DIAGNOSTIC GUIDE
Matching Algorithm Readiness Assessment
Executive Summary
Algorithms fail when inputs fail—sparse skill tags, missing availability, and unweighted 'nice-to-have' fields produce nonsense recommendations. Readiness means structured founder intake, normalized mentor taxonomy, defined weights for must-have vs preferred criteria, and human override workflow. Do not enable auto-matching until a blind test on last cohort's data produces sensible top-three suggestions for 80% of founders.
Key Takeaways & Benchmarks
- Validate taxonomy: fewer than 10% uncategorized mentor skills
- Define weights: stage, sector, function, availability, conflicts
- Run retrospective match test against known good pairings
- Train coordinators on override logging for algorithm tuning
- Start with recommendations, not autonomous pairing
Frequently Asked Questions
How many data fields are enough?
Five to eight weighted fields beat twenty unweighted checkboxes. Quality over quantity.
Can we tune algorithms mid-cohort?
Adjust weights for new matches; avoid rematching entire cohort unless crisis—document changes.
What if recommendations look wrong?
Inspect input data first—90% of bad suggestions trace to stale or missing mentor tags.
Next Step for Your Mentor Program
Explore matching quality, roster coverage, and intake workflows on your cohort.
Check roster readiness →