Features
Explore the platform capability map.
Explore →GEOLAYERS ARTICLE
1) Core principles
A practical playbook for building transparent, defensible suitability models: data quality, normalization, weighting, constraints, calibration and sensitivity analysis.
Browser-based GISGeoAI + spatial analyticsSouth African context
GUIDE
1. Define the decision before the criteria
Write the decision in plain language and define the unit being ranked: parcels, grid cells, service areas or candidate sites. The model should answer one decision question rather than accumulate every available dataset.
GUIDE
2. Separate constraints from scores
A constraint removes an area from consideration; a criterion changes how attractive it is. Keeping the two separate prevents a high weighted score from masking a fatal constraint.
GUIDE
3. Match data resolution to the decision
High-resolution output does not create high-resolution evidence. Avoid resampling coarse source data into tiny cells and then presenting the result as parcel-level precision.
GUIDE
4. Normalize with meaning
Min-max scaling is convenient but not always appropriate. Some criteria have thresholds, diminishing returns or target ranges. The normalisation function should represent how the variable actually affects suitability.
GUIDE
5. Make weights reviewable
Record the source of each weight, whether it came from policy, expert judgment or stakeholder preference, and test alternative sets. The purpose is not to find a single "correct" weight but to understand how preferences shape the ranking.
GUIDE
6. Run sensitivity scenarios
Compare the top candidates under different weights and thresholds. Stable candidates deserve more confidence; unstable candidates require deeper investigation.
GUIDE
7. Validate against known examples
Where possible, compare the model with existing successful/unsuccessful locations or expert review. Validation does not prove the model is universally correct, but it can reveal obvious structural problems.
GUIDE
8. Publish the assumptions with the map
A suitability surface without criteria, weights and constraints is difficult to audit. In GeoLayers, keep the score fields and scenario context attached to the project so maps and dashboards can explain how the result was produced.
PRACTICAL CHECKLIST
Questions to ask before applying the approach
Use these questions to test whether the workflow is ready for a real project.
- Does each criterion have a defensible relationship to the decision?
- Does source resolution support the precision of the output?
- Are weights and thresholds documented?
- Would a different reasonable scenario materially change the shortlist?