Open-plan offices are well-known to be adversely affected by acoustic issues. This study aims to model acoustic dissatisfaction using measurements of room acoustics, sound environment during occupancy, and occupant surveys (n = 349) in 28 offices representing a diverse range of workplace parameters. As latent factors, the contribution of $\textit{lack of privacy}$ (LackPriv) was 25% higher than $\textit{noise disturbance}$ (NseDstrb) in predicting $\textit{acoustic dissatisfaction}$ (AcDsat). Room acoustic metrics based on sound pressure level (SPL) decay of speech ($L_{\text{p,A,s,4m}}$ and $r_{\text{C}}$) were better in predicting these factors than distraction distance ($r_{\text{D}}$) based on speech transmission index. This contradicts previous findings, and the trends for SPL-based metrics in predicting AcDsat and LackPriv go against expectations based on ISO 3382-3. For sound during occupation, $L_{\text{A,90}}$ and psychoacoustic loudness ($N_{\text{90}}$) predicted AcDsat, and a SPL fluctuation metric ($M_{\text{A,eq}}$) predicted LackPriv. However, these metrics were weaker predictors than ISO 3382-3 metrics. Medium-sized offices exhibited higher dissatisfaction than larger ($\geq$50 occupants) offices. Dissatisfaction varied substantially across parameters including ceiling heights, number of workstations, and years of work, but not between offices with fixed seating compared to more flexible and activity-based working configurations. Overall, these findings highlight the complexities in characterizing occupants' perceptions using instrumental acoustic measurements.