Independence

Check whether your multiple sources actually have different failure modes or are secretly the same source repeated.

Three colleagues, on three separate occasions, recommend the same agency for your project. It feels like consensus, so you're ready to sign. Then a stray comment reveals all three attended the same conference talk by the agency's founder. Your three sources are one source, wearing three faces.

The move is to check whether your multiple sources actually have different failure modes, or are secretly the same source repeated. The cue is agreement that feels like independent confirmation. Steps: trace each recommendation back to its origin and ask what each person observed firsthand versus absorbed secondhand; if all of them would be wrong for the same reason (same talk, same dataset, same underlying model), collapse them into a single vote; then deliberately seek one source with a genuinely different error mode, like a former client who paid real invoices and lived with the results. When it works, false consensus deflates to its true weight, or the claim survives independent checking and earns real confidence. Don't use it when one high-quality source is already decisive for the stakes involved; hunting for independence has costs too, and a signed reference can outweigh three rumors.