Two years ago, Marcus wrote his master's thesis with AHP-OS, the free browser-based AHP tool that Klaus Goepel built at BPMSG. It was excellent. Fast, clear, a proper consistency ratio at the end, and it cost nothing. He recommended it to half his cohort and he was right to.
So when his company handed him eleven vendor proposals and a Monday deadline, he opened the same tool out of habit, alone in a glass meeting room at seven in the evening with the city lights coming on behind him.
This is the story of the four moments that evening when a tool he genuinely liked stopped fitting the job, and the one moment the following week when it was still exactly the right answer. I am telling it this way because the honest comparison between free and paid AHP tools is not a feature list. It is four questions about your specific decision.
Seven fifteen: the whiteboard arithmetic
Eleven vendors. Six criteria the steering group had already agreed. Marcus started the pairwise comparisons and, being the kind of person who does the arithmetic before he suffers the consequences, stopped after the third one and went to the whiteboard.
Pairwise comparison needs n(n-1)/2 judgments. Eleven vendors is 55 comparisons per criterion. Six criteria makes 330. Add 15 more to weigh the criteria against each other, and he was looking at 345 separate judgments before the tool would give him a ranking.
He knew himself well enough to know how that would go. He would answer the first thirty honestly, the next hundred adequately, and the last two hundred by clicking whatever got him to the end. And the tool would hand him a number that looked just as confident as if he had been careful the whole way.
The fix is a mode that skips pairwise comparison for the alternatives and takes direct 0 to 100 ratings instead, normalising them within each criterion. Our AHP Software ships this as Direct Rating AHP, and for Marcus it turned 345 judgments into 72 numbers: 11 vendors on 6 criteria, plus 6 criterion weights. The pattern that works is to screen the long list with direct rating and then run proper pairwise comparison on the two or three that survive. I walked through that with real arithmetic in the supplier selection story.
If you are ranking four things, none of this applies to you. Pairwise is fine and you do not need us.
Seven forty: the workshop notes that would not collapse
The six criterion weights were supposed to come from a stakeholder workshop Marcus had run the week before. He opened his notes. On the question of whether integration effort mattered more than licence cost, half the room had said "moderately more" and the other half had said "strongly more".
Standard AHP wants one crisp number. Moderate or strong, pick. Marcus sat there for a while, because either choice threw away information the workshop had actually given him. The disagreement was real and it was data.
Fuzzy AHP keeps the range instead of collapsing it. Our software implements this with the fuzzy geometric mean method, alongside three crisp methods: approximate eigenvector, largest eigenvector, and standard geometric mean. Which lets you do something more useful than picking one method and hoping. You run the same model through several and see whether the ranking is robust to the method or an artefact of it.
If your ranking changes depending on how you calculated it, you want to know that before you present it, not from the person in the back row who asks.
Eight o'clock: the question his boss always asks
Marcus had presented to his director before. He knew the question that was coming, because it came every time: what if you are wrong about the weights?
Producing a ranking is the easy half. The hard half is knowing how much of your ranking is load-bearing. This is where a dedicated tool earns its price, and it is the thing Marcus could not get from a browser session that evening.
Our sensitivity analysis works three ways. The result recalculates live as you drag a slider, so you can feel which judgments actually move things. You can switch a criterion off in the weighted attributes chart and see whether the winner survives without it. And one-way sensitivity analysis ranks every variable by how far it would have to move to change the decision, reported as a sensitivity rank of 100 minus the percentage change required. A high rank means a small revision flips your conclusion.
Walking into the room already knowing that the answer holds unless integration effort's weight rises by more than 40 percent changes the meeting. The question gets asked, you answer it in one sentence, and the conversation moves to implementation.
Eight twenty: the line in the non-disclosure agreement
Marcus was about to type vendor pricing into the browser when he remembered the paragraph in three of the eleven proposals that said commercial terms were confidential and not to be transmitted to third parties.
For a coursework example this is a non-question. For a live procurement, a vendor evaluation with pricing under NDA, or research under an ethics approval, it is often the deciding factor on its own. A desktop tool keeps the model in a file on your machine or your network share. Nothing leaves. If your organisation has rules about where evaluation data can be processed, that ends the discussion regardless of how good the online tool is.
Marcus closed the browser tab at 8:22 and the rest of the evening was spent in a desktop application with the project file in a folder his IT department already backed up.
The three department heads who could not be in one room
The steering group had three members who each needed to weigh in, and they were in three time zones. Most free tools handle groups by having everyone join a shared online session at the same time, which is convenient when you can get people together and useless when you cannot.
Our approach is file-based, and it suits exactly this situation. Each decision-maker builds and saves their own project file to a shared network location. One person imports the member files and the aggregation calculates. No cloud setup, no session that expires, and no requirement that everyone is awake simultaneously. The one rule is that every member file has to use the same hierarchy and the same options.
You also get both standard aggregation methods, and the choice matters enough to write down. Aggregation of Individual Judgments combines everyone's pairwise matrices cell by cell with a weighted geometric mean, then derives one set of priorities from the combined matrix. It treats the group as one decision-making unit speaking with one voice. Aggregation of Individual Priorities calculates each member's priorities first and combines the results. It treats them as separate stakeholders whose conclusions are pooled.
Those two can disagree. Marcus's three department heads represented competing budgets, so they were not one voice, and he used individual priorities. He wrote one sentence in his report saying so and why. That sentence is the difference between a method and a black box.
What the free tool was still right for
The following Wednesday, Marcus had two summer interns who had never heard of AHP. He sat them down with AHP-OS.
Being honest about this is the only way the rest of this article deserves to be believed. Free online tools win on friction. Nothing to install, nothing to license, works on a locked-down machine, works on someone else's laptop in a workshop, and you can send a colleague a link rather than a file and an install instruction. When your model is small and your real constraint is getting other people to participate at all, that beats every feature I listed above.
They are also the best way to learn. Nothing sits between you and the pairwise matrix, and if you are still building intuition for what the consistency ratio is telling you, stay free until it clicks. Marcus's interns understood AHP by lunchtime.
The path Marcus would give you
Start free. Build your hierarchy, run the comparisons, read the consistency ratio, get a result you understand. Most decisions stop there and should.
Move to a dedicated tool when you can point at a specific reason, one of the four Marcus hit that evening: more alternatives than pairwise comparison can honestly carry, judgments with real uncertainty in them, a ranking you have to defend under questioning, or data that is not allowed to leave your machine. One of those four, named. Not a vague sense that paid must be better.
If you want to try the pairwise mechanics without installing anything, our free online AHP calculator gives you weights and a consistency ratio in the browser. When one of the four questions turns into a yes, AHP Software has a trial, and the honest test is whether it answers the specific thing your current tool will not.
