National dairy sustainability monitoring requires indicators that are scientifically defensible, policy-relevant, and feasible to construct from recurrent agricultural surveys. The aim of this study was to construct, interpret, and evaluate a national, survey-weighted, frame-aware, MESMIS-informed monitoring profile for Peruvian dairy cattle producers using ENA 2021–2024 microdata. The Peruvian National Agricultural Survey (Encuesta Nacional Agropecuaria, ENA) was explicitly treated as a source of proxy-based sustainability monitoring rather than as a complete farm sustainability assessment. Using survey design variables and harmonized items on livestock, markets, producer capacity, and resource management, the analysis constructed survey-weighted formative profiles for nationally monitored dairy cattle producer units. The analytical sample comprised 42,331 agricultural producer units selected based on complete interview status, reported livestock activity, and harmonized evidence of dairy cattle activity. Because ENA documentation indicates a sampling-frame transition in 2023, the 2021–2022 and 2023–2024 rounds were treated as separate sampling-frame regimes. Indicators were scaled to 0–1 formative profiles, with explicit checks for missingness, indicator availability, exclusion of derivative outputs, exclusion of producer age, minimum-availability thresholds, and sampling-regime-specific scaling. The median indicator availability was 12 of 14 economic and productive indicators, five of five social access and capacity indicators, and eight of eight environmental and resource-management indicators. Survey-weighted descriptive estimates and adjusted within-regime contrasts showed that the strongest signal was a lower social access and capacity profile in 2022 than in 2021, mainly associated with lower agricultural information use, livestock training, and technical assistance. Within the 2023–2024 sampling-frame regime, the main adjusted profiles were statistically stable. The results show that ENA can support a reproducible national monitoring baseline, provided that the resulting profiles are interpreted as proxy-based formative summaries constrained by survey coverage rather than as direct environmental performance scores, causal trends, or complete sustainability measures.

Subsistence agriculture persists across Peru with markedly heterogeneous patterns among ecological regions, yet multi-regional analyses using predictive methods remain scarce. This study identified determinants of household subsistence using random forest and gradient boosting classifiers. Microdata from the 2024 Peruvian National Agricultural Survey yielded 34,827 agricultural units after quality control. Subsistence was operationalised as the share of self-consumption in total production value and categorised as low (0% - 20%), medium (20% - 60%) and high (> 60%). On a held-out test set of 10,448 units both algorithms converged (random forest 82.9%, gradient boosting 82.6% accuracy), yet a nested sensitivity analysis showed that this performance depends substantially on predictors that enter the definition of the dependent variable. Once income-derived and market-participation variables were removed, accuracy settled at 65.7% against a 57.9% majority-class baseline, and harvested area and input expenditure emerged as the leading structural determinants. Three findings run against prevailing narratives. First, productive richness does not accompany commercialisation: Andean units combined the highest richness (7.9 versus 5.8 species on the Coast) with the highest subsistence (39.1% versus 15.0%). Second, direct-to-consumer selling was associated with higher subsistence (34.2%) than collector intermediation (14.3%), consistent with intermediaries operating as a pragmatic commercialisation route rather than a barrier. Third, accuracy was lowest precisely in the Andes (77.0%), where subsistence concentrates, and weakest for the transitional medium class (60.7%). Reporting predictive performance without separating definitional from structural predictors overstates what such models explain, a distinction rarely made in this literature.