There is no single metric that captures “real” scientific impact. Contemporary research‑metrics literature converges on multi‑dimensional, hybrid frameworks that combine traditional bibliometrics (citations, h‑index, field‑normalized indicators) with altmetrics (online attention and engagement) and qualitative evidence (policy change, guidelines, practice impact) [1][2]. Frameworks such as the Leiden Manifesto, Snowball Metrics, and the EMPIRE Index for medical publications explicitly advocate this integrated approach [1][3].

Key metric types and what they capture

1. Traditional citation-based metrics

• Raw citations (per paper, per author):

  • Strength: direct measure of scholarly uptake in the literature.

  • Limitation: field‑, age‑, and language‑biased; susceptible to citation cartels and size effects.

• Field-normalized metrics (e.g., FWCI – Field‑Weighted Citation Impact):

  • Compare citations of a paper to the world average in its field and year [4][5].

  • FWCI ≈ 1: average; >1: above average influence.

  • Better for cross‑field and early‑career comparisons than raw counts.

• h-index and variants:

  • Measure sustained output plus citation impact.

  • Useful as a coarse screening tool but heavily age‑ and field‑dependent and insensitive to a few very high‑impact works [4].

1. Altmetrics (article-level metrics of online attention)

• Include:

  • Social media mentions (Twitter/X, Facebook).

  • News and blog coverage.

  • Mendeley readers, downloads, bookmarking.

  • Policy document mentions in some systems.

• Systematic review evidence [1][2]:

  • Altmetrics are best seen as complements, not replacements, to citations.

  • Moderate correlations with later citations in some fields (e.g., clinical/translational science), but they capture different dimensions—speed and breadth of attention, including outside academia.

• Platforms:

  • Altmetric Attention Score, PlumX Metrics, publisher dashboards [1][6][7].

1. Multi-component impact frameworks

• EMPIRE Index (medical publications):

  • A value-based, multi-component metric distinguishing:

  • Social impact: news, blogs, Twitter, Facebook, Wikipedia.

  • Scholarly impact: Mendeley readers, citations, F1000Prime recommendations.

  • Societal impact: mentions in clinical guidelines, policy documents, patents [3].

  • Benchmarked so that 100 equals the mean impact of NEJM Phase III trial papers in 2016.

  • Validated on thousands of trials; predictor scores based on early altmetrics explain ~70% of variance in later total impact (r² ≈ 0.69) [3].

• Snowball Metrics & institutional dashboards:

  • Combine publication counts, field‑normalized citations, collaboration indicators, and usage/altmetric data into institution‑level profiles [1].

• Leiden Manifesto and NISO altmetrics recommendations:

  • Emphasize responsible metrics: transparency, field normalization, and a portfolio view rather than a single score [1].

1. Qualitative and societal impact evidence

• Policy citations, clinical guideline inclusion, and documented changes in practice are often under‑captured by numeric metrics alone [1][3].

• Many frameworks now recommend including:

  • Case studies of policy or clinical impact.

  • Evidence of open‑source software re‑use, standards adoption, and spin‑offs.

So, what “best” measures real impact now?

The best current practice is not a single metric but a concise, interpretable bundle of indicators tailored to the decision context:

1. For individual researchers (hiring, tenure, awards)

• Use:

  • A field-normalized citation metric (e.g., FWCI, percentile ranks) to quantify scholarly uptake [4][5].

  • Selected article‑level indicators (citations to key papers; EMPIRE‑like component scores in medicine where available).

  • Altmetrics snapshots for major works, to show reach and timeliness (news, policy, Mendeley, downloads) [1][2].

  • Short impact narratives linking outputs to policy, practice, or technology changes.

1. For individual articles

• Combine:

  • Citations, field‑normalized where possible.

  • Altmetric profile (are clinicians, policymakers, or the public engaging?).

  • In medicine and clinical research, guideline/policy mentions or EMPIRE Index‑style societal scores, when available [3].

1. For journals and institutions

• Avoid relying solely on Journal Impact Factor.

• Prefer:

  • Composite dashboards with CiteScore, FWCI, citation distributions.

  • Collaboration and open‑science indicators (data/code sharing, preprints).

  • Altmetric and societal‑impact metrics where relevant [1][5].

Counterarguments and cautions

• Over‑complex dashboards risk obscuring rather than clarifying impact.

• Altmetrics can be gamed, biased towards English‑language and social‑media‑active communities, and may reflect popularity more than quality [1][2].

• Even sophisticated indices like EMPIRE depend on expert weighting choices and proprietary data sources [3].

• For early‑career researchers, many metrics are noisy; evaluation should heavily weight expert qualitative judgment alongside metrics.

Actionable guidance

For scientists preparing evaluations:

• Report no more than 5–7 metrics, e.g.:

• Total citations; h‑index.

• FWCI or similar field‑normalized indicator.

• For 3–5 key papers: citations + Altmetric score + any guideline/policy mentions.

• Explicitly contextualize:

• Field norms (e.g., typical citation rates).

• Co‑authorship patterns.

• Open science practices (preprints, data/code sharing).

• Use recognized principles (Leiden Manifesto) to argue against crude metric use and to support portfolio‑based assessment.