**Abstract.** Blue-green infrastructure (BGI) has been widely recognized as a multifunctional solution for cities since it can simultaneously contribute to stormwater management, heat mitigation, and biodiversity conservation. However, deciding on BGI elements, such as green roofs or ponds, which perform different functions, is complex. Elements are interconnected, can be combined, their performance is uncertain, and societal preferences about the desired function also vary. This study tackles the complexity and uncertainty of assessing BGI combinations with respect to 13 technical, social, environmental, and economic objectives using participatory multicriteria decision analysis (MCDA). We assessed 61 combinations of green roofs, bioretention cells, detention ponds, and pervious pavement at different implementation rates (10%, 25%, 50% and 100%) in a municipality near Zurich, Switzerland. Using multi-attribute utility theory (MAUT) and the concept of expected expected utility (EEU), we integrated probabilistic performance predictions from domain-specific models with expert judgement and elicited preferences of six local stakeholders. EEU provides a rationale to select options when predicted performance and preference data are probabilistic. Overall, bioretention cells combined with detention ponds performed well in face of heterogeneous stakeholder preferences, indicating potential consensus for our case. Lower spatial coverage performed better than higher coverage because fewer associated drawbacks, e.g., costs and potential for invasive species, outweighed marginal performance gains of higher coverage. We contribute a non-fully compensatory MAUT framework that propagates uncertainty in both performance predictions and elicited stakeholder preferences, enabling robust recommendations for planners and decision-makers. **Contents of this package.** The package contains the data and code required to reproduce the multicriteria decision analysis reported in the associated publication, including all figures and tables in the paper and its supplementary material. It comprises: (i) predicted performances of all 61 BGI options on the 13 attributes, derived from domain-specific models (EPA SWMM for combined sewer overflow and micropollutant discharge, Urban Tethys-Chloris for pedestrian-level thermal comfort, EnergyPlus for building energy demand, species accumulation and generalized additive models for biodiversity, a life-cycle-assessment meta-analysis for greenhouse gas emissions, and an equivalent-annual-cost model for life cycle costs), together with the probability distributions fitted to represent their uncertainty; (ii) the elicited preference data of six stakeholder representatives — feasible objective weight sets, the aggregation (compensation) parameter of the weighted power mean model, and risk attitude parameters of the single-attribute utility functions — in anonymized form; (iii) the R code implementing the sampling-based expected expected utility decision model, adapted from Haag et al. (2019, *Journal of Environmental Management* 252, 109652), covering value function construction, Monte Carlo sampling of performances and preference parameters, hierarchical aggregation, and ranking analysis; (iv) a stepwise sensitivity analysis carried out ahead of the stakeholder workshop to prioritize which preference parameters to elicit; and (v) the resulting outputs, i.e., expected expected utilities, option rankings per stakeholder representative, rank correlations between decision model variants, and the scripts producing the published figures. Associated publication: submitted to *Sustainable Cities and Society* The DOI of the article will be added once available.