The post-LIBOR transition has emphasised the importance of forward-looking risk-free term benchmarks, not only as an accurate asset pricing input but also as an institutional proxy reflecting market quality via liquidity, quote dispersion, and market regimes. In the UK, although Term SONIA Reference Rates are increasingly embedded in loan contracting, derivatives valuation, and collateral and discounting practices, literature on their microstructure quality remains limited. This paper analyses Term SONIA as a layered benchmark and examines (i) how information and noise are allocated across its construction chain (L1, L2, L3/voice broker data, and the published fixing) and tenors, and (ii) whether multi-asset investor sentiment helps explaining time variation in benchmark representativeness. To address these research aims, we adapt the sequential state-space framework to the intraday pipeline of Term SONIA. By estimating phase-specific innovation variances and layer-specific variances of the benchmark via a Kalman filter on rolling windows, we summarize the benchmark quality using information-noise share mechanism. Furthermore, to connect the structural decomposition to quantified magnitudes, we construct representativeness metrics based on deviations of Term SONIA from L1-L3 layers and investigate whether multiple-asset sentiment indices explain benchmark quality across normal and tail regimes. Empirically, the SONIA published rates represent consistently high information content and negligible noise relative to underlying data layers, underlining that the benchmark could act as a noise-reducing aggregator. The raw layers, especially L1, serve as informative yet overwhelmingly noisy inputs, consistent with microstructure theories that observed prices combine efficient components and transitory disturbances. Moreover, representativeness deviations are statistically significant but compress over time, suggesting improving alignment as the post-LIBOR ecosystem takes effective. Meanwhile, the regression evidence shows that sentiment effects are strongly state-dependent with FX sentiment as the most persistent explanatory channel, while interest-rate sentiment is most informative in upper tail, coinciding with large dislocations during rapid policy-path repricing periods. The study provides one of the first systematic, distribution-sensitive assessments of Term SONIA’s internal construction quality, linking benchmark microstructure to sentiment-driven regime dynamics, as well as offering practical implications for benchmark surveillance and governance in Sterling money markets.