Researchers
Your paper is published. Whether it is found is a separate question.
Most work that goes unread does not fail on quality. It fails on terminology, structure, availability, and metadata — all of which are fixable, and most of which are cheapest to fix before submission. This is instrumentation for that, built for the person who wrote the paper.
Start here
Audit a paper and get a baseline.
- 01Audit a new paper
- Point the product at a DOI, a preprint, or a manuscript you have the rights to process. It produces a baseline: which questions retrieve the paper, which retrieve something adjacent, and which do not reach it at all.
- 02Retrieval, citation, and use are measured separately
- A paper can be retrieved constantly and never used to support an answer. Collapsing those into one number would hide the only diagnosis that leads anywhere.
- 03Check how it is being described
- Where the paper is summarized, the summary is compared against your own stated population, endpoint, and limitations. Overstatement is reported as a discrepancy against the source.
Diagnosis
Understand why the work is not being surfaced.
A number that says your paper is under-discovered is not useful on its own. What matters is which of these is responsible, and whether you can do anything about it.
- 01Terminology mismatch
- The most common cause. The paper uses the vocabulary of its subfield; the question uses the vocabulary of the person asking. Both are correct, and the paper is not reached.
- 02Abstract structure
- An abstract that buries the population, the intervention, and the outcome in a narrative paragraph gives a retrieval system very little to match against.
- 03Availability
- Where no lawful open version exists, far less of the paper is available to be read by anything. Availability is reported as a factor because there is a plausible mechanism, not because a causal effect has been proven.
- 04Metadata defects
- Unregistered identifiers, missing subject terms, unlinked preprint and version-of-record records, and ambiguous author identity all reduce the chance the right paper is the one that surfaces.
- 05A better-positioned neighbour
- Sometimes the paper is fine and another paper simply answers the question more directly. That is worth knowing too, and it is reported rather than explained away.
Before submission
Improve terminology while it is still editable.
- 01Improve terminology before submission
- Run the diagnosis on a manuscript rather than a publication. Terminology, title specificity, and abstract structure are trivially editable before submission and effectively frozen afterwards.
- 02Suggestions, never rewrites of the science
- The product proposes vocabulary and structure. It does not alter findings, strengthen claims, or generate results, and every suggested change is shown as a diff you accept or reject.
- 03Subject terms that match how people ask
- Controlled vocabulary terms and author keywords are proposed from the questions your work actually answers, not from what is currently trending.
- 04No keyword stuffing
- Clarity is the mechanism. Suggestions that would make the abstract less accurate are not offered, and repetition tactics are explicitly out of scope.
A boundary worth stating
Trace does not write your paper, generate results, or strengthen a claim beyond what your data supports. It works on how clearly the work is described, and it flags any suggestion that would make the description less accurate.
Publish
Give the work a machine-readable home.
- 01A machine-readable research page
- Generate a structured page for a paper: the claim, the population, the method, the limitations, the identifiers, and links to the version of record and any open version. Structured markup is included so the record is parseable rather than merely readable.
- 02Attributable by construction
- The page is published under your name and your institution's. There is no anonymous channel and no synthetic amplification anywhere in the product.
- 03Transparent launch materials
- Plain-language summaries, structured briefs, and slide-ready statements that are checked against the source for overstatement before you use them.
After publication
Monitor visibility and benchmark against similar work.
- 01Monitor visibility over time
- The same query set, run on a schedule, on a stated instrument version. Changes are annotated with events you already know about — a preprint going to press, a repository deposit, a correction.
- 02Change is reported, not explained
- A shift after an event is shown as a shift after an event. No causal claim is made from an observational time series.
- 03Benchmark against similar work
- Compare against a set of papers matched on field, age, and study type. The comparison set is listed explicitly so you can judge whether it is fair before you read the result.
- 04Notice distortion early
- Where a summary of your work drifts from what the paper says, that discrepancy is surfaced with the source text alongside it.
Access
Individual plans are in private beta.
Individual accounts are limited while the measurement is validated. If you work at an institution or publisher that would benefit from portfolio-scale measurement, an enterprise pilot is the faster route.