Moonboard Data API

Current status

There is no publicly available Moonboard Data API. Moonboard operates backend infrastructure that serves its own application, but that is internal plumbing — not an offering. It has no published contract, no documented authentication, no rate-limit policy, no versioning commitment and no stability guarantee.

Documenting it as if it were a product would create integrations that break without warning, and would mean this site was describing something that does not exist in the form described. Neither is acceptable on a domain whose entire proposition is accuracy.

What is planned

A list of intentions, not a specification. No paths, parameters or response schemas are published here, because none are settled — and publishing invented ones would be worse than publishing nothing.

Design constraints being worked through

These are the reasons this is taking longer than standing up an endpoint, and each one is a genuine design problem rather than a delay:

• Percentile metrics are meaningless without their universe. Four of the seven scores are ranks against the currently scanned set. Any response carrying a rank has to state which universe produced it and when, otherwise consumers will compare ranks that are not comparable. See percentile ranking.
• Missing and zero must be separable. Internally an absent social value is read as zero, which makes a collection failure look like an inactive community. That is tolerable inside a dashboard where a human judges context, and not tolerable in a feed a program acts on. Fixing it at the API boundary is a prerequisite, not a refinement.
• Freshness has to be explicit. Data refreshes on fixed cycles bounded by upstream rate limits. Every response needs to say how old it is rather than implying it is live.
• Redistribution rights are limited. Upstream licences govern what may be passed on and in what form. This constrains what can be offered more than the engineering does.
• Historical scores cannot be recomputed from stored fields. A score recorded in the past was computed against the universe as it stood then. Serving historical scores means serving what was recorded, and being clear that it is not reproducible from the accompanying fields alone.

Available now instead

The methodology is specified in enough detail to reimplement every metric against your own data:

• Methodology overview — normalisation, weighting, time windows, missing data
• Moonboard Score — every threshold and cap, with worked examples
• Data categories — field definitions, units and failure modes
• Score calculator — a working reference implementation

Register interest

Concrete use cases shape what gets built first. Write to office@neob.ai describing the fields you need, the latency you need them at, and what you would build.