How this works

Methodology & provenance

This platform is only useful if you can check it. Nothing here is generated, inferred or estimated: every record traces to a registered source, and every computed number has a formula printed below.

Non-negotiables

  • We never invent an institution, person, position, project or publication. If we have no source, the field reads “not stated” rather than being filled in.
  • Every entity carries a source URL, discovery timestamp, last-checked timestamp and a verification status. Open the “Evidence” drawer on any record to see all of it.
  • Absence is not evidence. An empty section means we hold no sourced record — it does not mean nothing exists.
  • Deadlines, salaries and positions are mirrors of what the source said at capture time. The official page always wins.

Verification statuses

verified
A human or automated check confirmed the claim against the official source within the recorded verification window.
auto discovered
Captured by a connector from a registered source, not yet re-checked by hand.
needs review
Conflicting or incomplete source evidence. Displayed, but flagged.
possibly outdated
The source page changed or stopped responding since last check.
closed
The source states the call or role has ended.
archived
Kept for the historical record; not part of live counts.
unverified
Present in our database with no confirming source row. Treat as a lead, not a fact.

Deduplication

  • Publications are keyed on DOI first, then on normalised title plus year.
  • Researchers are keyed on ORCID first, then normalised name plus institution.
  • Institutions are keyed on ROR/OpenAlex identifier, then normalised name plus country.
  • Positions use a deterministic dedupe key of institution, normalised title and deadline, so the same call syndicated across three job boards stays one record with three sources.
  • Fuzzy candidates are queued rather than merged automatically; a merge is recorded and reversible.

Momentum & trend signal

growth_ratio = pubs_last_12m / max(pubs_prev_12m, 1)
trend_signal = growth_ratio × log(1 + pubs_last_12m)
  + 0.5 × active_projects + 0.5 × open_opportunities

Caveats worth stating: counts reflect only publications, projects and calls we hold and have classified against a topic, so coverage bias is real — a topic can look quiet because our source list is thin, not because the field is. Small denominators inflate growth ratios. The signal is a reading aid for spotting movement, not a bibliometric ranking, and it must not be used for evaluation of people or institutions.

Source registry

Trust level 5 means an authoritative first-party source (an institution's own page, a society, a funder). Lower levels are aggregators, which we use for discovery but not as the final word on a fact.

Found something wrong? Records marked “demo data” exist to demonstrate structure while connectors are being wired and are labelled everywhere they appear. Anything else that looks incorrect can be traced through its Evidence drawer to the exact page it came from — start there, then tell us which claim and which source disagree. Back to the pulse.