ACH is a structured way to test explanations against each other. List every plausible hypothesis, rate each piece of evidence against all of them, and favor the hypothesis with the least evidence against it, not the one with the most evidence for it.
Your analyses are saved only in this browser. Nothing is sent anywhere. Use Save file to keep a copy or share it with a colleague.
Saved in this browser
Next steps
Conclusion
Say how likely each hypothesis is relative to the others, and why the less likely ones were rejected.
Milestones
Future events or findings that would show your conclusion is wrong, or that events are taking a different course.
Export
Preview
How ACH works
Analysis of Competing Hypotheses was developed by Richards J. Heuer Jr. at the CIA in the 1970s to counter a common trap: settling on the first plausible explanation and then noticing mostly the evidence that fits it. ACH makes you consider every reasonable explanation at once, and judge them by the evidence against each one.
- Hypotheses. List every plausible explanation, including ones you think are unlikely and ones involving deception. A group brainstorm works well here.
- Evidence. List the evidence and arguments that bear on the question. Include your assumptions, and the absence of things you'd expect to see.
- Matrix. Rate each item against every hypothesis by working across a row, not down a column. Ask: if this hypothesis were true, how likely is it I'd see this?
- Refine. Reword or combine hypotheses as you learn more. Set aside evidence that's consistent with everything, because it doesn't help you choose.
- Tentative conclusions. Look for the hypothesis with the least evidence against it. Try to disprove hypotheses rather than prove them.
- Sensitivity. Ask which few items your conclusion depends on. What if they're wrong, misread, or planted? Check their sources.
- Report. Discuss the relative likelihood of every hypothesis, not just the winner, so readers see what was considered and rejected.
- Milestones. Note what you'd expect to see if your conclusion is wrong, and watch for it.
Ratings
How the scores work
Weighted inconsistency. Each Inconsistent rating counts 1 point against a hypothesis and each Very inconsistent rating counts 2. Each point is multiplied by the evidence's credibility and relevance (Low 0.5, Medium 1, High 1.5), so a highly credible, highly relevant item that's very inconsistent counts 4.5. Lower is better. Consistent ratings don't add points, because evidence that fits several hypotheses doesn't help you choose between them.
Diagnosticity measures how much an item's ratings differ across hypotheses. Ratings are scored from Very consistent (+2) to Very inconsistent (−2); the spread between the highest and lowest is 0 for "not diagnostic" and 3 or more for "high."
Sensitivity removes each item of evidence in turn and checks whether a different hypothesis would lead. Items that change the leader are flagged as critical.
The scores are an aid to judgment, not a verdict. Heuer's advice was to use the matrix to see your reasoning clearly, then decide for yourself.
Credits
The method comes from Richards J. Heuer Jr., Psychology of Intelligence Analysis (CIA Center for the Study of Intelligence, 1999), chapter 8. This app is a new implementation inspired by the open-source collaborative ACH by Matthew Burton and Joshua Knowles (Burton/Analysis-of-Competing-Hypotheses, 2010). Like that project, it's free software under the GNU GPL v3.