Documentation
From a table of numbers to a result you can defend
Three guides and one reference shelf. The guides walk the analysis from the file you upload to the report you cite; the reference shelf holds every method with its assumptions and its source paper.
Where to start
Pick the row that matches what you are trying to do right now. Each one names what you will come away with.
- 013 min read
New here
Run your first analysis
Preparing the decision matrix, setting criterion direction and data type, running, and reading the result screen.
- 021 min read
Choosing a method
Compare candidates before you commit
How Method Lab puts several methods side by side on one finished analysis, and how to read the fitness radar.
- 031 min read
Learning the ideas
The concepts behind the methods
Weighting, normalization, aggregation, outranking, robustness, and how the method families differ.
Reference
Lookup surfaces, for when you already know what you need.
Frequently asked questions
Short answers, each one taken from how the platform actually behaves. The links go to the page that covers it in full.
- What is multi-criteria decision making (MCDM)?
- MCDM is the family of methods that compares alternatives against several criteria at once, including criteria that pull in opposite directions. The input is a decision matrix: one row per alternative, one column per criterion. Every criterion carries a direction, benefit when higher is better or cost when lower is better, and a weight. The method turns that matrix into a single ranking or score. DecisionMind runs 555 such methods on one shared contract.
- When should I use TOPSIS?
- TOPSIS is a distance based compromise method: it scores each alternative by how close it sits to the ideal solution and how far it sits from the anti ideal one. It fits when your criteria are independent of each other, when compensation is acceptable so a strong score on one criterion may offset a weak score on another, and when the decision matrix has no missing cells. It is the wrong choice when criteria are strongly correlated, where DEMATEL or ANP handle the interdependence, or when preferences are non compensatory, where outranking methods such as ELECTRE or PROMETHEE fit better.
- What is the difference between AHP and TOPSIS?
- They do different jobs, so the question is rarely either one or the other. AHP is a weighting method: it derives criterion weights from pairwise comparisons through an eigenvalue calculation, and its output is a set of weights. TOPSIS is a ranking method: it consumes weights and scores the alternatives by their distance to the ideal solution, and its output is a ranking. That is why AHP together with TOPSIS is one of the most common pairings in the library: AHP sets the weights, TOPSIS produces the ranking.
- Why is sensitivity analysis necessary?
- Because the ranking depends on the weights you entered and on the method you picked. Sensitivity analysis perturbs the weights in a controlled way and measures how far the ranking moves, which is what tells you whether the winner is genuinely stable or stable only at the exact weights you typed. Several methods are also known to reverse ranks when the set of alternatives changes; TOPSIS and AHP both carry that limitation in their library records. A single run is a result, not yet a defensible one.
- Which data formats are supported?
- As files: CSV with comma, semicolon, tab or pipe separators (detected automatically), Excel (.xlsx), JSON and TSV. Cells do not have to be plain numbers. Besides crisp values the platform reads intuitionistic (0.6, 0.3), Pythagorean (0.7, 0.4), Fermatean (0.8, 0.5) and neutrosophic (0.6, 0.2, 0.3) fuzzy values, grey or interval values such as [4, 7], and hesitant sets such as {0.4, 0.6, 0.7}.
- How do I choose a method?
- Three routes, and they work together. The analysis flow has a recommendation step that proposes methods matching your data type and your goal, and you can override it at any point. Method Lab takes a finished analysis and puts several candidate methods side by side on the same data, with a fitness radar that shows which part of a score was measured from your data and which part is a property of the method. And every library entry states the method's assumptions and the cases where it is not recommended.