Multi-criteria decision analysis: AHP and value functions

Most consequential decisions are not about one thing. Choosing a security vendor is about cost, features, support, and risk all at once. Choosing where to live trades price against location, space, and schools. The objectives conflict, none of them is obviously dominant, and "go with your gut" is hard to justify after the fact.

Multi-criteria decision analysis (MCDA) is the discipline of making exactly these choices in the open. Rather than collapsing everything into a single instinct, it makes the criteria explicit, weighs them deliberately, and scores the options against each one. The result is a decision you can explain, repeat, and challenge. The Analytic Hierarchy Process is one of the most widely used methods inside that family, so it is worth understanding on its own terms.

What AHP is

The Analytic Hierarchy Process (AHP) was developed by the mathematician Thomas Saaty in the 1970s. Its central idea is simple. People are poor at rating ten things at once on an absolute scale, but quite good at comparing two things at a time. AHP leans on that strength. It breaks a decision into a hierarchy, gathers preferences as a series of small pairwise comparisons, and then uses matrix algebra to turn those judgements into weights and a ranking.

What makes it useful in practice is that it handles the soft criteria as comfortably as the hard ones. Cost and throughput are easy to measure. "Reputation" and "cultural fit" are not, yet they often decide the outcome. AHP gives those qualitative factors a defensible numeric weight instead of leaving them as a thumb on the scale.

How AHP works

1. Build the hierarchy

You structure the decision as a tree with at least three levels:

• The goal, the decision you are actually making, for example "choose the best software vendor".
• The criteria, the factors that matter, such as cost, features, support, and security.
• The alternatives, the real options on the table, such as vendor A, vendor B, and vendor C.

Laying the problem out this way forces a shared definition of what "best" even means before anyone argues about which option wins.

2. Compare things two at a time

This is the part that distinguishes AHP. Instead of rating everything at once, you compare pairs and say which one matters more, and by how much, on a scale from 1 (the two are equally important) to 9 (one is absolutely more important than the other).

You answer questions like "for this decision, is cost or features more important, and by how much?" You do this for every pair of criteria to derive their weights, and then for the alternatives against each other under each criterion. Each comparison is small and concrete, which is far easier to reason about than assigning an abstract score in isolation.

3. Calculate the weights, and check the logic

Behind the scenes the comparisons form matrices, and the method derives the final weights from their principal eigenvectors. The output is a clean ranking of the alternatives, with the contribution of each criterion visible rather than buried.

The step that sets AHP apart is the consistency ratio. Human judgements are rarely perfectly coherent. If you say A is more important than B, and B more important than C, but then rate C as far more important than A, your preferences contradict each other. AHP measures this. When the consistency ratio rises above 10 percent, the method flags that the judgements are too contradictory to trust and asks you to revisit them. It is a built-in check on your own reasoning, not just a calculator.

AHP and value functions are not the same tool

Value and utility functions, the methods behind multi-attribute value theory (MAVT) and multi-attribute utility theory (MAUT), sit under the same MCDA umbrella but take a different route to the answer. The distinction matters when you are choosing which to reach for.

On method, AHP works through relative pairwise comparisons, ranking the options against one another. Value functions work through absolute scoring, mapping each option's raw data onto a value, typically between 0 and 1, using a defined curve.

On data, AHP is built for subjective judgement expressed on the 1 to 9 scale. Value functions blend hard, objective measures such as exact costs or speeds with subjective scales just as easily, because everything is converted onto a common value axis first.

On effort and scale, AHP is light for humans to start with, since pairwise comparison feels natural, but the number of comparisons grows quickly as options multiply. Value functions cost more up front, because you have to define the scoring curves, yet they then handle large sets of options without extra burden.

On consistency, AHP actively measures whether your judgements hang together and tells you when they do not. Value functions assume the decision-maker is already rational and consistent, and offer no equivalent check.

Neither is universally better. AHP suits smaller sets of options where judgement dominates and you want a consistency check. Value functions suit larger, data-rich problems where the scoring logic can be defined once and reused.

Why AHP earns its place

A few properties explain why AHP remains popular decades after it was published.

It reduces cognitive overload. By asking you to weigh only two things at any moment, it keeps each judgement small enough to make well, then assembles those small judgements into a coherent whole.

It takes qualitative criteria seriously. Factors like brand reputation or team chemistry resist direct measurement, yet they routinely drive decisions. AHP gives them a transparent weight rather than letting them act unrecorded.

It works well for groups. When a leadership team has to agree, the pairwise comparisons become the agenda. People debate one concrete trade-off at a time, and consensus is built from those small agreements rather than argued over an abstract final score.

The point is the structure, not the number

The value of any MCDA method is not the figure it produces at the end. It is that the reasoning behind the figure is on the table, where it can be examined and improved. A ranking you can trace back to stated criteria, explicit weights, and a consistency check is far more defensible than a confident assertion, and it holds up when someone asks why.

That is the same principle behind quantitative risk work. Structuring a decision so its assumptions are visible, weighted, and open to challenge is exactly what Antan IRM is built to support. If you are trying to bring this kind of rigour to a security or investment decision, let's talk.