Illustrative example: a representative scenario, not a customer story
Hospital Site Selection
Municipal health authority compares 8 candidate locations using demographic, accessibility, environmental, and cost factors with non-compensatory outranking.
The decision problem
A health authority has eight candidate plots for a new district hospital and has to defend the choice in public. Travel time for the population served, land cost, ground and flood risk, and the distance to existing hospitals all pull in different directions. Some of these do not trade off at all: a plot on unstable ground does not become acceptable because it is cheap. The method has to respect that refusal instead of averaging it away.
What the decision matrix looks like
The typical shape of a study like this. Your own matrix can be larger or smaller; nothing here is fixed.
8
Alternatives
5
Criteria
Non-compensatory criteria, measurement thresholds
Uncertainty type
| # | Criterion |
|---|---|
| C1 | Population within reach |
| C2 | Travel time |
| C3 | Land cost |
| C4 | Ground and flood risk |
| C5 | Distance to existing hospitals |
Which method, and why
ELECTRE III is an outranking method with indifference, preference and veto thresholds, so a plot can be excluded on one criterion no matter how well it scores elsewhere. PROMETHEE II runs beside it and turns the same preference structure into a complete ranking through net flows. Entropy derives the weights from the spread in the data itself, which keeps the weighting out of the political argument.
What you get out
Ranking and scores
Every alternative with its score, its rank and the intermediate matrices that produced them.
Weight sensitivity
One criterion weight moves at a time, so you see exactly where the leader changes.
Method agreement
When more than one ranking method is run, the orders are compared with Spearman rho and Kendall W.
Report and citation
PDF, DOCX and XLSX output carrying the seminal source of every method used.
Run this flow on your own data
Load the matrix, choose the weighting and ranking methods, read the sensitivity.