Analysis of Competing Hypotheses (ACH) is a structured technique for deciding between several possible explanations of the same evidence. Richards Heuer developed it at the CIA in the 1970s, and it is still one of the most widely taught methods in intelligence analysis, investigative journalism and fraud investigation.
ACH exists to counter a natural habit. People usually pick the most likely explanation first and then collect evidence that supports it. ACH reverses that: you list every reasonable explanation, test each piece of evidence against all of them, and focus on the evidence that rules hypotheses out.
The eight steps
- Identify the hypotheses. Brainstorm every plausible explanation, ideally with colleagues who see the problem differently. Include explanations you think are unlikely but that would matter a lot if true.
- List the evidence and arguments. Include facts, absences (“we would expect X, and it is missing”), and assumptions. Note how reliable each item is.
- Build a matrix. Hypotheses across the top, evidence down the side. For each cell, ask: is this evidence consistent (C), inconsistent (I) or not applicable (N/A) if this hypothesis is true?
- Refine the matrix. Remove evidence that is consistent with every hypothesis. It does not help you choose. Add hypotheses or split them if the matrix shows they are too broad.
- Draw tentative conclusions. Look for the hypothesis with the fewest inconsistencies, not the most consistencies.
- Test sensitivity. Ask which few items of evidence your conclusion depends on. What if one of them were false, mistranslated or planted?
- Report the conclusion with the alternatives you considered and why you rejected them.
- Set indicators. List what you would expect to see in future if your conclusion were wrong, and watch for it.
A worked example
The question: a mid-sized electronics company notices that its biggest competitor has suddenly cut prices on one product line by 25%. Why?
Hypotheses:
- H1: The competitor has a new, cheaper manufacturing process.
- H2: The competitor is clearing stock before launching a new model.
- H3: The competitor is trying to push smaller rivals out of the market (predatory pricing).
- H4: The competitor is in financial difficulty and needs cash.
The matrix:
| Evidence | Reliability | H1 new process | H2 new model | H3 push rivals out | H4 needs cash |
|---|---|---|---|---|---|
| Price cut only on one product line | High | C | C | C | I |
| A trademark application for a new product name filed two months ago | High | N/A | C | N/A | N/A |
| Job postings for “production line ramp-up” at a new plant | Medium | C | C | N/A | I |
| Quarterly report shows strong cash position | High | N/A | N/A | C | I |
| Distributors say the discount is “while stocks last” | Medium | I | C | I | C |
| No change in the competitor’s supplier contracts | Medium | I | N/A | N/A | N/A |
Reading the matrix:
- H4 has three inconsistencies. Most important is that the competitor reports plenty of cash. It drops to the bottom.
- H1 is inconsistent with “while stocks last” and with unchanged supplier contracts. A cheaper process would usually mean permanent price cuts and new suppliers.
- H3 is inconsistent with “while stocks last”. A predatory price cut is meant to last until rivals give up.
- H2 has no inconsistencies. It is the leading hypothesis.
Sensitivity check: the conclusion leans on two medium-reliability items: the distributor comments and the job postings. If the “while stocks last” message was a sales tactic, H3 comes back into play. So the report should say “We assess it is likely that the competitor is clearing stock before a new model launch.” It should name H3 as the main alternative.
Indicators to watch: a product launch announcement within three months (supports H2); price cuts spreading to other product lines or staying in place after stock runs out (supports H3).
Common mistakes
- Too few hypotheses. If the true explanation is not in the matrix, ACH cannot find it. Always add a “something else” column and revisit it.
- Counting Cs. Twenty consistent items do not prove a hypothesis if one solid item contradicts it.
- Treating all evidence as equal. Mark reliability, and be suspicious of a conclusion that hangs on a single low-reliability item.
- Hiding assumptions. “The financial report is accurate” is an assumption. Write it down as an item in the matrix.
- Doing it alone. ACH works best as a team exercise. Disagreement about a cell is useful information.
When to use ACH
ACH takes time, so keep it for questions that matter. Use it when the stakes are high, when the evidence is ambiguous, when there is a risk of deception, or when a group seems to have settled on an answer too quickly. For routine questions, a short key-assumptions check is often enough.
A simple spreadsheet is all you need. The value is not in the software but in the discipline of writing down alternatives and actively trying to disprove them.