What does a bundle of measures really achieve? Causal identifiability of heating optimisation and heat recovery in an anonymised nonresidential building analysis

The label given to a renovation package often suggests a single technical cause for its energy saving. When the package changes several systems at once, however, this attribution is not supported by the package comparison. This paper examines an anonymised, 69-page energy consultancy report using structured intervention coding and a formal identifiability test. In the transition studied, hydraulic balancing, heating-circuit temperatures, control and heat recovery are changed together. The documented final energy demand falls by 31.7 per cent relative to the immediately preceding envelope state. This contrast is observable within the report; the individual effects of the heating and ventilation measures are not. A saturated two-factor model has four parameters, while only two distinct treatment combinations are available. The missing combinations are not replaced by assumed saving shares. Instead, a complete simulation design is formulated that would make main effects and interaction identifiable under constant boundary conditions. A supplementary mechanism check shows that the reported reduction in ventilation losses of about 87 per cent must not be equated with the stated heat recovery efficiency of 80 per cent. The result is a verifiable demonstration of limited attributability, not a new estimate of real savings. The paper links the diagnosis to a measurement and documentation plan for later technical verification.