Analytic Hierarchy Process (AHP) Calculator

Introduction to AHP pairwise priority weights

The Analytic Hierarchy Process, usually called AHP, converts comparative judgement into priority weights. Rather than forcing a team to assign absolute points to several competing concerns, it asks a more manageable question: between two criteria, which matters more for this decision, and by how much? This calculator handles that criteria-weighting step for three, four, or five criteria. It is useful when a choice must be explained later, such as a supplier selection, site choice, hiring framework, project portfolio, or product trade-off.

Each answer becomes an entry in a square reciprocal matrix. If Cost is judged three times as important as Quality, the Cost-over-Quality entry is 3 and the Quality-over-Cost entry is 1/3. AHP then extracts a set of weights from the matrix. The weights add to 1, so they describe relative emphasis rather than scores: a weight of 0.40 means that criterion accounts for 40% of the decision emphasis. These weights are helpful because they make a decision rule explicit without pretending that difficult trade-offs can be measured perfectly.

The calculator also audits whether the comparisons fit together. For example, if A is five times B and B is three times C, a perfectly coherent set of judgments would make A fifteen times C. Human decisions rarely fit exactly, especially when the scale stops at 9, but the consistency ratio identifies whether the mismatch is modest or worth revisiting. It is a prompt for conversation, not a substitute for professional judgement.

How to use the AHP comparison matrix

Start by selecting the number of criteria and naming them plainly. A label such as “quality” should mean the same thing in every comparison; if it mixes durability, appearance, and compliance, different people may silently answer different questions. Then enter only the upper-right half of the matrix. The diagonal is fixed at 1 because a criterion is equally important to itself, and the calculator fills every lower-left cell with the reciprocal.

1. Choose three, four, or five criteria. Those sizes require 3, 6, or 10 judgments respectively.
2. Use the row label first. A value of 4 in Cost versus Quality means Cost is four times as important as Quality in this decision.
3. Compute priorities. Review the largest weight, λmax, CI, RI, and CR together rather than relying on a ranking alone.
4. If CR is high, revisit the named triple of criteria. It identifies a conflicting chain, not merely a cell with an unattractive number.

For group work, it is often better to record each participant’s matrix first and discuss the largest differences before producing a shared matrix. Ratio judgments are commonly combined with a geometric mean, not an arithmetic mean, because the inputs are multiplicative comparisons. Keep a note of the meeting context and definitions as well: a matrix built during a cash shortage may reasonably change after funding, regulation, or strategy changes.

AHP is most useful when the criteria are distinct and decision-makers can explain their comparisons with evidence. Before beginning, state the decision, the time horizon, and the people affected. A comparison of price and reliability for a one-time purchase may reasonably differ from the same comparison for a long-term critical supplier relationship.

Saaty’s 1–9 comparison scale for AHP judgments

Use values from 1/9 through 9. Values greater than 1 favor the row criterion; values below 1 favor the column criterion. The familiar verbal anchors are 1 for equal importance, 3 for moderate importance, 5 for strong importance, 7 for very strong importance, and 9 for extreme importance. The even values 2, 4, 6, and 8 are useful intermediate positions. A value such as 1/5 means the column criterion is strongly more important than the row criterion.

The scale is deliberately not a measurement instrument with decimal-level precision. Avoid treating 5.3 as meaningfully different from 5.4 unless there is a defensible ratio behind it. Extreme entries also constrain the rest of a matrix tightly. Before entering 9, ask whether the favored criterion would still dominate under a realistic exception. If not, 5 or 7 may communicate the evidence more honestly.

AHP formulas for weights, CI, and CR

Let A be the positive n × n comparison matrix, where aᵢⱼ is the judgment of criterion i over criterion j. The matrix contains one entry for every ordered pair of criteria:

Reciprocity and the diagonal rule preserve the meaning of each pairwise statement:

A perfectly consistent matrix satisfies the multiplication rule below for every chain i, j, and k:

Equivalently, a completely consistent comparison matrix can be generated from underlying positive priorities:

The displayed AHP weights are the normalized principal right eigenvector. The calculation uses power iteration for the positive reciprocal matrix, then verifies λmax from the row ratios.

At each iteration, the temporary vector is scaled so that its entries remain comparable and sum to one:

The table also gives row geometric-mean weights as a cross-check. This is a different estimator, often called logarithmic least squares. Close agreement with the eigenvector is reassuring for a near-consistent matrix.

For a positive reciprocal matrix, λmax is at least n and equals n only under perfect consistency. The consistency index measures the excess, while the consistency ratio compares it with Saaty’s Random Index for a matrix of the same size:

After criteria weights are available, an alternative’s weighted score can be calculated from normalized local priorities or comparable performance scores:

This page uses RI values 0.58, 0.90, and 1.12 for matrices of size 3, 4, and 5. A CR at or below 0.10 meets the usual general guideline. The calculator flags the tighter convention of 0.05 for a 3 × 3 matrix and 0.08 for a 4 × 4 matrix. These thresholds are decision aids, not automatic approval stamps.

Worked example: supplier priorities from four criteria

Suppose a purchasing team compares Cost, Quality, Reliability, and Responsiveness. It judges Cost over Quality as 3, Cost over Reliability as 5, Cost over Responsiveness as 7, Quality over Reliability as 2, Quality over Responsiveness as 4, and Reliability over Responsiveness as 3. Those are the sample values loaded in the form. After computing, Cost receives the largest weight because it wins every comparison, but Quality and Reliability still receive meaningful shares.

The example is not perfectly consistent. Cost over Reliability of 5 and Reliability over Responsiveness of 3 imply Cost over Responsiveness of 15, while the scale entry is 7. The resulting CR can still be acceptable because decision-makers do not make exact algebraic judgments. The useful follow-up is to ask whether 7 was chosen because the difference is genuinely less dramatic than the chained estimate, or whether the team used inconsistent meanings for responsiveness. That discussion is more valuable than mechanically changing 7 just to improve a metric.

Once the team accepts its criteria weights, it can assess each supplier separately under each criterion. For instance, lower price must be converted to a higher-is-better local priority before it is combined with quality and service priorities. The final weighted score is only as credible as those local assessments, so the team should retain quotations, test results, and service history that support them.

Interpreting AHP weights and consistency output

Read the top weight as the criterion receiving the greatest relative emphasis, not as proof that it is objectively most important. Close weights can be practically tied; the 1–9 scale cannot justify fine distinctions that the evidence does not support. λmax and CI are intermediate diagnostics. CR is the concise consistency signal, while the reported worst triple makes that signal actionable by showing which three statements conflict most.

Use the weights only after defining how alternatives will be scored. A typical next step is to rate each alternative on a common higher-is-better scale, multiply each score by the corresponding criterion weight, and add the products. Convert lower-is-better measures such as price or lead time before combining them. Finally, vary an important comparison by one scale step. If the preferred alternative changes easily, the decision is marginal and needs stronger evidence or a clearer policy, not more decimal places.

A result with a low CR deserves scrutiny as well. Consistency does not establish that the criteria are complete, independent, or ethically appropriate. A group can be perfectly consistent about an assumption that is poorly supported. The strongest use of the output is transparent: show the criteria, judgments, sources, and sensitivity checks so that others can understand and challenge the reasoning.

Limitations of this AHP consistency estimate

This calculator supports one AHP criteria matrix with three to five criteria; larger decisions are usually clearer when related criteria are grouped into a hierarchy. A low CR checks arithmetic coherence only. It does not remove bias, prevent double-counting, validate evidence, or make overlapping criteria independent. AHP can also show rank changes when alternatives are added or removed, depending on the synthesis approach. Revisit judgments whenever the decision context changes, and document why each strong comparison was made.

The Random Index values and CR thresholds are conventions rather than universal scientific cutoffs. A high ratio may be acceptable where evidence is uncertain and the inconsistency is openly documented. Conversely, a low ratio should not excuse an unsupported extreme preference. Use the calculator to structure a decision discussion, then apply domain knowledge, legal obligations, budgets, and stakeholder responsibilities before acting.

Frequently asked questions about AHP weights

Sources for AHP methods and consistency guidance

The scale, eigenvector method, consistency index, and Random Index conventions are associated with Thomas L. Saaty’s foundational AHP publications. For further study, see Saaty’s 1977 article, “A scaling method for priorities in hierarchical structures,” Journal of Mathematical Psychology, 15(3), 234–281; Saaty’s 1980 book The Analytic Hierarchy Process; and Crawford and Williams’ 1985 discussion of logarithmic least squares. These sources explain both the mathematical method and the practical caution that internally consistent preferences are not automatically good policy.