Communications Biology

Methodology

Where these numbers come from

No number on this page was entered at a guess. Each one is gathered from external databases, matched by identifier, validated and computed by a fixed formula. Below is the whole path: where the data comes from, how it is matched and checked, and what the metrics behind the final score measure. Every other metric states its own formula — in the metric grid above.

The data path

Five steps from someone else's record of an article to the number on this page.

1. 1

   Harvesting Records of articles, references and authors are pulled from external sources on a schedule: each source carries its own, stored alongside it. A failure of one source does not stop harvesting from the rest. For every record we track how many independent sources confirmed the data.

2. 2

   Matching A DOI match means the same object, an ISSN match the same journal. A match does not create a new record — it confirms the one already collected and raises the number of confirming sources; where identifiers are missing, titles are compared.

3. 3

   Normalisation Titles, dates and formats are brought to a single form; the anchor identifiers are DOI for articles, ISSN for journals, ORCID for authors and ROR for organisations. Subjects are not derived anew: the topic hierarchy is loaded into a reference book, and ready-made subject assignments are imported and aggregated up to the journal level. Under a separate rule, the language versions of one work are linked into a group rather than merged into one record, and uncertain matches go to manual review.

4. 4

   Scoring Every record gets a confidence score from 0 to 100. It is raised by confirmed identifiers, agreement between several independent sources, completeness of the mandatory fields and freshness of the update; it is lowered by conflicts between sources, missing identifiers and a low-trust source. From 80 the record is published without extra review; 60–79 is published but at reduced confidence, with a manual check advisable; below 60 it is not published until a review confirms it.

5. 5

   Calculation Metrics are computed by the formulas below and published together with their calculation window and confidence score. A metric that does not clear the confidence threshold is left out of the final score; an object that does not clear the publication threshold enters neither the ranking nor the cohort the metrics are normalised against.

Formulas in words

The same explanation that opens next to each metric, collected into a single table. The formulas are fixed: a change of methodology is introduced as a new version with an effective date, and earlier results are kept with their version mark and never rewritten retroactively.

The data quality behind every number

Confidence is a service score of our data about an object, not an expert judgement of the object itself. A low value means the sources disagree with each other, mandatory fields or identifiers are missing, or the data came from a low-trust source.

The platform reconciles differences between sources regularly: critical fields and the metrics that form the ranking at least once a week, less sensitive reference data at least once a month. A small discrepancy is only logged; a significant one creates a manual review task with an assigned owner.

You can report an error in the data through the feedback form: the request enters the tracking system and is followed through to a reply to the person who sent it.