Matching and Ranking

Initial matching and ranking should be rule-based, transparent, and explainable.

Ranking formula

Example candidate ranking:

ranking_score =
  40 * payment_reputation_score
  + 30 * affordability_score
  + 20 * application_completeness_score
  + 10 * landlord_preference_match

Ranking inputs

Input	Description
Payment reputation	Consent-based tenant score.
Affordability	Rent amount compared with declared income or verified affordability data.
Completeness	Required application data and documents submitted.
Preference match	Match against landlord-defined criteria.

No ML initially

Do not start with machine learning. A rule-based system is easier to validate, explain, tune, and audit.

Explainability requirement

For every ranking result, the platform should be able to explain:

• why the candidate ranked where they did
• which factors improved the result
• which factors reduced the result
• which data was missing
• whether consented reputation data was used

Guardrails

• Do not use sensitive attributes for ranking.
• Do not expose hidden scores without consent.
• Do not rank users based on unverified or disputed data.
• Keep a manual override path for landlords.