Analysis of Competing Hypotheses (ACH)

How the ACH matrix works

ACH is usually described as a sequence of steps, and modern write-ups from threat-intelligence practitioners expand it to seven or eight. First, generate the full set of hypotheses, deliberately including ones you find unlikely, since the method only works if the true explanation is on the list. Second, list the relevant evidence and arguments, including significant absences of evidence. Third, build a matrix with hypotheses across the top and evidence down the side.

The analytical work happens in the cells. For each item of evidence, you ask not whether it fits a hypothesis but whether it is consistent or inconsistent with each one. Evidence that is consistent with every hypothesis has little diagnostic value and can be set aside; evidence that is consistent with one hypothesis and inconsistent with others is diagnostic and does the real discriminating. You then tally inconsistencies rather than confirmations, run a sensitivity check on the few items driving the conclusion, and report the surviving hypothesis together with the alternatives you rejected and why. Documenting the rejected hypotheses is part of the method, it leaves an auditable trail.

Why disconfirmation is the point

The core insight behind ACH is that evidence which confirms a hypothesis is weak, because the same evidence usually confirms several hypotheses at once. A competitor posting senior engineering roles is consistent with building a new product, and also with backfilling attrition, and also with signaling strength to investors. Confirmation alone cannot separate those.

What separates them is inconsistency. If a hypothesis requires something that the evidence contradicts, that hypothesis is weakened in a way that pointing to supporting evidence can never match. ACH operationalizes this by scoring each hypothesis on how much evidence is inconsistent with it and favoring the one with the fewest contradictions. This is where the technique fights cognitive bias directly. Analysts naturally seek confirmation and anchor on an early favorite; forcing every hypothesis into the same matrix and grading on disconfirmation makes it harder to quietly protect a preferred conclusion. It also surfaces the assumption that the real answer might not be among the hypotheses considered: a gap the matrix makes visible.

ACH versus adjacent structured techniques

ACH is one of dozens of Structured Analytic Techniques, and it is easy to confuse with its neighbors. The Key Assumptions Check interrogates the implicit assumptions underneath a single prevailing view; ACH instead lays out multiple explicit alternatives side by side and grades evidence against all of them. The two are complementary and often run together rather than substituted for one another. Devil's Advocacy and Team A / Team B assign a person or group to argue against a consensus conclusion; ACH needs no assigned dissenter and can be run solo or in a group.

It is also distinct from SWOT analysis, a common point of confusion in the CI literature. SWOT profiles one entity's strengths, weaknesses, opportunities, and threats; ACH tests competing explanations for an observed event against a shared evidence matrix. And ACH is not statistical hypothesis testing, it is a qualitative judgment aid built for messy, incomplete, deception-prone evidence, though academic extensions have layered Bayesian or subjective-logic scoring on top of the basic matrix.

Using ACH on competitor behavior

In competitive intelligence, ACH fits situations where a competitor's move admits several readings and picking wrong is costly. A sudden price cut could mean a push for market share, a response to churn, an attempt to clear inventory ahead of a new tier, or a reaction to a specific lost deal. Rather than defaulting to the most threatening interpretation, a team enumerates those hypotheses and grades the observable evidence, pricing-page history, hiring patterns, press activity, review-site sentiment, roadmap signals, against each.

This is where continuous monitoring becomes the raw material for the method. ACH is only as good as the evidence rows in its matrix, and those rows are exactly what competitive-monitoring workflows produce: dated snapshots of competitor websites and pricing pages, job postings, and news. Tools like meertrack that timestamp changes across a competitor's public surface make it possible to fill the matrix with evidence and its absence rather than recollection, and to re-run the analysis when a new signal lands that is inconsistent with the hypothesis you had been favoring.

Limitations and criticisms

ACH has been examined critically in the peer-reviewed literature. Mandeep Dhami's 2019 study in Applied Cognitive Psychology analyzed how ACH is actually used and where it falls short, and a 2024 article in Intelligence and National Security offered a critical review with lessons for the intelligence community. Tim van Gelder has argued that the technique demands too many discrete pairwise judgments and misconceives the relationship between evidence and hypotheses.

The practical limitations follow from the method's structure. If the correct explanation is never written down as a hypothesis, no amount of matrix work will find it. Judging each evidence-hypothesis cell as consistent or inconsistent can impose false precision on genuinely ambiguous signals. The exercise is labor-intensive, which tempts teams to shortcut the evidence-gathering step that gives it value. And like any snapshot method it can go stale; a matrix built on last quarter's evidence should be revisited when the competitive picture moves. ACH reduces bias, it does not eliminate it, and its output is a reasoned judgment rather than a proof.

Frequently Asked Questions

Related terms