Rapid urbanization intensifies demand for urban green space that is not only well distributed on paper but also aligned with residents' lived experience of fairness, accessibility, and everyday usability. As a policy-led test bed for China's "Park City" programme, Chengdu offers an informative case for examining whether objectively measured park provision aligns with residents' lived evaluations of park performance. We linked GIS-derived park indicators with 1866 household surveys from eight typologically distinct sub-districts and analysed overall satisfaction using binary logistic regression and a CHAID decision tree. Across models, park size, perceived park provision, transport convenience, and facility friendliness emerged as the most important predictors of satisfaction. Logistic regression identified strong average effects for usable scale, reliable access, and friendly facilities, whereas CHAID revealed age-segmented heterogeneity: transport convenience is the first discriminator among minors; park size comes first among young adults, with access further differentiating satisfaction within the medium/large-park branch; and among middle-aged and elderly adults, perceived park provision precedes park size in defining the highest-satisfaction pathway. Overall, objective provision alone is insufficient: residents' satisfaction depends on whether parks are large enough, reachable enough, and usable enough in daily life. The findings support a sequenced strategy of securing adequate park scale, removing access bottlenecks, and upgrading inclusive amenities. More broadly, the logit-CHAID framework offers a transferable, decision-oriented approach for evaluating whether green-space investments deliver both equity and effective use.