Residential biomass combustion can substantially contribute to local air pollution, with gaseous and particulate emissions responding nonlinearly to operating strategy and combustion phase. We report a controlled experimental campaign performed on plate-steel stoves designed for seasoned hardwood combustion (12 kWth nominal). Two stove configurations were investigated, differing in combustion-air distribution; one configuration was additionally operated with and without an accumulation layer. Stove operation was governed by a dedicated control algorithm enabling reproducible test conditions and systematic variation of combustion regimes. The experimental rig provided continuous measurements of flue-gas outlet temperature and gas-phase concentrations of CO, CO₂, O₂, NOx and SO₂. Particulate matter (PM) was quantified gravimetrically and characterised chemically using filter samples collected over defined firing phases. Combustion performance was diversified by adjusting airflow and fuel feed rate to obtain three operating conditions. Each experiment combusted 15 kg of logwood or briquettes in five batches spanning ignition (cold start), successive combustion steps, and afterburning. We developed a Python-based data pipeline to harmonise time-resolved gas data with phase-resolved PM and composition results and to enable systematic screening of dependencies across fuels, stove configurations, and operating regimes. Exploratory analysis includes phase-aware comparisons and correlation-based mapping between operational descriptors, meteorological covariates, gaseous metrics and PM outcomes. The resulting curated dataset and dependency analysis provide a foundation for ongoing surrogate modelling to support rapid scenario screening and subsequent climate–health trade-off assessment.