Neighbourhood Signals investigates how hyperlocal digital discourse maps onto offline crime dynamics across UK communities. Using a large longitudinal dataset from the hyperlocal Nextdoor platform linked with police-reported crime and socio-demographic data, the project models predictive relationships between online behavioural indicators (post volume, sentiment, topics, latent social signals) and crime outcomes such as VAWG, knife crime, and anti-social behaviour, aligning with Strand 3. The study applies transparent, reproducible spatial–temporal modelling and Natural Language Processing techniques to identify early warning indicators of emerging safety risks. It will produce an open, extensible dataset and methodological framework offering robust empirical evidence on how neighbourhood-level online activity can support crime prediction and safer streets initiatives.