Publishers
Understand how your corpus performs in AI-mediated research discovery.
Readers increasingly arrive at an article through a system that retrieved it, summarized it, and decided whether to cite it. That layer is measurable. This gives a publisher a portfolio-level view of it, and an itemized list of what is holding specific articles back.
Capabilities
Portfolio measurement, down to the individual record.
Everything below reports at both levels. A portfolio number that cannot be traced to the articles that produced it is not useful to an editorial team.
- 01Journal portfolio heatmap
- One view across every journal, imprint, and subject area you publish. Retrieval, citation, and evidence use are shown separately, because a corpus can be found often and used rarely.
- Drill from portfolio to journal to issue to article without changing instrument.
- 02Under-discovered article detection
- Articles whose measured discoverability is well below what comparable work in the same field and age band achieves. These are the highest-yield remediation targets in any catalogue.
- Ranked by the size of the gap and by whether the cause is fixable.
- 03Metadata quality diagnostics
- Missing or unresolved identifiers, absent structured abstracts, thin or inconsistent subject terms, unlinked corrections, ambiguous author records, and license fields that do not match the article's actual availability.
- Every defect is itemized against the record it belongs to, with the field to change.
- 04Query coverage
- Versioned query sets built for each subject area. Coverage reports which questions your corpus answers, which it partly answers, and which it does not reach at all.
- Query sets are quotable objects, so last quarter's figure can be re-derived.
- 05Competitor journal comparison
- For a given question set, see which journals surface instead. Comparison is on public bibliographic records and measured retrieval, not on any private data from another publisher.
- Field-normalized, so subject-area volume differences do not drive the result.
- 06Author-facing reports
- A per-article report an editor can send to a corresponding author: what was measured, what the specific weaknesses were, and which of them the author can act on.
- Carries the methodology version and the run window on every page.
- 07Post-publication monitoring
- Visibility changes over time, with annotations for the events you already know about — a press release, an indexing change, a correction. Changes are reported as changes, not as effects.
- Correlation is shown; causation is not inferred.
- 08API and export
- Programmatic access to portfolio measurements and metadata diagnostics, so results can be joined to your existing production and editorial systems.
- CSV and JSON exports carry provider set, sampling parameters, and version stamps.
Reports and marks
A report you can share. Not a public quality score.
Publishers ask for something displayable. There is a real version of that idea and a harmful version, and the distinction matters enough to state plainly.
What the report is
A dated, versioned document describing a measurement.
- States what was measured, over which query set version, in which window, using which providers.
- Reports retrieval, citation, evidence use, and fidelity separately rather than collapsing them into one number.
- Names its own limits, including sampling error and the parts of the corpus that could only be analysed from metadata.
- Is shareable with authors, editorial boards, and societies as evidence of work done on discoverability.
What it is not
A public badge asserting quality.
- The Trace Visibility Index is a product heuristic under active validation. It is not a validated scientific metric.
- An unvalidated score is not surfaced publicly as a quality metric, and no public-facing badge asserting article or journal quality is issued.
- Scores are internal instrumentation for finding remediable problems. Treating them as a ranking would repeat the failure mode of every metric that outgrew its evidence.
- If validation work later supports a public measure, it will be published with its methodology and its error bars first.
Stated once, plainly
Trace does not publish an article-level or journal-level quality score. Any visibility figure shown in the product carries its methodology version and its provisional status, and is intended for the people doing the remediation work — not for display as a mark of scientific merit.
How it runs
Measure, remediate, re-measure.
- 01Register the corpus
- Identify the portfolio by identifier ranges, journal records, or a supplied list. Public bibliographic sources supply the record; full text is used only where reuse rights are confirmed.
- 02Measure against versioned query sets
- Retrieval, citation, evidence use, and interpretation fidelity are measured separately and reported separately.
- 03Fix what is fixable
- Most of what limits discoverability in a mature catalogue is metadata and availability. Both are within a publisher's control, and both are itemized as concrete record-level changes.
- 04Re-measure and report
- The same query set, run again, on the same instrument version. That is the only way to say whether a change did anything.
Next step
Start with one journal.
A single title with a well-understood readership is enough to judge whether the measurement describes something real. Enterprise pilots are available now.