As ACH is grounded in the principles of the scientific method, it isn't fit for all problems. Appropriate applications of ACH include:
ACH will stimulate and guide an inquiring mind but will not force open a closed mind. It assumes analysts are truly interested in identifying and questioning assumptions and in developing new insights about the issue in question. It is unlikely that ACH will resolve an impasse between analysts who are firmly entrenched in their views about the issue and have a strong commitment to maintaining those views. If an analyst is unable to see alternative perspectives, the evidence will always be interpreted in a way that supports that analyst's preferred view. ACH can still be useful, however, in helping to pin down the exact basis for the disagreement.
If your goal is mathematical accuracy in calculating probabilities for each hypothesis, there are other versions of ACH that may better meet your needs. They use Bayesian inference or Bayesian belief networks and may require a methodologist trained in Bayesian statistics to assist you through the process. Although the Bayesian probability calculations are mathematically correct, the results cannot be any more accurate than the multitude of subjective judgments about the evidence that go into the Bayesian calculation. Open Source ACH, on the other hand, emphasizes what is practical and understandable for the average analyst to use. Its payoff comes from the analytical process it takes the analyst through, not from precise probability calculations for each hypothesis. The final judgment is made by the analyst, not by the computer.
ACH is not appropriate for all types of decisions. It is used to analyze hypotheses about what is true or what is likely to happen. One might also want to evaluate alternative courses of action, such as alternative business strategies, which computer to buy, or where to retire. In such cases, this software is of limited value. The ACH matrix can be used to break such a decision problem down into its component parts, with alternative choices (comparable to hypotheses) across the top of the matrix and goals or values to be maximized by making the right choice (comparable to evidence) down the side. However, this type of analysis requires a different type of calculation. The principle of refuting hypotheses (in this case alternative courses of action) cannot be applied to a decision based on goals or personal preferences. One would need a more traditional analysis of the pros and cons of each alternative.
Original source: ACH manual by Richards Heuer