GastroMatch methodology
GastroMatch is built around a simple principle: trust matters more than coverage.
Why methodology matters
- Recommendation systems often optimize for coverage and activity.
- Food choice in unfamiliar contexts has a high cost of error.
- A wrong confident answer is worse than a careful silence.
Trust over coverage
- The system does not need to answer every time.
- Silence is acceptable.
- Explainability is more important than aggressive recommendation.
- Confidence should be visible in system behavior.
What the system tries to do
- Reduce dish-level uncertainty.
- Interpret food choices through Taste DNA.
- Preserve memory of useful choices across places and time.
What the system does not try to do
- Not cover every restaurant.
- Not guess with false confidence.
- Not replace taste with popularity.
- Not behave like a mass restaurant marketplace.
Why growth must be slow enough to stay trustworthy
- Network effect is useful only if local and interpretable.
- Similar users are a hypothesis, not a certainty.
- Aggressive aggregation weakens explainability.
How confidence works
- Local dish signal — from people with similar taste nearby.
- Personal Taste DNA mode — recommendations from your own model.
- The user can choose what to trust.
- The system should remain predictable.
Practical product consequence
- If data is weak, the system should narrow, defer, or stay silent.
- A smaller but reliable system is stronger than a larger noisy one.
Next steps
Open the app, read how it works, or learn more about the project.