between different forms of triangulation, including data and methodological triangulation, and is consistent with Lincoln and Guba’s emphasis on credibility and trustworthiness in qualitative inquiry (Denzin, 1978; Lincoln and Guba, 1985). Triangulation is therefore used to make the interpretation more credible and transparent by showing where evidence streams converge, where they diverge and how these divergences are addressed. Without eliminating bias or replacing professional judgment, it makes the basis for that judgment visible. In FSC assessment, convergence across evidence streams can support more defensible gap statements, while divergence should be treated as an informative finding rather than as a failure of the method. Such divergence may reveal misalignment between management expectations, frontline reality and documented system performance (Jespersen and Wallace, 2017). Additionally, triangulation supports linking upstream signals (leading indicators) to downstream outcomes (lagging outcomes). In this study, leading indicators include leadership behaviours, communication routines, training quality, resource adequacy, and risk awareness, whereas lagging outcomes include non-conformances, repeat deviations, complaints and verification failures.

According to Jespersen and Wallace (2017), a practical gap remains in the FSC literature, as many studies that assess food safety culture using mixed methods neither define what they mean by “triangulation” nor explain how evidence from surveys, observations, and records is integrated into a final judgement (Jespersen and Wallace, 2017). Without explicit decision rules for handling disagreement between streams, internal assessors, auditors and competent authorities may reach different conclusions, which creates uncertainty for decision-making, weakens audit defensibility, and delays improvement planning. The risk is highest when the streams contradict each other, for example, when surveys suggest a positive culture but observations show inconsistent application of critical food safety practices. The issue is compounded in workplaces without a just/no-blame reporting climate, where deviations may be under-reported and survey responses overly positive, inflating maturity impressions and delaying corrective action (Wiśniewska, 2023; FoodDrinkEurope, 2023).

The gap is not the absence of FSC assessment tools. The gap is the lack of explicit integration rules for combining evidence streams and reporting confidence in FSC conclusions. This paper addresses this gap by combining a focused mapping review of six empirical studies in manufacturing and processing plants with a practical protocol for FSC assessment. Specifically, the review mapped the evidence streams, triangulation depth and integration approaches reported in the six included studies. Τhese findings are presented in the Results section. The extracted findings were then organized using an EU-aligned 5D framework, and a practical protocol that produces two actionable outputs: a gap register (what is underdeveloped and why) and a confidence rating (how strong the evidence is) was proposed. The intention is to support both food businesses and competent authorities by offering a clear, traceable approach to provide evidence of FSC and to prioritise targeted improvements.

2 Materials and methods

2.1 Study design and framework

A focused mapping review was used followed by a practical protocol synthesis. The mapping review describes how triangulation has been applied in FSC assessment in manufacturing and processing settings. The protocol synthesis converts the cross-study patterns into a step-by-step, traceable workflow for FSC assessment.

The work is aligned with the EU requirement for food businesses to establish, maintain and provide evidence of an appropriate FSC (Commission Regulation (EU) 2021/382). It is consistent with the practical guidance of Commission Notice 2022/C 355/01, which includes an example FSC questionnaire structured around five core elements and explicitly recognises flexibility according to business size and nature (EC, 2022). These five elements are operationalised in this paper as a five-dimension framework: commitment, leadership, awareness, open communication and resources.

The EU-aligned 5D framework was used as a coding and reporting structure, rather than as a complete explanatory model of how food safety intentions are translated into action. It provided a practical way to organize the evidence across these five dimensions. However, FSC in daily work is also shaped by work routines, supervisory practices, organizational learning, production pressure, reporting standards and local decision-making conditions. For this reason, the protocol combines perception-based evidence (e.g., surveys and interviews) with practice-based and system/performance evidence (e.g., observation, records and verification data). The 5D framework therefore provides the structure for reporting the findings, while triangulation helps examine how these dimensions are reflected in practice.

2.2 Focused mapping review

Candidate studies were identified through targeted searches in Google Scholar and Elsevier’s ScienceDirect (publisher platform), complemented by backward and forward citation tracking. The search focused mainly on the last decade and used combinations of the terms “food safety culture” OR “food safety climate” with “assessment”, “triangulation”, “mixed methods”, and plant terms such as “manufacturing”, “processing”, “plant” or “factory”. Studies were prioritised when they provided sufficient methodological detail to extract triangulation practices and audit-relevant evidence streams. Because this was a focused mapping review rather than a systematic review, the search was targeted rather than exhaustive. The aim was not to measure how common each assessment approach is in the literature. Instead, the aim was to identify empirical studies with enough methodological detail to show how different FSC evidence streams were combined.

Database searches were last updated on 30 November 2025. The abstracts, methods and conclusions of all 33 candidate publications were reviewed. Six studies providing detailed information on evidence streams and integration approaches were selected for in-depth methodological synthesis. Study selection and data extraction were performed by a single reviewer, and references were managed in Mendeley. The study selection process is summarised in Figure 1.

The inclusion and exclusion criteria are summarised in Table 1. These criteria define the scope of the synthesis and are reflected in the “Selection focus” box in Figure 1. Eligible studies also had to include enough evidence streams to allow cross-stream comparison. The requirement for a survey plus at least two additional evidence streams was intentional. It reflected the purpose of the review: to examine triangulation design and integration logic, rather than to map all FSC assessment studies.

TABLE 1. Inclusion and exclusion criteria for the focused mapping review.

Mixed-setting studies were included only when the manufacturing or processing component could be clearly identified. The exclusion criteria are not mutually exclusive.

For each included study, the study context, participant groups, evidence streams, any reported leading indicators and lagging outcomes, and the integration approach (including any stated decision logic) were recorded. Evidence streams were coded as follows: survey/questionnaire (S), interviews (I), focus groups (FG), storytelling (ST), on-site observation (O), documents/records (D), knowledge test (K), microbiological or verification performance data (M), and food safety management system diagnostic instrument (FSMS-DI). To facilitate consistent interpretation and avoid confusion between perceptions and performance, the data sources were grouped into three practical categories (Table 2).

TABLE 2. Classification of evidence streams used in this study.

Knowledge tests (K) were treated as competence-supporting evidence (primarily relevant to the Awareness dimension). These tests can strengthen the evidence basis when aligned with practice-based or system/performance evidence. However, these tests do not on their own demonstrate applied food safety practices or FSMS, performance.

Practice-based evidence refers specifically to on-site observation (O) of food safety practices, routines and decision points during the assessment. These observations are usually time-limited. They are therefore not interpreted as evidence of stable behavioural patterns unless supported by repeated observations or by converging evidence from other streams. System/performance evidence refers to documents/records (D), microbiological or verification performance data (M), and FSMS diagnostic instruments (FSMS-DI). Together, practice-based and system/performance streams were treated as more direct evidence of applied food safety practice and FSMS functioning than perception/self-report evidence. In this paper, “compliance” is used only where the activity refers to a regulatory, certification or audit requirement, or where it is the term used by the original study. For routine workplace activities, terms such as “conformance with procedures”, “procedural adherence”, “implementation of food safety practices” or “observed food safety practices” are used instead. This helps avoid implying a formal compliance requirement where the evidence refers more generally to workplace practice. Triangulation depth was coded on a four-level scale (L0–L3) as summarised in Table 3.

TABLE 3. Triangulation depth scale (L0–L3).

2.3 Practical protocol synthesis

The practical protocol was developed by combining the EU-aligned 5D framework with cross-study patterns from the mapping review and focused on feasibility, transparency and impact. The workflow is summarised in Table 4.

TABLE 4. Workflow linking the focused mapping review to the practical triangulation protocol.

Evidence-stream abbreviations are defined in Section 2.2.

The term “confidence rating” refers to a pragmatic judgement of the strength of the evidence supporting an individual FSC gap. It considers the type and quality of evidence, the number of distinct streams, their convergence or divergence, and how clearly the integration rationale is documented. It is not a statistical confidence level and does not replace qualitative criteria such as credibility, dependability and confirmability. In this protocol, confidence ratings are intended to support CAPA prioritisation, management review, and clear and traceable documentation of the supporting evidence (e.g., for internal review, third-party audits, or competent authority review). The criteria used in this paper are summarised in Table 5.

TABLE 5. Pragmatic confidence rating criteria.

Knowledge tests (K) support competence/awareness inference but do not replace practice-based or system/performance evidence. Ratings also consider evidence quality, relevance, sampling coverage and dependence between sources. Multiple evidence streams do not automatically provide independent corroboration, and limited coverage or unresolved contradictions may justify a lower rating.

In this protocol, leading and lagging indicators were treated as complementary. Leading indicators describe upstream conditions, routines and practices linked to the dimensions of the 5D framework. They are closer to daily management and thus more controllable. From the other side lagging outcomes reflect downstream failures and are often detected through audits, complaints or verification results. Both indicator types are relevant for FSC, but they answer different questions. The proposed protocol and its reporting outputs are presented in Section 3.3.

3 Results

3.1 Included studies and triangulation map

Of the 33 candidate publications reviewed, six were selected for detailed analysis (Figure 1). All studies were conducted in manufacturing or processing settings and covered different organisational contexts (large companies and small and micro enterprises). The final set of included studies comprised Nyarugwe et al. (2018), Watson et al. (2022), Spagnoli et al. (2023), Joomun et al. (2024), Spagnoli et al. (2024) and Jespersen and Wallace (2017).

Tables 6, 7 summarise the mapped studies. Table 6 presents the evidence streams, triangulation depth (L0-L3) and reported indicators, whereas Table 7 presents the integration approach and the contribution of each study to the proposed protocol.

TABLE 6. Evidence streams, triangulation depth (L0-L3) and reported indicators in included FSC studies.

Abbreviations: S = survey/questionnaire; I = interviews; FG, focus groups; O = on-site observation; D = documents/records; K = knowledge test; M = microbiological or verification performance data; FSMS-DI, food safety management system diagnostic instrument; ST, storytelling; NR, not reported; EFSCM, enlighten food safety culture model; CoQ = cost of quality; PAF, prevention, appraisal and failure;^a CoQ is included as a broader performance-related measure.

TABLE 7. Integration approaches in the included studies and their relevance to protocol development.

Evidence-stream abbreviations are defined in Table 6.

3.2 Evidence streams, triangulation depth and integration approaches

Across the six studies, survey/questionnaire data were used in all cases (6/6). Interviews were used in most studies (5/6). Focus groups were used in one study (1/6). On-site observation was used in three studies (3/6) and document/record evidence in four studies (4/6). Two studies included structured FSMS diagnostic instruments (2/6). Knowledge tests were used in two studies (2/6). Microbiological performance data were used in one study (1/6). Storytelling was used in one study (1/6).

Using the L0-L3 coding defined in this paper, four studies were classified as L3 (Nyarugwe et al., 2018; Spagnoli et al., 2023; Spagnoli et al., 2024; Jespersen and Wallace, 2017). Two studies were classified as L2 (Watson et al., 2022; Joomun et al., 2024). All six studies combined at least three evidence streams, although the extent of cross-stream integration and the detail with which it was reported varied across studies.

Practice-based or system/performance evidence was present in five of the six studies (Table 6). Nyarugwe et al. (2018) included microbiological performance data and complaint and non-conformance records, while Spagnoli et al. (2024) examined cost-of-quality data. In the other four studies, lagging outcomes were not reported as measurable indicators.

Integration approaches differed across the six studies. Watson et al. (2022) used a sequential mixed-methods design. A survey was followed by interviews and then focus groups while integration was done by comparing quantitative scores with qualitative themes. Joomun et al. (2024) used parallel scoring across survey, observation and knowledge. They explicitly reported disagreement between survey and observation, but did not provide a detailed weighting rule. Nyarugwe et al. (2018) used cross-stream comparison to classify companies into reactive, active or proactive levels but without providing an explicit weighting scheme. Spagnoli et al. (2023) and Spagnoli et al. (2024) used explicit gap analysis and integration to produce maturity and gap outputs. Jespersen and Wallace (2017) applied method triangulation by combining a quantitative self-assessment scale with qualitative analyses of performance documents and semi-structured interviews. Findings from each stream were plotted on a food safety culture maturity model and compared across sources; however, no explicit decision rule for prioritising evidence when sources disagreed was reported.

3.3 Proposed protocol and reporting outputs

Practical protocol application starts with a simple site plan. The assessor defines the boundaries of the assessment (site, line, product family, shift pattern) and the purpose (baseline assessment, follow-up, or investigation of a known problem). Sampling decisions should be recorded in advance. As a minimum, the sample should include different job roles (operators, sanitation, maintenance, supervisors, quality) and at least two shifts where relevant.

Data collection is organised around at least three evidence streams, which capture complementary aspects of food safety culture, from what people report to what is done in practice and what the system shows. In this way, perceptions, observed practice and system functioning are all represented. The practical role of each stream in the protocol is summarised in Table 8.

TABLE 8. Evidence streams used in the protocol.

Evidence-stream abbreviations are defined in Section 2.2 and Table 2. CAPA, corrective and preventive action.

Figure 2 summarises the protocol workflow and shows how evidence streams and integration logic relate to the L0–L3 triangulation-depth scale.

Evidence is coded into the FSC dimensions, not to force each finding into a single category, but to organise the evidence in a way that supports action. For example, repeated hygiene deviations may suggest gaps in awareness and/or resources, while weak escalation of deviations may indicate gaps in open communication and leadership. Coding should be explicit and traceable; however, a short evidence summary per stream and per dimension is sufficient.

Regarding integration, a documented decision rule is used. At minimum, the assessor should state how conflicting streams will be interpreted. In this paper, a simple decision rule is proposed. When relevant practice-based or system/performance evidence indicates a weakness in applied practices or FSMS functioning, the gap is reported even if survey findings are positive, and any divergence is documented. This does not mean that survey data are ignored. Survey findings remain evidence of staff perceptions. When streams disagree, the assessor examines the relevance, quality and sampling coverage of the conflicting evidence. Greater weight is given to relevant practice-based or system/performance evidence when the judgement concerns applied practices, provided that the sampling is adequate and the documentation is traceable. If these conditions are not met, the level of confidence is reduced.

Finally, the protocol produces two outputs. The first is a 5D gap register expressed as short, actionable gap statements per FSC dimension, which identify the expected practice, the observed weakness, and the supporting evidence streams. The second is a confidence rating per gap, reported with a brief justification. The confidence-rating criteria are defined in Section 2.3 and summarised in Table 5.

In the protocol, leading indicators are used to explain and target gaps. In parallel, lagging outcomes are used to examine failure patterns and monitor changes following improvement actions over time. Examples of leading and lagging indicators are provided in Supplementary Table S1. When an indicator can be interpreted in more than one way (e.g., near-miss reporting), interpretation relies on triangulation across methods and data sources rather than on a single number. A reporting template linking gap statements, supporting evidence, confidence ratings and follow-up actions is provided in Supplementary Table S2.

4 Discussion

4.1 Why triangulation matters under EU requirements

Under EU legislation, food business operators must establish, maintain and provide evidence of an appropriate FSC (European Commission, 2021). The practical challenge is how to demonstrate that these requirements are reflected in day-to-day work. Annex III, Section 8 of Commission Notice 2022/C 355/01 describes FSC verification using surveys, interviews, observation and checks of practices, records and performance data (European Commission, 2022). The proposed protocol builds on this approach by comparing evidence from different streams and examining whether they support, qualify or contradict the same conclusion. For example, perceptions of open communication can be examined alongside evidence of how deviations are reported and followed up. In this context, triangulation is intended to make both the evidence of implementation and the basis of the assessor’s judgement more explicit, rather than simply increasing the amount of information collected. Neither triangulation nor the proposed confidence rating is prescribed by these provisions.

In practice, surveys capture perceptions across many people. Interviews and focus groups can explain why people respond as they do and can surface concrete examples. Observation captures food safety practices, routines and decision points at the time of assessment. Record review can show whether routines exist, whether problems are detected, and whether CAPA and management review loops close in practice. The complementary use of these methods is illustrated by Watson et al. (2022), who compared survey findings with interview and focus-group themes, and Jespersen and Wallace (2017), who compared self-assessment, document and interview evidence.

Questionnaires can therefore serve as a useful practical starting point for assessing food safety culture. However, when the aim is to examine how food safety practices are carried out in day-to-day work, the findings should be interpreted alongside relevant practice-based and system/performance evidence.

For a latent construct like FSC, evidence should cover both what people say and how food safety is managed in daily work. This includes decisions under pressure, reporting practices, management follow-through, and the effectiveness of feedback and CAPA loops. The combination of perception streams with practice-based and system/performance streams can offer a more defensible route to meeting the requirement to provide evidence and can help connect gaps to specific processes. However, its contribution depends on evidence quality, sampling coverage and the transparency of the integration rationale. The five EU-aligned dimensions provide a coding and reporting structure, not a complete explanation of how food safety intentions are translated into practice. Findings should therefore be interpreted in the context of work routines, supervisory practices and local reporting conditions.

For auditability, an evidence package within the proposed protocol could include: (i) a summary of survey results mapped to the five FSC dimensions, (ii) short observation notes linked to specific practices and routines, (iii) selected records that demonstrate detection and follow-up (e.g., repeat NCs and CAPA closure), and (iv) a written integration note that explains how the conclusion was reached. The aim is to use a focused set of relevant evidence, especially when the assessment is narrowly scoped to high-risk processes.

For competent authorities and third-party auditors, a confidence-rated gap register can make the basis of each reported gap explicit. Each gap statement is linked to specific evidence streams and follow-up indicators. This can support a more proportionate and transparent assessment. The scope and level of detail of the supporting evidence can be adapted to the nature and size of the business, consistent with Regulation (EU) 2021/382 (European Commission, 2021).

4.2 Practical decision rules when evidence streams disagree

Disagreement between evidence streams can be informative rather than problematic. Joomun et al. (2024) reported a discrepancy between survey and observation findings, while Jespersen and Wallace (2017) compared self-assessment, document and interview evidence within a food safety culture maturity model. These examples illustrate why both the comparison between sources and the basis for the final judgement should be reported. A formal weighting rule is not the only way to integrate evidence, and its absence does not necessarily mean that cross-stream comparison did not occur.

Without a clearly documented integration rationale, it may be difficult for external reviewers to trace how a judgement was reached, to compare successive assessments or to understand the choice of improvement actions. Changes in how evidence is interpreted may also complicate the distinction between actual improvement and differences in assessment practice. The proposed protocol therefore requires the assessor to document which evidence streams support a gap, which contradict it and how any disagreement was handled. This keeps divergence visible rather than allowing it to be obscured within a single average score. The intention is to make interpretive judgement explicit, not to eliminate it.

Operationally, evidence is organised within the five FSC dimensions and examined for each potential gap. Each evidence stream is summarised as indicating a strength, a weakness, or an unclear finding. These summaries are then integrated using an explicit rule. When relevant practice-based or system/performance evidence indicates a weakness in applied practices or FSMS functioning, the gap is reported even when survey scores are high. Greater weight is given to this evidence provided that it is adequately sampled and documented in a manner that ensures its traceability. Positive survey findings remain part of the assessment as evidence of staff perceptions. Where the evidence streams diverge, the disagreement is documented as a separate finding. This is because the divergence may indicate a mismatch between stated priorities and applied practices, or between management expectations and frontline experience.

This integration logic is intended for plant settings and depends mainly on clear and consistent documentation rather than complex numerical weighting. The documentation focuses on what evidence was reviewed, what each stream suggests, and why a particular judgement was reached. Any weighting applied locally should be stated and justified. For example, observation and records may carry greater weight when the judgement concerns the implementation of a specific practice, although their relevance, quality and sampling coverage should also be considered. A limited observation window or incomplete records may justify a lower confidence rating and the collection of further evidence. The integration rationale should therefore explain both the evidence supporting the conclusion and any limitations that may affect its interpretation.

Confidence ratings complement this approach by making the strength and limitations of the supporting evidence explicit (Table 5). They help distinguish well-supported gaps from findings that require further observation, record review or verification. A medium-confidence gap may still require action, while also needing follow-up to confirm the interpretation or assess the effectiveness of the response. Low-confidence findings are treated as hypotheses that require additional evidence. The rating describes the strength of the evidence supporting the gap. The urgency of the operational response is considered separately. Confidence therefore supports, rather than replaces, professional judgement.

4.3 Linking gap statements to leading and lagging indicators

The proposed integration rule and confidence-qualified reporting format are intended to connect assessment findings with practical priorities for improvement. A gap register is more than a list of weaknesses. It links each reported gap to the relevant FSC dimension or dimensions, supporting evidence and targeted improvement actions.

The studies included different types of performance-related information, ranging from microbiological performance and complaint and non-conformance records (Nyarugwe et al., 2018) to cost-of-quality measures (Spagnoli et al., 2024). This variation highlights the need to clearly define what each indicator represents and how it relates to the reported gap. Inconsistent reporting of lagging outcomes makes comparison across the included studies more difficult. It also limits what can be concluded about whether changes in FSC are accompanied by changes in performance over time.

Leading and lagging indicators are useful at this step, but they answer different questions. Leading indicators concern daily management routines and practices that are more amenable to targeted action. Examples include the quality of training and supervision, the availability of time and tools, and the timeliness and effectiveness of CAPA closure. Lagging indicators reflect failure outcomes, such as internal and external non-conformances, repeat deviations, complaints and failed verification results. Their combined use allows the assessment to consider both the conditions supporting food safety practices and the failures identified through the FSMS.

In practice, the protocol encourages each gap statement to be linked to at least one leading indicator and, where available, one relevant lagging outcome. For example, a gap in open communication could be tracked using data on escalation timeliness and repeat deviations, as illustrated in Section 4.5. These links can support management review by showing whether improvement actions have been implemented and whether related failures continue to occur. Follow-up findings can then guide further action or reassessment of the gap. The reporting template in Supplementary Table S2 provides a common format for recording these links. The proposed approach is intended to monitor the implementation of improvement actions and changes in food safety performance. Such monitoring, however, does not by itself show whether better performance results from improvements in FSC.

Near-miss reporting illustrates why indicator trends require contextual interpretation. High reporting counts may reflect openness and organisational learning, but they may also signal frequent process weaknesses or weak preventive control. Interpretation should therefore combine near-miss data with interviews, observation and record review. An increase in reporting should not, on its own, be taken as evidence that FSC has either improved or deteriorated. Examples of leading and lagging indicators, including the qualification relating to near-miss reporting, are provided in Supplementary Table S1.

4.4 Implementation considerations for manufacturing plants

The proposed protocol is intended for use in manufacturing plants and can be adapted to the nature and size of the business, consistent with Regulation (EU) 2021/382 (European Commission, 2021). From a practical point of view, implementation starts with a clear scope (the assessor defines which product lines, shifts, and functions are included). Including frontline and supervisory staff and, where relevant, employees from different shifts helps reduce the risk that the assessment reflects only one group or one pattern of work. It can also reveal differences in how food safety is experienced across organisational levels.

For a focused initial assessment, the proposed evidence package can combine a short survey with observation during routine work and at pressure points, together with a targeted review of relevant records. Interviews or short focus groups can then help explain the findings and explore barriers to safe practice. The scope and coverage of these sources should be matched to the questions being examined.

A short observation period can be informative when focused on hazard-relevant activities, although it cannot establish consistency over time. Relevant activities include start-up, product changeovers, particularly those involving different allergen profiles, rework and cleaning. Observation should focus on how food safety procedures are applied, rather than on housekeeping alone. It should include supervisory practices as well as operator practices. Examples include how supervisors respond to unsafe practices, whether they follow hygiene procedures themselves, and how they decide whether production should be stopped when a food safety problem arises.

As regards record review, it should be selective and purposeful. The aim is to see whether the management system functions as a learning system. This can be done by focusing on records that include repeat non-conformances, CAPA closure rates, time-to-close actions, verification results, training records, and management review minutes. Repeated non-conformances with limited follow-up may prompt closer examination of commitment or leadership. Weak escalation may point to communication difficulties, while constraints related to equipment, staffing or time may suggest resource-related gaps. These interpretations should be checked against observations and staff accounts rather than inferred from records alone.

The quality of the assessment also depends on the local reporting context. Fear of blame or negative consequences may discourage staff from reporting deviations and near-misses. Interpretation should therefore consider the conditions that encourage or inhibit reporting (Wiśniewska, 2023; FoodDrinkEurope, 2023). Interview prompts should be phrased neutrally and focus on barriers to safe practice rather than on attributing blame to individuals. For example, an assessor might ask: “What makes it easier or harder to report a deviation during a busy shift?” This can help bring out issues related to working conditions and reporting barriers without placing the focus on individual fault.

Spagnoli et al. (2023) combined diagnosis and gap analysis to identify underdeveloped aspects of FSC as a basis for improvement planning. The proposed gap register serves a similar planning purpose by linking findings to actions and follow-up. To support auditability and follow-up in the workplace, Supplementary Table S2 provides a standard format for concise gap statements, supporting evidence and the confidence judgement. A useful way to implement the protocol is to align it with the FSMS’s existing procedures. For example, the gap register can be reviewed during management review, and confidence-rated gaps can be linked to the CAPA plan, with actions assigned to the relevant functions. The aim is to make FSC assessment part of routine FSMS improvement activities, rather than a separate exercise led only by the quality team.

Timing and frequency can also be adapted to the plant’s circumstances. A full triangulated assessment could, for example, be scheduled annually or following major organisational or process changes, while shorter follow-up checks could focus on previously identified gaps. The frequency should reflect the purpose of the assessment, operational risks and previous findings, rather than a fixed annual schedule. Repeated assessments can then show how identified gaps are being addressed and whether the relevant practices are maintained over time.

4.5 Illustrative example of a confidence-rated gap statement

To illustrate how the protocol can be used in practice, Example 1 presents a hypothetical example of a confidence-rated FSC gap statement. The example is not based on data from an actual production facility. It is used only to demonstrate how disagreement between evidence streams can be documented and translated into a CAPA-ready output.

Table 9.

S = survey/questionnaire; O = on-site observation; D = documents/records; I = interviews; NCs, non-conformances; CAPA, corrective and preventive action.

Example 1 Hypothetical confidence-rated FSC gap statement.

This example illustrates the importance of the integration rule. The positive survey findings are useful as evidence of staff perceptions, but they do not override observations, records, and interviews when these streams indicate a recurring operational weakness. The evidence supports targeted action on handover and reporting procedures. Given the limited observation coverage, a medium confidence rating is appropriate, with follow-up needed after CAPA implementation.

4.6 Strengths, limitations and future work

This study used published literature only. No personal data were collected by the authors, and ethics approval was not required. Because this was a focused mapping review based on targeted searches in Google Scholar and Elsevier’s ScienceDirect complemented by citation tracking, some relevant studies may have been missed. Screening and data extraction were performed by a single reviewer, which may increase the risk of selection bias and data-extraction errors. Selection prioritised studies with detailed methodological reporting, which may favour well-documented assessments. A complete study-level record of reasons for non-selection is not available, so their relative frequencies cannot be reported reliably.

The included studies differed in sector, company size and assessment instruments. Their findings were therefore synthesised descriptively rather than quantitatively. Lagging outcomes and performance metrics were also reported inconsistently across studies.

This work translates mixed-methods FSC research into a practical protocol intended to support audits in food manufacturing and processing settings. It provides a typology of evidence streams, a triangulation depth scale and pragmatic confidence-rating criteria. The review was deliberately focused on empirical FSC or food safety climate studies conducted in manufacturing or processing settings and using multiple evidence streams that allowed cross-stream comparison. The patterns mapped from these six studies should therefore be interpreted as illustrative rather than exhaustive.

Finally, the L0–L3 triangulation depth scale and the pragmatic confidence rating criteria proposed in this paper have not yet been validated. They are intended to support transparent, traceable and verifiable reporting, rather than to replace professional judgement. The proposed triangulation protocol aims to improve transparency, but it does not eliminate interpretive judgement. Observation, interviews and document review remain influenced by sampling decisions, organisational context, assessor experience and the quality of available records. Evidence streams may also draw on overlapping sources, so agreement between them should not automatically be treated as independent corroboration. For this reason, confidence ratings should be understood as practical statements about the strength of the available evidence, rather than as expressions of statistical certainty.

Future work should assess reliability between assessors. It should also evaluate whether the protocol improves the targeting and effectiveness of improvement actions. Another priority is to test the protocol in different regulatory and organisational contexts, including small businesses with limited data systems. Future research should also explore how lagging outcomes can be reported more consistently in FSC studies to support assessment of changes in food safety performance over time. According to Spagnoli et al. (2025), intervention portfolios and expert-consensus approaches can support improvement planning after diagnosis by providing options for developing human-organisational FSC dimensions. Design-thinking intervention studies in food service settings (e.g., Ferreira et al., 2025) also illustrate possible strategies for change, although they fall outside the manufacturing/processing inclusion criteria of the present mapping.

From a reporting perspective, future FSC papers could improve comparability by consistently documenting the evidence streams used, the sampling plan, the triangulation depth and the integration rule. Reporting should also specify the practice-based and system/performance evidence considered, together with any lagging outcome indicators included. Even a short checklist could make both the assessment process and the basis of the conclusions easier to evaluate.

Future applied work should examine whether confidence-rated gap reporting influences management decisions and follow-up actions. Relevant questions include whether documenting the evidence supporting each gap increases management engagement, and whether linking actions to leading and lagging indicators improves CAPA follow-up and effectiveness. Addressing these questions would help establish the practical value of the protocol for businesses and regulators.

A further priority is to test the protocol during periods of organisational change. Regulation (EU) 2021/382 emphasises the need to maintain the integrity of the food hygiene system during change, together with appropriate training, supervision and sufficient resources (European Commission, 2021). Future studies could examine how the protocol performs during new product introduction, staff turnover or equipment changes. For example, observation during start-up could assess whether temporary staff receive adequate supervision, while record review could determine whether training and verification arrangements have been updated accordingly. Such studies could also assess whether the gap register helps identify and justify resource needs.

5 Conclusion

Across the six included studies, the approaches to cross-stream integration and the reporting of lagging outcomes varied. Surveys were present in all included studies and remain useful for assessing food safety climate. However, questionnaire findings should not be interpreted in isolation when the aim is to draw conclusions about how FSC is applied in daily work.

To improve validity and practical usefulness, an EU-aligned triangulation protocol is proposed that links the 5D requirements to multiple evidence streams, leading indicators and lagging outcomes. The protocol produces two audit-defensible outputs: a 5D gap register and a confidence rating for each reported gap. These are intended to support food businesses and competent authorities by making the evidence supporting each gap explicit and informing targeted improvement planning. Future work should test feasibility, reliability and practical impact of the protocol in real audits and improvement cycles, including how confidence-rated reporting influences decision-making and sustained FSC improvement.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frfst.2026.1952161/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Nikolaos D. Andritsos, University of Patras, Greece

Reviewed by: Juan Moreira, Colorado State University, United States

Jason Barnes, La Trobe University, Australia