The Hypothesis Set and the Matrix

Generate before you evaluate. The set must include the explanation you find boring: error, coincidence, ordinary practice. It must also include deliberate deception, because leaving deception out is how a prepared source walks straight through a careful process. A set containing only versions of wrongdoing cannot return an innocent answer; the conclusion was fixed when the set was written.

Then build the matrix: this is Richards Heuer's Analysis of Competing Hypotheses. Evidence down the side, hypotheses across the top, and in each cell a judgment: consistent, inconsistent, or not applicable.

The scoring rule does the work, and it is counterintuitive. You are not counting support. Evidence consistent with every hypothesis distinguishes nothing and comes out of the matrix, however true it is. The surviving hypothesis is the one with the least evidence against it.

This matters because confirming evidence is cheap. Many explanations fit the same facts, so a hypothesis that accumulates support is just a hypothesis that got attention. Inconsistency eliminates, and elimination is what narrows a case.

Then find the sensitive items: the few pieces the conclusion rests on. Recheck those hardest, because that is where a single verification failure takes the whole finding with it.

And keep the alternatives in the report. A finding that never names what it rejected has not shown its work.