From Extraction Yield to Defensible Route-Level Claims: A Critical Integrative Review of Laboratory-to-Industrial Translation in Plant-Derived Bioproduct Valorization

1. Introduction

Agri-food processing byproducts and other plant-derived resources can supply bioactive compounds and functional ingredients [1,2,3]. Valorization studies commonly begin with laboratory extraction and screening. Yet gross extraction yield alone cannot establish specification-conforming output, benchmark-equivalent service, or environmental improvement when laboratory findings are extrapolated to pilot- or industrial-scale configurations. Gross extraction yield, marker concentration, and related screening metrics remain intermediate process measures [4,5]. Accordingly, these measures are treated as intermediate evidence rather than as proof of downstream route-level claims.

Route-level evaluation begins with the resource: source, tissue, handling, composition, moisture, stability, safety, and processability determine recoverability [2,6,7,8]. Side streams also differ from dedicated crops because their existing uses and burden origins matter. For comparative claims, a zero-burden convention at generation does not remove baseline fate, foregone beneficial functions, or residual-management consequences when diversion displaces feed, composting, anaerobic digestion, or another use [9,10]. Functional unit, system boundary, allocation, and displaced-function assumptions can therefore change LCA conclusions [11].

Route identity must be maintained beyond extraction and across scale. Sequential separation changes target retention, whereas concentration, drying, and carrier addition change final product mass and properties [12,13]. In this review, laboratory-to-industrial translation denotes the evidentiary progression from laboratory measurements to pilot-scale, modeled scale-up, or industrially relevant claims; it does not imply that every retained package documents direct scale-up. Each change in scale is therefore treated as a distinct route configuration or an explicitly reconciled variant, with feed concentration, unit operations, product basis, utilities, and residual fate documented [14,15]. A conformity claim concerns a final product assessed against a predeclared specification. An equivalence claim additionally requires benchmark-equivalent service under matched test and use conditions. Modeled substitution potential requires a defined supply–demand context and realization assumptions, whereas observed displacement requires empirical evidence that uptake reduced the procurement, production, or consumption of a named incumbent within a defined market or supply relationship [16].

Evidence needed to evaluate this progression spans seven streams: feedstock identity; recovery and profiling; downstream processing and formulation; product quality and readiness; circularity and substitution; techno-economic assessment (TEA); and life-cycle and counterfactual assessment. These streams serve different purposes and often terminate at different outputs. Product studies report extracts, products, or application performance, whereas TEA and LCA model economic or environmental outcomes under defined assumptions [2,15]. Circularity metrics also embed choices about resource state, scope, and allocation; increased circulation alone therefore does not establish environmental improvement [17,18,19,20,21]. The central challenge is whether load-bearing evidence refers to the same feedstock, process sequence, scale, product state, function, and comparison basis or whether differences are explicitly reconciled.

Building on recent syntheses of material readiness, green recovery and profiling, and integrated ingredient processing [22,23,24], the present review addresses a distinct analytical question: whether experimental, downstream-processing, product-readiness, TEA, LCA, substitution, and market evidence jointly support the same route configuration and claim. The principal unit of evidence coding was the evidence package; route configurations and claim-support chains were retained as nested interpretive records only when source-reported outcomes were separable. Specification-adjusted route accounting is used as a synthesis and claim-calibration aid: successive output states are separated, evidence is linked through a common route key, and each claim stops at the first unsupported or incompatible load-bearing link within the reviewed evidence record. Technical realization, physical resource productivity, economic plausibility, environmental improvement, modeled substitution potential, and observed displacement remain distinct, while economic and environmental pathways remain parallel. Application to selected cases illustrates the audit logic but does not validate it.

This distinction is consequential because evidence may be strong within one domain yet nontransferable to the next. A well-characterized extract does not by itself establish final-product conformity; a favorable TEA does not establish environmental improvement; and a modeled substitution credit does not demonstrate market displacement. A route-level review therefore requires both a stable unit of comparison and explicit stopping rules. Without them, results obtained for different product states, scales, functions, or counterfactuals may be combined into a conclusion that none of the underlying sources supports. Recent reviews across biomass- and waste-valorization domains have synthesized feedstock-specific technologies and application pathways [25,26,27,28,29,30]. The present contribution instead examines whether heterogeneous evidence supports the same configuration-level claim.

The audit was applied to 17 author-curated evidence packages comprising 45 nested analytical units across molecular-identity, composition-constrained, and function-constrained product archetypes. The selected set includes agri-food byproduct routes alongside other plant-derived systems that provide contrasting product, process, scale, and evidence structures. The scope comprised selected non-energy plant-derived products rather than all biomass-valorization pathways. Fuels, biochar, compost, construction materials, protein- or starch-dominant products, and broad multi-output biorefineries were outside the primary case set except when considered as baselines, residual pathways, or comparators. The 45 units were nested within 17 packages and were not treated as independent studies or replication units. The set examines decision consequences; it does not estimate field-wide prevalence, route-success probability, or audit validity.

The review asks which feedstock, process, scale, product, and comparison fields must align; when laboratory evidence can support pilot- or industrial-scale claims; how recovery should advance to specification-conforming output and benchmark-equivalent service without being conflated with market uptake; and what evidence supports each of six route-level claim classes. The objective is not to rank routes or generate a sustainability score, but to identify the strongest reporting- and linkage-compatible statement and the next evidence needed to advance it. Laboratory-to-industrial translation is accordingly treated as an evidence-alignment problem rather than an automatic consequence of higher extraction yield or increased production scale.

2. Review Design and Evidence Architecture

2.1. Review Design, Source Identification, and Criterion-Based Case Selection

This critical integrative review combined cross-stream synthesis with route-level evidence audits to determine how unsupported or incompatible load-bearing links constrain the strongest defensible claim. The design supported conceptual integration and claim calibration, not exhaustive enumeration, meta-analysis, field-wide frequency estimation, or formal risk-of-bias assessment [31,32,33]. Each of the seven substantive streams—feedstock identity; recovery and profiling; downstream processing and formulation; product quality and readiness; circularity and substitution; techno-economic assessment (TEA); and life-cycle and counterfactual assessment—required at least two independent anchors and one source addressing its principal contested boundary. Independent anchors excluded publications authored by the present author, while those reviews were retained only as continuity sources.

Searches supporting coverage verification and route auditing were conducted in Scopus on 11 August 2026 and in the Web of Science Core Collection and PubMed on 12 August 2026. In Scopus, Advanced document search used the TITLE-ABS field; in Web of Science, Advanced Search used separate Title and Abstract fields; and in PubMed, searches used Title/Abstract [tiab]. Where applicable, no user-defined filters were applied for date, language, document type, subject area, species, access status, or full-text availability. The final searches retrieved 63,465 records in Scopus, 48,706 in Web of Science, and 19,592 in PubMed. These platform-specific result sets served as audit snapshots for verifying predefined concepts and route coverage; they were neither pooled nor treated as a screening denominator or eligibility corpus. Supplementary Table S1 provides the exact search strings, fields, dates, restrictions, record counts, anchor-gate results, export scope, and rationale for duplicate handling.

Thirty-eight predesignated audit sources supported the cross-stream synthesis and route audit: 15 were checked in full text, one additional load-bearing source [34] was checked in official publisher-hosted unedited full text, and 22 were checked through official abstracts, metadata, or other accessible records. The remaining cited sources provided methodological or contextual support and were not part of this source-access tally. Load-bearing references were checked against official publisher or Version of Record metadata and, where accessible, full text to verify bibliographic identity, claim scope, numerical conditions, system boundaries, source locators, and publication status. Coding for access-limited sources was restricted to claim elements explicit at the available level. The original searches and source-status checks were conducted on 11–13 August 2026; the access status of CC-01 and MI-01 was reverified on 26 August 2026, and the publication status of CC-02 was reverified on 3 September 2026. Package-level access status and interpretive limits are documented in Supplementary Workbook S1. Targeted reference-list and cited-by checks resolved specific gaps; no exhaustive citation-network search was undertaken. NR was assigned only when a sufficiently complete record supported a nonreporting judgment. When full text was unavailable, absence from an abstract or metadata record was not treated as evidence of nonreporting; the field was coded P and labeled not verifiable unless the accessible record explicitly established absence.

The targeted-audit backbone and scope-challenge candidates were provisionally assembled and checked by 8 August 2026, before the 11–12 August database snapshots; source-specific verification and targeted route-completion checks continued through 13 August 2026. The database snapshots were used only to verify predefined stream coverage and resolve documented gaps, not as an eligibility corpus or sequential screening pool. The candidate-source audit comprised 17 retained evidence packages, five contextual sources, and eight excluded candidates. Retention as a route-level evidence package required (i) a clearly defined non-energy plant-derived product, (ii) sufficient feedstock, process, and product information to define at least one configuration, (iii) at least one claim-relevant evidence link, and (iv) a publicly verifiable source record. Candidates were excluded because of product-scope mismatch, insufficient route identity, duplication without a distinct evidence link, an incompatible product or process basis, or unverifiable load-bearing information. The 17 packages were author-curated from a preassembled targeted-audit backbone and were not selected by screening the database snapshots. They constitute a purposive demonstration set chosen to expose contrasting evidence structures; completeness of candidate capture and freedom from selection bias cannot be inferred. Core and boundary-condition labels indicate analytical roles only, not representativeness, evidence quality, or validation status.

2.2. Evidence Units and Route-Linkage Rules

Two organizational levels were distinguished to prevent evidence from being assigned to a finer level of inference than its source supported: evidence packages and nested analytical units. An evidence package comprised a single publication or a set of publications explicitly linked by a documented common route. A route family was used only as a descriptive grouping for a shared feedstock–product objective and was not counted as an analytical unit. The nested analytical units comprised parent route configurations, source-reported configuration variants, modeled scenarios, and paired comparisons. A parent route configuration was defined by feedstock condition, unit-operation sequence, scale, final product state or grade, utility basis, and residual fate. A configuration variant was a source-reported experimental or operational alternative nested under a parent configuration. A modeled scenario represented a source-defined change in one or more decision variables within a configuration. A paired comparison represented a within-source contrast between two separable configurations and was retained as a comparison-level unit rather than treated as an independent study. The functional unit, comparator, and counterfactual were recorded as additional linkage fields when required by the claim. Sources were grouped into an evidence package only when the source records established a common route identity or provided sufficient information to reconcile load-bearing differences in feedstock basis, unit-operation sequence, product definition, and scale or model basis. Grouping did not itself establish compatibility; each claim-relevant linkage was coded separately, and any reconciliation was limited to the documented conversion or assumption. In this review, laboratory-to-industrial translation was operationalized as a cross-scale linkage question rather than as an assumed progression. Laboratory measurements, pilot-scale observations, modeled scale-up scenarios, and industrial-scale evidence were each interpreted only at their documented unit of inference; linkage across them required alignment or explicit reconciliation of load-bearing differences in feed concentration, unit-operation sequence, product basis, utilities, residual fate, and scale or model basis.

Cross-stream evidence supported a common route-level claim only when claim-required fields were explicit—or partial for a bounded conclusion—and every load-bearing cross-source or cross-basis linkage was aligned or explicitly reconciled. An R code required a traceable record of original and converted bases, the transformation or assumption, and the affected claim. Unresolved differences in feedstock basis, scale, unit operations, product grade, utilities, residual fate, functional unit, comparator, counterfactual, geography, time, or impact method were incompatible when load-bearing and stopped the path. Thus reference-substance prices were not assigned to crude extracts, laboratory and industrial drying were not linked without reconciliation, and beneficial side-stream uses were not assigned a default zero burden.

The 17 evidence packages were decomposed into 45 nested analytical units: 17 parent route configurations, 15 source-reported configuration variants, 12 modeled scenarios, and one paired comparison. These units were nested within the packages and were not treated as independent studies, replication units, or denominators for estimating field-wide prevalence. Supplementary Figure S1 and the Package Register, Configuration Index, and Unit of Inference sheets in Supplementary Workbook S1 document package roles, configuration and scenario labels, parent–child linkages, source-reported separability, and the permitted and prohibited inference for each analytical unit. Package-level matrices summarize evidence coverage only. Configuration-specific records were retained where outcomes were separable, but this retrospective audit does not provide a complete configuration-by-claim implementation of A01–A08. The reported ceilings should therefore be interpreted as structured qualitative package-level judgments rather than fully reproducible route-configuration calculations.

2.3. Coding, Provenance, Compatibility, and Internal Consistency

Each of the 170 package–field decision cells (17 packages × 10 evidence fields) was coded separately along three dimensions: availability and completeness (E, explicit; P, partial or bounded; NR, required for the evaluated claim path but not reported; NA, not applicable); provenance and derivation status (SR, source-reported; AD, author-derived; MS, source-reported modeled scenario; SR/AD, source-reported inputs combined with an author-derived value); and linkage compatibility (A, aligned; R, explicitly reconciled; I, incompatible). Availability and provenance were coded for every available package–field evidence record, whereas compatibility was assigned only when a cross-source or cross-basis linkage could meaningfully be evaluated. MS was used only for a value or statement directly reported from a source-defined modeled scenario. SR/AD denoted a value calculated from traceable source-reported inputs, whereas AD denoted an author-constructed scenario, classification, or other derivation that introduced author-specified assumptions. NL was assigned when an evidence record existed but no cross-source or cross-basis linkage was invoked; it was not an availability or completeness code. A blank compatibility cell was reserved for the absence of an evidence record to which compatibility could meaningfully be assigned. A P code permitted only a bounded or conditional conclusion. NR was assigned only when the field was required for the evaluated claim path but not reported, whereas NA was assigned only after the field was determined not to apply to that path. No unreported load-bearing parameter was imputed.

The Cell Decision Master and Evidence Record Log sheets in Supplementary Workbook S1 contain the primary cell-level records for the ten Figure 4 fields, including source records, rationales, provenance, compatibility, calculations, and reconciliations. The Availability, Provenance, and Compatibility Matrix sheets summarize those records, while the Claim Path Stop Audit records claim-specific stopping decisions and qualitative auxiliary considerations. Figure-based summaries are presented with the cross-package synthesis in Sect. 4.7. All source selection, extraction, classification, and appraisal were conducted by the single author, who then performed a structured second-pass internal-consistency audit of source locators, units, denominators, package and configuration identifiers, author-derived calculations, and reported values across the manuscript, tables, figures, and workbook. This procedure constituted an internal audit; it did not constitute independent duplicate source selection or data extraction, inter-rater calibration, or external validation.

2.4. Specification Evidence and Claim-Relevant Appraisal

Product-related evidence was appraised across four distinct product-evidence states: (1) a characterized endpoint; (2) a predeclared acceptance specification; (3) demonstrated conformity of the assessed batch or product stream with that specification; and (4) repeat-batch conformity or documented commercial release. A post hoc comparison with values reported in the literature or for a commercial product was not treated as a predeclared specification, and conformity of a single batch did not establish repeatability or routine release. Documented commercial release, where available, was treated as product-release evidence rather than evidence of observed displacement; observed displacement required documented replacement of an incumbent within a defined market or supply relationship.

Reporting completeness, applicability, and source-role checks were documented separately using source-type-specific signaling questions in Supplementary Workbook S1. These checks were not formal risk-of-bias or certainty ratings, and descriptive concern labels did not raise claim ceilings. An NR or I code—or explicit failure of a required acceptance or equivalence criterion—stopped the affected path, whereas P allowed only a bounded conclusion. The architecture is a review-specific audit aid, not a validated risk-of-bias tool, GRADE assessment, certification scheme, reporting standard, or certainty instrument.

2.5. Claim Paths, Accounting Bases, and Uncertainty Treatment

Six claim classes were evaluated separately: technical realization, physical resource productivity, economic plausibility, environmental improvement, modeled substitution potential, and observed displacement (Table 1). These classes define distinct claim paths rather than a single maturity ladder. A defined feedstock–process–product route and a traceable, claim-appropriate technical basis were treated as shared prerequisites. Empirical technical realization was required only when realization was claimed; purely modeled TEA or LCA results remained admissible as model-conditional evidence and were not reclassified as realized routes. Comparative claims additionally required an explicit comparator and, where product grade, function, or use level differed, evidence of product or service equivalence. Economic plausibility and environmental improvement were evaluated in parallel, and evidence for one path was not used to establish the other. The maximum reporting- and linkage-compatible conclusion on each path was capped at the first unsupported or incompatible load-bearing link, including a required-but-not-reported field or explicit evidence that a claim-required criterion was not met. Modeled substitution potential required an explicit comparator, benchmark-equivalent service, a defined supply and demand context, and stated post-cap realization assumptions. Observed displacement was evaluated independently and required empirical evidence linking uptake of the recovered product to reduced procurement, production, or consumption of a named incumbent within a defined market or supply relationship; it did not require a prior substitution model.

Table 1 summarizes the shared prerequisites, load-bearing evidence requirements, stopping conditions, and maximum reporting- and linkage-compatible conclusion for each claim path.

A TEA result was admitted only at its reported unit of inference and required, as applicable, a route-matched throughput and utilization basis, currency and price year, CAPEX/OPEX boundary, coproduct and residual treatment, quality-control and rejection assumptions, decision metric, and sensitivity or uncertainty analysis. Missing load-bearing fields limited the conclusion to the last supported economic statement. A deterministic point estimate and a probabilistic risk result were retained as different model outputs rather than merged into one scenario ranking.

Accounting fields were grouped into feedstock bases, output states, physical resource inputs, and residual or co-product outputs. Physical resource productivity was defined only for a named output state paired with a matched physical input over the same route, mass basis, and system boundary; the numerator was not advanced from stage recovery to specification-conforming output or benchmark-equivalent service without the corresponding evidence. Concentration was not treated as recovery, gross solvent circulation was not treated as fresh-solvent use, and land productivity was considered defensible only when upstream attribution to the audited route was explicit. No package-level physical resource-productivity ratio was calculated in the retrospective case audit because the workbook did not contain a complete, route-aligned output numerator and attributable physical-input denominator for any applicable claim path; those paths were therefore recorded as not evaluated or capped, rather than inferred from gross recovery or inventory availability.

Multi-output systems were handled through a review-specific sequence for internal route consistency, not a universal allocation hierarchy. Subdivision was preferred when operations and inventories were separable; otherwise, the source-reported system-expansion or allocation method was accepted only when explicit and justified. Substituted product and ratio, or the allocation basis, had to be reported. Shared flows and burdens were partitioned or retained within the complete product basket so that avoided baseline fate and displaced function were not credited twice. Mutually exclusive scenarios were not summed or treated as independent evidence [11,16]. Product-specific productivity ratios were reported only when inputs could be assigned consistently; otherwise, the basket and shared input profile were retained.

Uncertainty was represented using only source-reported ranges, source-reported probabilistic outputs, and explicit scenarios. Author-constructed scenarios were labeled, and no new probabilistic analysis was performed. Sequential recoveries were multiplied only for conditionally nested denominators with preserved target identity. Service adjustment required matched tests and defensible proportionality; partial evidence was P, missing required evidence NR, and an incompatible basis or failed criterion capped the path. Preference was described as stable only within a reported range or scenario set. A quantified switching point or bounded interval supported a conditional preference, not a route-independent ranking. No unreported distributions, cross-case averages, meta-analysis, or composite score were introduced; Supplementary Workbook S1 records transformations and denominator risks.

Equations (1)–(9) are deterministic accounting relations and operational definitions, not calibrated predictive models. They require matched units, boundaries, denominator structures, and explicit route assumptions; without these inputs they do not predict process performance, market uptake, displacement, or impacts. Equation (1) was not retrospectively instantiated when generated mass, route-accessible fraction, and incoming-feed acceptance were not reported on a common basis. Among Equations (1)–(9), only the conditionally nested recovery relation was instantiated numerically in the retrospective cases; the remaining equations served as proposed accounting definitions and claim-eligibility boundaries.

3. Evidence-Linkage Synthesis Through Specification-Adjusted Route Accounting

3.1. Product Archetypes and Specification Anchors

The audit logic distinguishes three product archetypes because the evidence required to establish specification-conforming output depends on the nature of the claimed product. These archetypes are analytical categories used to select claim-appropriate evidence rather than an exhaustive product taxonomy. Molecular-identity products, including purified compounds and reference substances, may require evidence of identity, assay or purity, stereochemical control, relevant impurities and degradation products, residual solvents, contaminants, and traceability, depending on the claimed grade and intended use [35,36]. For reference materials, purity assignment should identify the measurement basis and associated uncertainty and may integrate mass balance, quantitative nuclear magnetic resonance, thermal methods, and other direct assays [37]. Composition-constrained products, such as standardized botanical extracts, require predeclared compositional acceptance criteria, including marker ranges or fingerprints, together with contaminant limits, stability-indicating evidence, and batch-level conformity [38,39]. Function-constrained products, including pectin and dietary-fiber or polysaccharide fractions, require predeclared performance criteria evaluated under defined test and use conditions. When equivalence or substitution is claimed, the comparator and the product use level required to deliver the same defined service must also be specified; stability and safety require separate evidence [16,40,41]. The cited standards and guidance serve as intended-use- and grade-specific specification anchors rather than universal release criteria. In particular, specifications issued by the Joint FAO/WHO Expert Committee on Food Additives (JECFA) for pectin are relevant only when food-additive-grade pectin is claimed.

Within the route audit, these archetypes are used to identify category errors in the assignment of product grade, price basis, functional unit, comparator, or substitution credit. A standardized extract should not inherit the price basis or substitution credit of a purified compound unless grade equivalence is demonstrated, and an insoluble recovered fiber should not be treated as equivalent to a commercial fiber unless functionality and intended use are matched. For multi-marker or multifunctional products, satisfaction of a single favorable criterion does not establish conformity with all predeclared specifications or benchmark-equivalent service.

3.2. From Generated Feed to Specification-Conforming Output and Benchmark-Equivalent Service

Throughout Section 3, $t$ denotes a target compound or constituent, $p$ a final product stream, $s$ a defined service or application, $j$ a unit operation, $i$ a physical resource input, and $h$ an environmental impact category.

For readability, route-configuration, scale, and temporal indices are suppressed in Equations (1)–(9). Their omission does not permit cross-scale pooling: inputs from laboratory, pilot, modeled industrial, and operating industrial configurations may be transferred or combined only when the load-bearing feedstock, unit-operation, product, utility, residual-fate, temporal, and system-boundary bases are aligned or explicitly reconciled. Calculation of an equation-defined quantity does not by itself raise a claim ceiling; the required inputs and linkages must also satisfy the availability, compatibility, and stopping rules defined in Section 2.

Route accounting begins with the mass generated or produced within the defined temporal and system boundaries, rather than only with the mass introduced into a laboratory experiment (Figure 1). Accepted feed is calculated using Equation (1):

$$M_{\mathrm{a}\mathrm{c}\mathrm{c}}=M_{\mathrm{g}\mathrm{e}\mathrm{n}}\times f_{\mathrm{a}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e}}\times f_{\mathrm{i}\mathrm{n}}$$

(1)

where $M_{\mathrm{g}\mathrm{e}\mathrm{n}}$ is the mass generated within the defined temporal and system boundaries; $f_{\mathrm{a}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e}}$ is the fraction accessible to the route after accounting for collection, seasonality, storage, and existing uses or prior commitments that constrain diversion; and $f_{\mathrm{i}\mathrm{n}}$ is the fraction of accessible feed that satisfies the predeclared incoming-feed specification. $M_{\mathrm{a}\mathrm{c}\mathrm{c}}$ therefore denotes accepted feed. $M_{\mathrm{g}\mathrm{e}\mathrm{n}}$ and $M_{\mathrm{a}\mathrm{c}\mathrm{c}}$ must be expressed on the same wet- or dry-mass basis, and $0\le f_{\mathrm{a}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e}},f_{\mathrm{i}\mathrm{n}}\le 1$. Availability denotes access to the route; it neither confers a zero burden nor eliminates opportunity cost. If diversion foregoes a beneficial use, the foregone function must be represented in the environmental and, where relevant, economic counterfactuals. Any constraint represented in $f_{\mathrm{a}\mathrm{v}\mathrm{a}\mathrm{i}\mathrm{l}\mathrm{a}\mathrm{b}\mathrm{l}\mathrm{e}}$ must not be applied again through $f_{\mathrm{i}\mathrm{n}}$.

For a route defined by a discrete target compound or constituent, Equations (2) and (3) quantify the initial target mass present in the accepted feed and the final target mass retained after sequential recovery:

$$T_{0,t}=M_{\mathrm{a}\mathrm{c}\mathrm{c}}\times c_{0,t}$$

(2)

$$T_{\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l},t}=T_{0,t}{\displaystyle {\prod }_{j=1}^{J_t}}{\eta }_{j,t}$$

(3)

where $c_{0,t}$ is the mass fraction of target $t$ in the accepted feed, expressed as kilograms of target per kilogram of accepted feed on the same wet- or dry-mass basis as $M_{\mathrm{a}\mathrm{c}\mathrm{c}}$. Other units must be converted before use, and $0\le c_{0,t}\le 1$. ${\eta }_{j,t}$ is the conditional recovery of target $t$ through operation $j$, relative to the amount of that target entering the operation; $0\le {\eta }_{j,t}\le 1$, and $J_t$ is the number of nested recovery operations. Recovery fractions may be multiplied only when their denominators are conditionally nested and target identity is preserved across operations. Equation (3) applies only to a single, conditionally nested target-recovery sequence; routes involving parallel target-bearing streams, recycle, or recombination require explicit stream-wise mass balances, with recovered target masses summed only across nonoverlapping streams. Critically, $T_{\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l},t}$ is the mass of target $t$ retained after recovery, whereas $Q_{\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l},p}$ is the total isolated mass of product stream $p$ before release assessment and may include non-target constituents. These quantities are not interchangeable.

For a function-constrained route that is not defined by a discrete target constituent, accounting may instead proceed from accepted feed to total isolated product mass and then to specification assessment.

Specification-conforming output is calculated using Equation (4):

$$Q_{\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c},p}=Q_{\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l},p}\times f_{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{e},p}$$

(4)

where $f_{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{e},p}$ represents the fraction of isolated product stream $p$ that satisfies all predeclared release specifications. For a single homogeneous batch, $f_{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{e},p}$ is binary: 1 if every specification is met and 0 otherwise. A fractional value may be used only when conforming and nonconforming fractions are physically separated and quantified or when expected output is estimated across multiple batches or a defined annual production period. Across multiple batches or a defined production period, the grade fraction must represent the mass-weighted conforming fraction of total isolated output rather than the unweighted proportion of batches passing. If either final product mass or a predeclared acceptance rule is unavailable, $Q_{\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c},p}$ cannot be established, and the claim ceiling remains at the last directly supported output state. Establishing $Q_{\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c},p}$ does not, by itself, establish benchmark-equivalent service or market uptake. The formal domain is 0 ≤ $f_{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{d}\mathrm{e},p}$ ≤ 1.

For any product comparison in which equivalent service depends on relative use level or functional performance, benchmark-equivalent service may be calculated using Equation (5) only after specification-conforming output and comparable performance data have been established:

$$S_{\mathrm{e}\mathrm{q},p,s}=Q_{\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c},p}\times {\displaystyle \frac{u_{\mathrm{r}\mathrm{e}\mathrm{f},p,s}}{u_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t},p,s}}}$$

(5)

where $u_{\mathrm{r}\mathrm{e}\mathrm{f},p,s}$ and $u_{\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{t},p,s}$ are the respective amounts of the reference and recovered products required to provide the same defined service $s$ under matched test and use conditions. The use ratio in Equation (5) is admissible only when the compared tests address the same service, the relevant conditions are matched, and proportional scaling is defensible over the range to which the ratio is applied. $S_{\mathrm{e}\mathrm{q},p,s}$ is expressed in reference-product-equivalent units for service $s$; it is neither recovered-product mass nor evidence of market uptake. Both amounts must be strictly positive and expressed on compatible measurement bases.

Market uptake remains distinct from both specification-conforming manufacturing output and benchmark-equivalent service. Within the broader claim class of modeled substitution potential, scenario-constrained substitution potential is calculated using Equation (6) only after benchmark-equivalent service has been established and a defined demand context and realization scenario have been specified:

$$S_{\mathrm{s}\mathrm{u}\mathrm{b},scenario,p,s}=\mathrm{m}\mathrm{i}\mathrm{n}\left(S_{\mathrm{e}\mathrm{q},p,s},D_{\mathrm{e}\mathrm{q},p,s}\right)\times f_{\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{l},p,s}$$

(6)

where $D_{\mathrm{e}\mathrm{q},p,s}$ is the addressable demand expressed in the same units as $S_{\mathrm{e}\mathrm{q},p,s}$, and $f_{\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{l},p,s}$ is the scenario-specific fraction of the supply- and demand-capped potential expected to reduce incumbent consumption; $0\le f_{\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{l},p,s}\le 1$. Because it is applied after the minimum operator, $f_{\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{l},p,s}$ must represent only post-cap realization constraints and must not duplicate the relative-functionality adjustment already incorporated in Equation (5) or the supply and demand caps incorporated in Equation (6). $S_{\mathrm{s}\mathrm{u}\mathrm{b},scenario,p,s}$ is therefore a scenario-conditional model output that supports, at most, a modeled-substitution-potential claim; it is not an observation of displacement. Neither modeled substitution potential, scenario-constrained substitution potential, nor an LCA substitution or avoided-burden credit constitutes observed displacement. For this review, observed displacement was operationalized as empirical evidence linking recovered-product uptake to reduced procurement, production, or consumption of a named incumbent within a defined market or supply relationship; this operationalization is consistent with the displacement-accounting boundary discussed by Vadenbo et al. [16]. The addressable demand must also be matched to the defined service, comparator grade, geographic or market boundary, and temporal basis.

3.3. Physical Resource Productivity and Environmental Counterfactuals

Manufacturing- and service-level physical resource productivity can be quantified only when the output and physical input refer to the same route, product stream, accounting basis, and system boundary. For a named physical resource input $i$, Equations (7) and (8) define output per unit of attributable input:

$$R{P}_{\mathrm{m}\mathrm{f}\mathrm{g},i,p}={\displaystyle \frac{Q_{\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c},p}}{R_{i,p}^{\mathrm{a}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}}}}$$

(7)

$$R{P}_{\mathrm{s}\mathrm{v}\mathrm{c},i,p,s}={\displaystyle \frac{S_{\mathrm{e}\mathrm{q},p,s}}{R_{i,p}^{\mathrm{a}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}}}}$$

(8)

where $R_{i,p}^{\mathrm{a}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}}$ denotes physical input $i$ assigned to product stream $p$ over the same route and system boundary as the numerator; for a single-output route, $R_{i,p}^{\mathrm{a}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}}=R_i$. A product-specific denominator is admissible only when input assignment follows a reported and justified subdivision or allocation rule. Under system expansion, the complete product basket and expanded functions must be retained rather than converted into unsupported product-specific ratios. If shared inputs cannot be defensibly assigned, the complete product basket and shared input profile should be reported, and ratios sharing an unallocated denominator should not be summed. A net input may be used only when the source explicitly defines its components and the corresponding netting procedure. The numerator and denominator must also refer to the same temporal basis, and the attributable input must be greater than zero. A zero or negative net input should be reported as a balance rather than converted to a productivity ratio.

Biomass input, freshwater input or explicitly defined net water use, fresh-solvent makeup, process energy, and attributable land are reported as separate input-specific profiles rather than collapsed into a composite score. A higher $RP$ value therefore indicates more specification-conforming output—or benchmark-equivalent service—per unit of the named input within the stated boundary; it neither compares unlike resource categories nor establishes environmental improvement. Gross solvent circulation is not equivalent to fresh-solvent makeup, and land may be attributed to a side stream only when the upstream attribution rule is explicit and defensible. Cost, labor, and capital are evaluated within techno-economic assessment and are not components of these physical resource-productivity ratios. Because an output–input ratio does not represent impact pathways or an environmental counterfactual, physical resource productivity cannot by itself establish environmental improvement. For impact category $h$, comparative environmental direction is evaluated using Equation (9):

$$\Delta I_h=I_{\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{e}\mathrm{d},h}-I_{\mathrm{c}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{f}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l},h}$$

(9)

where $I_{\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{e}\mathrm{d},h}$ and $I_{\mathrm{c}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{f}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{u}\mathrm{a}\mathrm{l},h}$ are the modeled impacts of functionally equivalent systems expressed using the same functional unit. For Equation (9), the compared systems must deliver the same quantified function and use a common functional unit, corresponding reference flows, and a compatible life-cycle impact-assessment method and version. To preserve functional equivalence, the proposed system includes the valorization route and any replacement required for a beneficial function foregone through feedstock diversion; the counterfactual includes the baseline feedstock fate or use and the incumbent system providing the matched service. Alternative system-expansion or avoided-burden conventions are admissible only when the represented functions, system boundaries, and sign conventions are internally consistent and no credit is counted more than once. A negative $\Delta I_h$ indicates a lower modeled point estimate for the proposed system only for the named impact category, functional unit, geographic and temporal context, system boundary, allocation or system-expansion rule, and assumptions. It must not be generalized to overall environmental superiority, and the category-specific direction should be described as robust only when its sign remains unchanged across reported uncertainty or sensitivity analyses. ISO 14040 and ISO 14044 provide the principles, framework, requirements, and guidelines for life-cycle assessment, whereas ISO 59020 addresses the measurement and assessment of circularity performance [42,43,44]. Circularity performance is therefore treated in this review as a distinct analytical dimension rather than as a substitute for category-specific environmental assessment. When geography, temporal reference, LCIA method/version, or claim-relevant uncertainty evidence was missing, partial, or incompatible, the environmental conclusion was respectively bounded, capped, or stopped at a category-specific source-reported result; missing uncertainty evidence precluded describing the direction as robust.

3.4. Claim-Specific Ceilings and Stopping Rules

Specification-adjusted route accounting assigns a separate ceiling to technical realization, physical resource productivity, economic plausibility, environmental improvement, modeled substitution potential, and observed displacement (Table 1; Figure 2). A defined route and traceable technical basis are shared prerequisites; modeled TEA or LCA remains admissible as model-conditional evidence when realization is not claimed. Product or service equivalence is required only for comparisons that depend on matched function or use. Physical, economic, and environmental paths then proceed separately, with TEA and LCA remaining parallel. Modeled substitution requires comparator, service-equivalence, supply, demand, and realization assumptions. Observed displacement is independent, requires empirical named-incumbent reduction within a defined relationship, and does not require a prior substitution model.

An unqualified claim required every load-bearing field to be explicit and each required linkage aligned or explicitly reconciled. P permitted only a bounded or conditional conclusion; NR, I, or explicit failure of a required criterion capped the path at the last supported point. A deficit unique to one branch did not lower another branch, and ceilings were not averaged into a composite score. Technical realization could therefore be supported without physical productivity, economic plausibility without environmental improvement, or modeled environmental direction without observed displacement.

4. Cross-Stream Synthesis and Route-Level Applications

4.1. Cross-Stream Synthesis: Recurring Dependencies and Contested Boundaries

Three cross-stream findings emerged. First, recoverable output is a route property rather than an extraction endpoint. Feedstock identity, genuine availability, incoming-feed acceptance, stage selectivity, downstream separation, formulation, and storage jointly determine the mass and state that can support a product claim [2,6,45]. A favorable gross yield or marker response can therefore remain valid as a process result while being insufficient for final-product conformity. The recurring boundary is between characterization and predeclared release: a measured endpoint becomes specification-conforming output only when the final stream, product basis, acceptance rule, and conformity evidence refer to the same configuration. Existing uses, burden origin, and baseline fate create a parallel boundary at the feedstock end. A route-accessible side stream is not automatically burden-free or economically uncommitted, because diversion may replace another beneficial function or change residual management. The synthesis therefore begins with generated, genuinely available, and accepted feed, then retains each downstream output state rather than treating all recovered mass as interchangeable. This ordering also prevents moisture correction, concentration, purification, carrier addition, or rejection from silently changing the numerator. It makes clear whether a reported percentage concerns target retention, isolated product, accepted product, or delivered service and whether the corresponding input and baseline remain attached to that state.

Second, analytical maturity within one evidence stream does not repair linkage gaps between streams. TEA and LCA can be detailed and internally admissible while inheriting a scale, grade, output numerator, utility basis, residual fate, functional unit, or counterfactual that differs from the experimental route. Their conclusions remain useful within the reported model assumptions, but they cannot establish release, benchmark-equivalent service, or another branch’s conclusion merely because the model is sophisticated. The same rule applies in reverse: a supported stage recovery or functional response does not supply missing economic, inventory, market, or displacement evidence. Circularity, retained quality, and value measures also answer distinct questions and may coexist with category-specific environmental trade-offs [17,46,47,48,49]. Thus evidence is cumulative only when load-bearing route and comparison bases are aligned or explicitly reconciled. This prevents laboratory, pilot-scale, modeled scale-up, and operating industrial configurations—or product and service bases—from being joined into a broader conclusion that none of the contributing records independently supports. Reconciliation must be traceable rather than assumed: the original bases, conversion or bridging assumption, and affected claim must be recorded. Otherwise, apparent agreement may reflect unlike denominators or boundaries rather than corroboration, and the broader path stops even when each source remains credible within its own scope.

Third, claim strength depends more on the alignment of numerator, product state, scale, function, comparator, and counterfactual than on whether an isolated result is favorable. Specification-adjusted route accounting converts that finding into explicit fields and claim-specific stopping rules. Availability and baseline-fate fields define the resource; product archetype, acceptance specification, and conformity distinguish characterization from release; and compatibility records show whether experimental, economic, environmental, substitution, and market layers can support the same route-level claim. Physical resource productivity, economic plausibility, and environmental improvement then proceed independently after their shared route prerequisites. Benchmark-equivalent service is required only when a comparison depends on matched function or use, and it precedes modeled substitution potential; observed displacement remains an independent empirical path [16]. A partial field can support a bounded conclusion, whereas a missing or incompatible load-bearing link caps the affected path without lowering a separate supported branch. This structure preserves narrow technical or model-conditional findings while identifying the next evidence needed for a broader claim. It also distinguishes a genuine evidence gap from a field outside applicability and avoids converting missing links into a composite sustainability or quality score. A supported finding is therefore neither upgraded by favorable evidence in another branch nor downgraded merely because a different claim is inapplicable. The decision record remains claim specific and exposes which new measurement, comparator, or reconciliation would change the ceiling. These distinctions govern every subsequent cross-stream and package-level audit decision. Table 2 summarizes the three findings and their contested boundaries; Supplementary Workbook S1 preserves the corresponding package-, configuration-, evidence-, and claim-path records.

4.2. Analytical Case Set and Unit of Inference

The route-level audit retained 17 author-curated evidence packages—11 core packages and six boundary-condition packages—and decomposed them into 45 nested analytical units: 17 parent route configurations, 15 source-reported configuration variants, 12 modeled scenarios, and one paired comparison. Package-level counts were used only for coverage summaries, whereas results from configurations, variants, scenarios, and the paired comparison were interpreted only when the corresponding conditions and outcomes were separable in the source. None of the 45 units was treated as an independent study, replication unit, or denominator for estimating field-wide prevalence. The retained set comprised four molecular-identity, seven composition-constrained, and six function-constrained packages and provided the analytical basis for the four worked decision demonstrations presented in Figure 3.

The author-curated set included agri-food byproduct routes alongside other plant-derived systems and was designed to expose variation in product archetype, feedstock, process completeness, scale, application context, and TEA and LCA coverage rather than to represent the wider field. Beyond the four worked cases, core packages included a Peucedanum reference-substance route; polyphenol and standardized-extract routes involving red-wine pomace, Ginkgo, and industrial extracts; and citrus-pectin processing and application routes [50,51,52,53,54,55,56].

The 45 units varied by technology, solvent-to-feed ratio, scale, source, pulsed-electric-field condition, solvent source, and pretreatment. Scale variation did not constitute a common longitudinal scale-up series. Records at different scales were therefore treated as distinct route configurations unless their load-bearing differences were explicitly reconciled; the case set supports evaluation of cross-scale evidence linkage rather than estimation of scale-up success. Package-level inference was retained when outcomes were not separable; configuration- or scenario-specific inference required distinct reported conditions and outcomes. Supplementary Figure S1 summarizes the architecture, while Supplementary Workbook S1 records identifiers, parent–child links, permissible inference, source values, and calculations.

4.3. From Endpoint Characterization to Specification-Conforming Output

All 17 packages reported a terminal chemical or functional endpoint, but evidence for specification-conforming output was uneven. Characterization did not necessarily include a predeclared specification, demonstrated pass, repeat-batch conformity, or commercial release. The Peucedanum and artemisinin packages provided detailed identity and purification evidence, and the pilot-scale citrus-pectin package provided the clearest predeclared food-grade criteria [52,56,57,58]. Several standardized extracts lacked complete predeclared ranges or repeat-batch rules. A processed measurement was therefore treated as characterization unless assessed against an acceptance rule set before outcome evaluation.

Endpoint characterization was not equated with release, repeatability, or benchmark-equivalent service. Configuration-specific process conclusions required a compatible route and product basis; TEA and LCA additionally required claim-appropriate scale, utility, boundary, inventory, and comparison alignment while remaining separate paths. Comparative service and substitution required grade, comparator, matched service, and use-level evidence. Thus the first unsupported applicable link—not the most favorable endpoint—set the claim ceiling. Cell-level decisions are recorded in Supplementary Workbook S1.

4.4. Worked Demonstrations of Route-Level Decision Changes

Four worked demonstrations show how the conclusion suggested by a favorable focal metric changes when downstream losses, product properties, comparison conditions, and claim-specific evidence requirements are evaluated at the route level (Figure 3; Supplementary Table S3). These demonstrations illustrate claim calibration within selected evidence packages and are not intended as a cross-case ranking. Three of the four demonstrations concern agri-food byproduct routes. The artemisinin example was retained as a molecular-identity contrast because its explicitly reported sequential recoveries permit direct examination of how stage-specific evidence constrains complete-route claims.

In the artemisinin route, conditional recoveries of 90.0% (maceration), 87.1% (flash chromatography), and 47.6% (crystallization) yielded cumulative retention of 90.0%, 78.4%, and 37.3%, respectively [57]. The 78.4% intermediate was author-derived; 37.3% rounded the source-reported 37.27% overall recovery and was independently reproduced. Treating maceration as complete-route retention would overstate final retention by a factor of 2.41. The evidence supports stage recovery and complete-route target retention within the technical boundary, while release remains partial. The broader physical, economic, environmental, and modeled-substitution paths were outside the MI-02 applicability map; observed displacement was undocumented.

For grapefruit pectin, the source reported a 27.34% yield on a mass basis that was not explicit in the accessible record, representing a 16.34% within-source relative increase over conventional heating extraction, while using a temperature 13.3 °C lower and an extraction time 37.78% shorter [59]. Yet intrinsic viscosity and molecular weight were lower, the degree of esterification was slightly but not significantly lower, and not all reported increases in branching and purity were supported by statistically significant contrasts. Without a predeclared release specification or benchmark-matched gelation or use-level tests, the defensible conclusion is a configuration-specific trade-off among extraction yield, operating conditions, and reported product properties; it does not establish benchmark equivalence, substitution, displacement, or general route superiority.

The red-grape-pomace TEA supports a model-specific preference, not general superiority [60]. PEF-02 was preferred below USD 249 kg−1 and SLE-01 at or above USD 250 kg−1, bounding the switching interval. At USD 233 kg−1, PEF-02 reduced unit production cost by 2.72% and increased return on investment by 8.11% relative to SLE-01. The defensible conclusion is a conditional economic preference within the reported route, price range, and model assumptions. The package did not establish environmental improvement, modeled substitution, or observed displacement.

For microencapsulated pomegranate-peel extract powder (PPEP), Case 3 provided the strongest deterministic economic result, whereas Case 2 had the highest reported probability of positive NPV, P(NPV > 0) = 91.9%, in the source-reported 10,000-run Monte Carlo analysis [34]. These outputs belong to one model-conditional economic path but answer different questions. A separate package-level LCA identified freeze-drying as the dominant hotspot. Because it was not attributable to a specific TEA scale and lacked a functionally equivalent counterfactual, it did not establish comparative environmental improvement. Neither branch established benchmark-equivalent service, modeled substitution, or observed displacement.

4.5. Decision Consequences and Claim-Path Audits

Figure 3 and Supplementary Table S3 present the four worked demonstrations as decision changes rather than route rankings. For each case, the focal metric is retained, then interpreted against the route, product state, comparison basis, and intended claim. The resulting ceiling may be a supported complete-route technical result, a process–property trade-off, a price-bounded economic preference, or parallel model-conditional economic and environmental findings. The audit does not reverse a source result; it limits the conclusion to the boundary supported by its numerator, denominator, configuration, and comparator. This distinction is important because the same favorable direction can imply different decisions depending on whether the unresolved link concerns final conformity, a matched physical input, price sensitivity, functional equivalence, or a counterfactual.

Supplementary Figure S2 makes the same logic explicit for MI-02 and CC-02. MI-02 supports conditionally nested recovery but only bounded release, so broader physical, economic, environmental, and substitution paths remain outside its documented applicability and displacement remains undocumented. CC-02 supports route definition and modeled economic cases, but release evidence is bounded. Its deterministic point estimates and Monte Carlo probability-of-positive-NPV results answer different questions within one economic path, while the package-level LCA supports hotspot identification without a functionally equivalent counterfactual or attribution to a specific TEA scale. The records therefore remain parallel and cannot be combined into an integrated superiority, substitution, or displacement claim. Separating the two packages also shows why NA and missing evidence are not equivalent: an inapplicable branch is excluded before coding, whereas a required but unsupported link actively limits the corresponding claim.

The stopping decision is claim specific. P permits a narrower or conditional conclusion only when no stopping condition is present; NR, an incompatible linkage, or explicit failure of a required criterion caps the affected path at its last supported point. NA is excluded only after applicability is determined. Missing values are not imputed, one branch does not fill another branch’s gap, and separate ceilings are not averaged. Observed displacement does not require a prior substitution model, but it does require empirical linkage between recovered-product uptake and reduced procurement, production, or consumption of a named incumbent within a defined relationship and period. No selected package met that requirement at the documented access level. The audit trail links the ten displayed fields and their available evidence records to claim-specific stopping decisions. Because auxiliary requirements A01–A08 were not coded systematically at the package–claim level, the trail supports a structured qualitative explanation of the reported ceilings rather than complete independent reproduction of every stopping decision.

4.6. Boundary-Condition Demonstrations Across Contrasting Evidence Structures

The six boundary-condition packages were assessed with the same availability, provenance, compatibility, and stopping rules as the 11 core packages. Their label denotes an analytical role across contrasting process, scale, and assessment structures; it does not denote an external validation set or support estimates of audit accuracy, reliability, generalizability, or route success. Package-level details remain in Supplementary Table S4.

Molecular-identity cases showed that recovery must retain its numerator, denominator, stage, and scale. Steviol-glycoside and curcuminoid records reported crystallization, residual-stream, complete-procedure, or scale-up recoveries on different bases; those values were not treated as interchangeable measures of route output [61,62]. The defensible conclusion followed the specifically documented stage and product state rather than the largest reported percentage. This prevents a high conditional recovery from being relabeled as complete-route retention.

Composition-constrained cases separated model outputs from release and service evidence. Pine-residue LCA supported a configuration-specific environmental comparison, while spruce-bark tannin TEA and LCA supported conditional economic and category-specific environmental findings in parallel [63,64]. Neither evidence structure, by itself, established routine release, benchmark-equivalent service, modeled substitution, or displacement. Model completeness within one domain did not raise another domain’s ceiling.

Function-constrained cases reached bounded conclusions through different paths. Prospective rhamnogalacturonan-I pectin models supported conditional economic and category-specific environmental results, whereas Kinnow-peel pectin evidence supported bounded product comparability and hotspot identification [65,66]. Across all three archetypes, heterogeneous evidence produced heterogeneous claim ceilings under one rule set. The packages demonstrate analytical applicability, not validation; package-specific findings and load-bearing limitations are retained in Supplementary Table S4. Their value is to expose boundary behavior, not to estimate how often a route succeeds.

4.7. Evidence Availability, Provenance, Compatibility, and Interpretive Limits

Figure 4 aggregates the availability and completeness decisions for 170 package–field cells comprising 17 author-curated evidence packages and ten audit fields. Feedstock identity and route configuration were explicit in all 17 packages. Stage recovery was explicit in three packages and partial or bounded in 14. Acceptance specifications were explicit in one package, partial in 11, and required but not reported in five; evidence of specification pass was explicit in one, partial in nine, and required but not reported in seven. No package contained explicit evidence of benchmark-equivalent service: three packages were coded as partial or bounded, ten as required but not reported, and four as not applicable to the documented package–claim applicability map. TEA was explicit in five packages, required but not reported in five, and not applicable in seven. LCA was explicit in 11 packages, required but not reported in one, and not applicable in five. Modeled substitution potential was partial or bounded in two packages, required but not reported in 11, and not applicable in four. Observed displacement was surveyed across all 17 packages as a cross-cutting question rather than treated as a uniformly applicable source-reporting requirement. No selected package provided verifiable observed-displacement evidence at the documented access level: 16 packages were coded NR, whereas MI-01 was coded P because the accessible abstract and metadata were insufficient to support a complete nonreporting judgment. These counts should not be interpreted as a prevalence estimate of missing reporting.

Figure 4. Evidence availability and completeness across 17 author-curated packages. E, explicit; P, partial or bounded; NR, required but not reported; NA, not applicable. Each E/P/NR stack uses the field-specific applicable denominator (17 − NA), with NA shown separately. TEA and LCA remain parallel domains. Counts describe the selected set, not field-wide prevalence, framework validation, or sustainability performance. Observed displacement was surveyed as a cross-cutting question across all 17 packages. No selected package provided verifiable evidence at the documented access level: 16 were NR and MI-01 was P because limited access precluded a definitive nonreporting judgment; these counts are not a field-wide prevalence estimate.

Figure 4. Evidence availability and completeness across 17 author-curated packages. E, explicit; P, partial or bounded; NR, required but not reported; NA, not applicable. Each E/P/NR stack uses the field-specific applicable denominator (17 − NA), with NA shown separately. TEA and LCA remain parallel domains. Counts describe the selected set, not field-wide prevalence, framework validation, or sustainability performance. Observed displacement was surveyed as a cross-cutting question across all 17 packages. No selected package provided verifiable evidence at the documented access level: 16 were NR and MI-01 was P because limited access precluded a definitive nonreporting judgment; these counts are not a field-wide prevalence estimate.

The ten-field matrix is a coverage summary, not the complete claim logic. The Claim Path Stop Audit summarizes matched inputs, multifunctionality, uncertainty, and counterfactual considerations in its rationale fields, but auxiliary requirements A01–A08 were not independently coded at the package–claim level across all packages. Accordingly, these records are a structured qualitative audit aid, not a complete risk-of-bias, certainty, or fully reproducible claim-decision matrix.

Route identity was more complete than release conformity, benchmark-equivalent service, or displacement in this selected set, but the counts are not field-wide prevalence estimates. An E code for one field does not complete a path: the first unsupported or incompatible required link still controls the ceiling. E/P/NR/NA describe availability and completeness, not evidence quality, risk of bias, route performance, or sustainability.

Figure 4 reports availability only. Supplementary Figure S3 and Supplementary Workbook S1 separately report provenance and linkage compatibility. Provenance does not imply completeness; compatibility matters only when records are linked for one claim. A or R permits evaluation within its documented boundary, I caps the path, NL means an evidence record without invoked linkage, and blank means no eligible evidence record. These matrices document the selected cases, not audit accuracy or external validity.

4.8. Incremental Contribution Relative to Adjacent Assessment Methods

Adjacent methods answer complementary questions. Circularity indicators describe circulation and resource states; quality and value frameworks describe context-dependent product attributes; TEA and LCA model economic and category-specific environmental outcomes; and substitution frameworks distinguish equivalence, potential replacement, and empirical displacement [16,17,46,48,67]. Each method can be internally rigorous without establishing that its feedstock, configuration, product, scale, or counterfactual matches another assessment layer. Specification-adjusted route accounting therefore operates between these methods as an evidence-linkage layer rather than replacing their calculations or quality criteria. It asks whether the outputs can support the same claim and, if not, where the evidentiary handoff fails.

The present framework combines a common route key, explicit separation of output states and claim types, and claim-specific stopping rules in one review-specific audit architecture. Its intended contribution is the integration of these elements as an evidence-linkage layer across experimental, product-specification, TEA, LCA, substitution, and market records. This integration is proposed for independent evaluation; the selected case set does not establish uniqueness, superiority, reliability, or decision benefit relative to adjacent methods.

Applied to the worked and boundary-condition cases, the framework preserved source-reported or transparently derived results while narrowing only the implication that could be defended. That distinction matters for research design and decision use. Investigators can predeclare the intended claim and collect the next load-bearing record; modelers can state where route or comparison bases remain conditional; and decision-makers can distinguish a supported technical result from economic, environmental, substitution, or market evidence. The audit does not rank technologies, certify sustainability, validate the framework, or estimate field-wide prevalence. Its value is the structured relationship between evidence state and conclusion. Supplementary Workbook S1 records route keys, evidence records, compatibility decisions, transformations, and stopping-point rationales; because auxiliary requirements A01–A08 were not coded systematically at the package–claim level, the workbook supports inspection of that relationship rather than complete independent reproduction of every ceiling. Independent prospective testing is still required to determine whether this structure improves consistency across reviewers and new route families.

5. Discussion

5.1. From Cross-Stream Synthesis to Defensible Route-Level Claims

For agri-food byproduct routes and other plant-derived systems, the main interpretive problem in laboratory-to-industrial translation is not a shortage of favorable metrics but a mismatch between individual metrics and the claims they are used to support. Extraction yield, characterization, TEA, LCA, substitution modeling, and market evidence describe different stages or domains. They support a common route-level conclusion only when the load-bearing feedstock, process, product, scale, and comparison fields are aligned or explicitly reconciled. In this critical integrative review, route accounting does not strengthen or pool source results; it identifies the strongest claim supported by the documented route-linked evidence and the first load-bearing link that remains unresolved.

Across the 17 packages and 45 nested units, six distinct claim paths with separate ceilings were preserved. After their shared route-definition prerequisites were met, economic and environmental claims proceeded in parallel, and neither supplied evidence for the other. Partial evidence permitted only bounded or conditional conclusions, whereas an unsupported or incompatible required link capped the affected path at its last supported point. The resulting conclusions illustrate decision consequences within the selected set; they do not estimate field-wide prevalence, validate the audit, or rank technologies.

5.2. Product Archetypes and Claim-Specific Evidence Requirements

The three product archetypes operationalize context-dependent product quality without reducing it to a single score. They function as claim-oriented specification anchors rather than universal quality rankings or regulatory classifications. Circular-economy literature treats quality as multidimensional, potentially encompassing technical properties, functionality, durability, market value, and environmental attributes; no single metric captures all of these dimensions [46,49]. Accordingly, the relevant evidence depends on the attribute that defines the claimed product and service.

For a molecular-identity product, a claim of specification-conforming output ordinarily requires evidence of identity, purity or assay, and claim-relevant impurity limits. For a composition-constrained product, the load-bearing specification comprises predeclared compositional ranges or fingerprints, together with evidence that the final product conforms to those requirements. For a function-constrained product, product conformity depends on predeclared functional acceptance criteria assessed under defined test conditions. Benchmark-equivalent service is a separate comparative requirement: it additionally requires a named comparator, matched test conditions, a common service basis, and an explicit use level. Functional characterization alone therefore does not establish either routine product release or service equivalence.

Evidence requirements should follow the intended claim rather than a universal maturity ladder. A characterized endpoint supports technical realization only for the reported stage and configuration. Specification-conforming output requires a defined final product stream, a predeclared acceptance specification, and demonstrated conformity with that specification. Physical resource productivity additionally requires a route-aligned output numerator and an attributable physical-input denominator expressed on the same accounting basis and within the same system boundary. Economic plausibility and environmental improvement require route-aligned, claim-appropriate TEA and LCA evidence, respectively, and remain parallel rather than interchangeable. Modeled substitution potential further requires benchmark-equivalent service, an explicit supply and demand context, and stated realization assumptions. Observed displacement is evaluated independently and requires empirical evidence that uptake of the recovered product reduced procurement, production, or consumption of a named incumbent within a defined market, procurement, or supply relationship; a prior substitution model is not required. This differentiation is consistent with substitution frameworks that distinguish physical resource potential and recovery efficiency from relative functionality and the expected change in competing-product consumption [16].

5.3. Distinguishing Physical Resource Productivity, Circularity Performance, and Environmental Improvement

Physical resource productivity pairs one named output with one physical input under a common basis and boundary. The numerator may be specification-conforming output or benchmark-equivalent service; the denominator may be accepted feed, net water, fresh-solvent makeup, energy, attributable land, or another explicit input. These ratios answer different questions and should remain separate. Better output per accepted feed can coexist with higher solvent, water, or energy use, just as favorable economics can coexist with a drying hotspot. No single ratio establishes total resource reduction, economic or environmental superiority, substitution, or displacement [16].

Circularity performance and environmental improvement likewise answer different questions. ISO 59020:2024 specifies requirements and guidance for measuring and assessing circularity performance within a defined economic system and is intended to be used with complementary methods when social, environmental, or economic impacts are assessed. ISO 14040:2006 and ISO 14044:2006 instead provide the principles and framework, and the requirements and guidelines, respectively, for life-cycle assessment. High circulation, recycling, or retained value therefore cannot alone establish lower environmental impacts. Environmental direction remains specific to the impact category, functional unit, system boundary, comparator, geographic and temporal context, and modeling assumptions, and improvement in one category may coexist with deterioration in another [19,42,43,44].

Within this review, an unqualified, impact-category-specific environmental-improvement claim required functionally equivalent systems and explicit load-bearing fields that were aligned or explicitly reconciled: the assessed route configuration and scale; functional unit and reference flow; system boundary; comparator and counterfactual; geographic and temporal context; LCIA method and version; evaluated impact category; and, where applicable, baseline feedstock fate, displaced function, and treatment of multifunctionality. When substitution or system expansion was modeled, the substituted product, substitution ratio, and evidential basis for that ratio were additionally required. Partial evidence permitted only a bounded or source-scenario-specific interpretation, whereas a required but unreported or incompatible field capped the claim at the last supported result. A direction was described as robust only when claim-relevant uncertainty or sensitivity analyses were reported and its sign remained unchanged across them. For multi-output routes, subdivision was preferred when operations and inventories were separable; otherwise, the source-reported system-expansion or allocation rule was retained and made explicit. To avoid double counting within each product-system comparison, shared inventory flows and burdens were partitioned among outputs or retained within the complete product basket in accordance with the declared treatment of multifunctionality; neither an avoided baseline fate nor a displaced function was credited more than once. Results from mutually exclusive scenarios were not summed or treated as independent evidence, even when the scenarios shared a common baseline assumption [11,16].

5.4. Implications for Study Design, Cross-Assessment Linkage, and Laboratory-to-Industrial Translation

Laboratory-to-industrial translation is treated here as an evidence-linkage problem rather than as an automatic extension of laboratory extraction performance. Specification-adjusted route accounting is proposed as a prospective aid for designing linked experiments and assessments and as a retrospective aid for auditing evidence claimed to support transitions across scales and assessment domains. At study outset, investigators should assign stable identifiers to the feedstock, route configuration, product state and grade, and comparison basis, and should predeclare the intended unit of inference and claim class. Claim-proportionate experimental records should distinguish generated, route-accessible, and accepted feed; state the applicable mass basis and feedstock condition; and preserve the complete unit-operation sequence. For target-defined routes, records should report initial target content and conditional and cumulative stage recoveries. Across route types, records should also report, where applicable, final isolated or formulated product mass, predeclared acceptance specifications, specification-pass decisions, disposition of nonconforming material, physical inputs, co-products, and residual fates. These records separate gross recovery from specification-conforming output and prevent later models from implicitly supplying missing experimental links.

Downstream assessments should inherit the same route identifiers, output numerator, product grade, and unit of inference or explicitly reconcile any differences. TEA should report the reference production basis, scale and throughput, capacity utilization, route-aligned mass and energy balances, quality-control, rejection and reprocessing assumptions, capital and operating cost boundaries and estimates, co-product treatment, currency and base year for costs and prices, grade-matched price basis, the economic decision metric and any claimed acceptance criterion, and claim-relevant uncertainty, sensitivity, or switching conditions. LCA should report the functional unit and reference flow, proposed and counterfactual systems, baseline feedstock fate and displaced function where applicable, allocation or system-expansion basis, route-aligned inventory, geographic and temporal context, LCIA method and version, category-specific results, and claim-relevant uncertainty or sensitivity analysis. TEA and LCA remain parallel claim paths; evidence for economic plausibility does not establish environmental improvement, and the converse also holds. When technical, economic, and environmental analyses are intentionally integrated as an environmental techno-economic assessment, ISO/TS 14076:2025 provides principles, requirements, and guidance for performing the assessment, specifies requirements for documenting its results, underlying assumptions, parameters, and methodologies, and applies to process systems of any size or production scale [68].

Modeled substitution additionally requires a named comparator and grade, a specified service, matched test and use conditions, a service-equivalence ratio, available benchmark-equivalent supply, compatible demand, and a stated realization scenario applied after the supply and demand caps. Observed displacement is a separate empirical path, not an inference from the substitution model, and requires procurement, production, contract, or sales records linking recovered-product uptake to a reduction in use of the named incumbent within a defined period and market, procurement, or supply relationship. Market size, recovered-product sales alone, modeled credits, and purchase intentions are insufficient. Each transition from laboratory measurements to pilot-scale observations, modeled scale-up, or operating industrial evidence should be treated as a change in route configuration or as an explicitly reconciled variant, rather than as an implicit extension of a laboratory route. Differences in feed concentration, equipment, unit-operation sequence, recycle, utilities, product state or grade, specification, functional unit, residual fate, or comparator require explicit alignment, modeling, or reconciliation [14,15].

5.5. Limitations, Appropriate Use, and Priorities for Independent Evaluation

The author-curated 17-package set supported cross-stream synthesis and examination of contrasting evidence structures, but it was not drawn from an exhaustive screening universe and did not constitute a representative sample. Package counts therefore do not estimate field-wide reporting prevalence, the probability of route success, or audit performance. The 45 units were nested within packages rather than representing independent studies. Differences in product archetype, maturity, scale, access, and purpose preclude direct ranking. The case set did not constitute a longitudinal validation cohort in which the same routes were followed from laboratory through pilot to operating industrial scale. Accordingly, this review evaluates the evidentiary conditions and linkage requirements for laboratory-to-industrial claims; it does not estimate the probability of successful scale-up or validate industrial performance.

One author conducted selection, extraction, classification, appraisal, compatibility assessment, and reconciliation. A structured second-pass internal-consistency check did not constitute duplicate screening, independent extraction, inter-rater calibration, or external validation. For access-limited sources, coding was restricted to claim elements explicit in official records available at the documented access level, which may still introduce selection and interpretive bias. Boundary-condition packages illustrated rule operation and did not constitute an independent test set. Independent reclassification and prospective multi-assessor evaluation are therefore required before reproducibility or transferability can be inferred.

Importantly, the audit evaluates claim-relevant reporting completeness and linkage compatibility rather than study-level risk of bias or certainty of evidence. Support at a given claim ceiling therefore indicates only that the load-bearing fields applicable to that claim were reported and compatible; it does not establish that the underlying effect estimates are unbiased, precise, or independently replicated.

Specification-adjusted route accounting is an evidence-linkage, claim-calibration, and study-design aid—not a consensus standard, certification instrument, validated guideline, predictive model, or composite score. Evaluation priorities include independent duplicate classification, inter-rater agreement, independently selected cases, prospective predeclaration, and sensitivity of claim ceilings to coding and reconciliation. A formal reporting guideline would also require stakeholder participation and tests of clarity, feasibility, decision impact, and burden.

Practically, researchers can assign route keys and intended claims before data collection; industry can treat the first unsupported link as the next scale-up, quality, TEA, LCA, or market requirement; and public decision-makers can separate documented outputs from unverified environmental or displacement claims. Priorities are complete route/specification records, matched denominators and application tests, uncertainty-aware TEA/LCA consistent with established assessment guidance [15,69], prospective multi-assessor evaluation, and empirical displacement records—not forecasts of adoption or policy benefit.

6. Claim-Specific Reporting Implications

Supplementary Table S2 presents a proposed 42-item claim-specific checklist spanning feedstock, product specification, manufacturing, scale, TEA, LCA, market, substitution, displacement, and multi-output counterfactuals. The checklist is intended to preserve the traceability of load-bearing route elements before a claim is advanced, drawing on related substitution and circular-quality frameworks [16,49]. For cross-scale claims, laboratory, pilot-scale, modeled scale-up, and operating industrial configurations are treated as distinct. Applicability and evidence-status coding are applied separately to each configuration and to the linkage between them; an item coded E at one scale is not thereby E at another unless load-bearing differences in feedstock, process sequence, product basis, utilities, residual fate, temporal basis, and system boundary are aligned or explicitly reconciled. Applicability otherwise depends on the product archetype and evaluated claim; the applicable items constitute a candidate minimum reporting set rather than a consensus standard.

Applicability is determined before evidence status is coded. Availability and completeness are recorded as E (explicit), P (partial or bounded), NR (required for the evaluated claim path but not reported), or NA (not applicable). Provenance and derivation status are recorded as SR (source-reported), AD (author-derived), MS (source-reported modeled scenario), or SR/AD (source-reported inputs combined with an author-derived value). MS is used only for a value or statement directly reported from a source-defined modeled scenario. SR/AD denotes a value calculated from traceable source-reported inputs, whereas AD denotes an author-constructed scenario, classification, or other derivation that introduces author-specified assumptions. Linkage compatibility is recorded as A (aligned), R (explicitly reconciled), I (incompatible), or NL (no cross-source or cross-basis linkage invoked). A, R, or I is assigned only when evidence records or assessment layers are linked in support of the same route-level claim. NL is assigned when an evidence record exists but no cross-source or cross-basis linkage is invoked, whereas a blank compatibility cell is reserved for the absence of an evidence record to which compatibility could meaningfully be assigned. Provenance identifies the origin or derivation status of a value or statement but does not grade its methodological quality, and an E code for one field does not establish that all load-bearing requirements for a claim have been met.

After applicability and shared route-definition prerequisites have been assessed, the six claim paths retain separate ceilings rather than forming a single cumulative hierarchy. Economic plausibility and environmental improvement remain parallel paths. Modeled substitution potential additionally requires benchmark-equivalent service and explicit supply, demand, and realization assumptions, whereas observed displacement is a separate empirical path that does not require a prior substitution model; product conformity or service equivalence is required for that path only when it is load-bearing for the defined product–displacement claim. A P code permits only a bounded or conditional conclusion when no stopping condition is present. An NR or I code, or explicit failure of a claim-required criterion, caps the affected path at its last supported point. NA is assigned only after an item has been determined not to apply to the documented product archetype, route configuration, or evaluated claim path. Reconciliation permits use only within its recorded boundary and cannot create missing evidence. Evidence from TEA, LCA, substitution modeling, market records, or adjacent domains contributes only to the claim path for which it is applicable and cannot replace any missing prerequisite required for that path. No load-bearing value is imputed, and evidence on one branch does not raise the ceiling of another.

The checklist is an audit and study-design aid, not a consensus standard, certification tool, validated guideline, risk-of-bias instrument, or sustainability score. Completion of applicable items does not itself establish methodological validity, product conformity, successful laboratory-to-industrial translation, superiority, environmental improvement, substitution, displacement, or generalizability. Developing a formal reporting guideline would require a separate process involving stakeholder consensus and prospective evaluation.

7. Conclusions

For agri-food byproduct routes and other plant-derived systems, gross extraction yield is an intermediate process measure and cannot by itself establish specification-conforming output, benchmark-equivalent service, physical resource productivity, economic plausibility, or environmental improvement. Laboratory-to-industrial translation should therefore be treated as an evidence-linkage problem across scale-specific route configurations rather than as an assumed progression from a favorable laboratory result. Defensible route-level interpretation requires the applicable load-bearing fields—including feedstock identity, condition, availability, and burden origin; process sequence, configuration, and scale; product state and grade; utilities and residual fate; and, where required by the evaluated comparison, the comparator, functional unit, and counterfactual—to be aligned or explicitly reconciled. Progression from recovery to specification-conforming output requires a defined final product stream and mass, predeclared acceptance criteria, and demonstrated conformity; benchmark-equivalent service additionally requires a named comparator, a defined service, matched test and use conditions, and an explicit use level. Modeled substitution potential further requires explicit supply, demand, and post-cap realization assumptions, whereas observed displacement requires empirical evidence linking uptake of the recovered product to reduced procurement, production, or consumption of a named incumbent within a defined market or supply relationship and period. Partial evidence permits only a bounded or conditional conclusion, whereas an unsupported or incompatible load-bearing link—or explicit failure of a required criterion—caps the affected claim path at its last supported point. Physical resource productivity, economic plausibility, and environmental improvement retain separate claim ceilings; evidence on one branch does not establish another, and circularity alone does not demonstrate environmental improvement.

Applying qualitative evidence-status and stopping rules to 17 author-curated evidence packages containing 45 nested analytical units did not alter source-reported or transparently derived results. Of Equations (1)–(9), only the conditionally nested recovery relation was instantiated numerically in the retrospective cases; the remaining equations served as proposed accounting definitions and claim-eligibility boundaries. The worked demonstrations illustrated that technical or model-conditional findings remain informative within their documented boundaries even when broader claims concerning product conformity or release, physical resource productivity, environmental improvement, substitution, or displacement are bounded or capped. Observed displacement was surveyed as a cross-cutting question, and no selected package provided verifiable observed-displacement evidence at the documented access level. The nested, nonrepresentative case set was not a longitudinal scale-up cohort; it therefore does not estimate field-wide prevalence, route-success probability, or scale-up success, and it does not validate the audit or establish a universal hierarchy. The framework and its proposed 42-item checklist are evidence-linkage, claim-calibration, and study-design aids, not composite scores, consensus reporting standards, certification schemes, risk-of-bias instruments, predictive models, or validated guidelines. Their practical contribution is to identify the strongest reporting- and linkage-compatible conclusion and the first unresolved load-bearing evidence requirement without treating an evidence gap as evidence of poor route performance. Independent reclassification, inter-rater assessment, prospective application, and evaluation in broader case sets are required before reproducibility, transferability, or decision benefit can be inferred. Future studies should predeclare route keys, product states, intended claims, and claim-specific stopping criteria and test whether classifications and reconciliations are reproducible across assessors and new route families and whether the proposed minimum reporting records materially improve decision consistency.