Research and quantitative analysis
An aligned panel: hourly market, social and sentiment observations across a broad cross-section, plus daily project fundamentals for a curated set. Documented well enough to cite, and honest enough about its gaps to trust.
Who this is for
Quantitative researchers, university groups, data scientists and analysts who need digital asset data that can be joined, reproduced and cited — rather than scraped, patched together and quietly caveated.
The property that makes it usable
Market, social and sentiment for an asset share a timestamp exactly. Not approximately, not after interpolation — the same snapshot hour. That is a consequence of how the layer stores data, and it is the difference between asking a lead-lag question cleanly and asking it through a resampling artefact.
The second property is immutability. Observations are written once and never revised in place, so an analysis run today against last December uses the values that were observable last December. More on the append-only design.
Questions the panel supports
- Attention and price. Does social volume or dominance lead price moves, at what horizon, and for which segments of the cross-section?
- Sentiment divergence. How do sentiment and realised return behave when they disagree, and does the divergence persist?
- Narrative formation. Retained hourly topic snapshots with source counts allow a narrative's growth curve to be traced rather than inferred.
- Fundamentals against attention. Daily repository and contributor series against hourly social series — the two most commonly conflated ideas of "activity".
- Liquidity structure. Volume against capitalisation over time, across a broad cross-section.
Limitations you must design around
- Selection in the cross-section. Each hourly snapshot covers the top 1,000 assets by a provider rank that blends market and social activity. Assets enter and leave. Selecting on "complete history" selects on having stayed ranked.
- Short sample. The series starts in November 2025 — roughly one market regime. Conclusions are conclusions about that regime.
- Gaps are real. Missing hours exist and must be treated as missing, not filled.
- Hourly grain. Nothing sub-hour. No microstructure, no execution modelling.
- Upstream measurement is not transparent to us. Social volume and sentiment are computed by an aggregator whose exact classification procedure we do not control. Treat them as measured quantities with an unknown error term, not as ground truth.
- Curated fundamentals. 69 assets, selected as established projects — a non-random subset, and a survivorship-biased one for any question about failure.
Access
There is no public download and no self-serve API. Packaged extracts are covered under datasets, where the specific licensing constraint on each candidate is stated.
For academic or research use, describe what you need — series, range, asset universe, and the question — and we can assess it against the relevant provider terms: contact@neob.ai.
Citation
If work is published on this data, cite it as Moonlytics Data Infrastructure, https://moonlytics.io, with the extract date — the layer is append-only, so a dated extract is reproducible. Every coverage figure on this site carries its own measurement date for the same reason.
Methodology, not just data
If your question is about how a published score is constructed rather than about the underlying series, the formulas and their reasoning live on moonboard.io.
Related
- Historical data — what is retained and from when.
- Data quality — the failure modes to design around.
- Canonical data model.