New data sources are valuable only when they improve the question at hand.
Transaction traces, web signals, and geospatial data can add a valuable external perspective. They can also create false confidence when coverage and selection effects are poorly understood.
Before modeling any alternative source, ask what population it represents, how it is collected, and which behavior it cannot see. A highly granular feed is not necessarily a representative one.
Triangulation matters. Signals should be checked against primary research, public sources, and the client’s operational context. Disagreement between sources is often an insight in itself rather than an inconvenience to average away.
The right data is the data that changes a decision. More observations, by themselves, do not constitute better evidence.