Requirements
Every question is optional — skip anything you genuinely have no view on and it is left out of the arithmetic entirely, rather than scored as neutral.
Recommendation
Reference matrix
Independent of your answers. For each requirement this shows where each candidate is at its best and where it is at its worst, derived directly from the same score tables the recommendation uses.
How the scoring works
Each answer awards every candidate a score from −3 (wrong tool for this) to +3 (headline strength). That score is multiplied by the weight of the question and summed. A few answers are hard constraints rather than penalties: choosing “fully managed” removes anything that ships as a binary you install, and those candidates are listed as ruled out with the reason, never quietly buried at the bottom of the ranking.
The “decisive answers” figures are not raw scores. They measure how far a candidate sits from the average of the surviving field on that same answer, so a requirement every gateway satisfies equally never registers as decisive however important it is to you.
These are opinions, encoded in a table at the top of
assets/scoring.js so that you can read them, disagree with them and
change them. The README documents the data model. For the reasoning behind the
dimensions themselves, see
gateway selection criteria
and
API gateway fundamentals and architecture.