Metrics

Citation impact

Impact Factor

What it shows. The Impact Factor (IF) of a journal is the average number of citations received by its articles over a defined period. It is calculated using data from the preceding two or five years.

When it helps. To compare how often journals in the same subfield are cited on average — for example, when choosing where to submit a paper.

How to read it. A higher IF indicates that the journal’s publications are cited more frequently on average. IF is normalised within the subject subfield (per the OpenAlex taxonomy): a value considered high in humanities may be low in medicine.

CiteScore

What it shows. CiteScore is a Scopus indicator similar to the Impact Factor, but calculated over a three-year window.

When it helps. When you need a measure less swayed by a single unusually good or bad year than the impact factor.

How to read it. CiteScore is less sensitive to anomalous years and covers a wider citation window. The Platform computes it from open Crossref and OpenAlex data, so the value may differ from the one Scopus publishes.

Citations

What it shows. The platform displays two citation-based indicators.

When it helps. Total citations show the overall scale of influence; citations per paper let you compare journals of different sizes.

How to read it. Enables more objective comparison between journals with different publication volumes. Should be interpreted in the context of the subject field.

Immediacy Index

What it shows. The Immediacy Index measures how quickly a journal’s articles begin to be cited.

When it helps. When it matters how quickly work gets noticed — in fast-moving fields where fresh results count.

How to read it. A higher Immediacy Index means the journal’s articles are picked up quickly — common in fast-moving fields like computer science. A low value is normal in disciplines where citation lifecycles are slow (e.g. mathematics).

Cited Half-Life

What it shows. Cited Half-Life is the median age of the articles cited in the current year. It indicates how long the journal’s content remains influential.

When it helps. To see how long a journal’s papers stay in demand — useful if your work is meant to be cited for years.

How to read it. A high Cited Half-Life signals long-lived references — classical or foundational work. A low value is typical in fast-moving fields where older work is rapidly superseded.

Field-normalised

FWCI

What it shows. Field-Weighted Citation Impact (FWCI) compares a journal’s articles against the citation rate expected for their field and year, so a citation in a low-citing field counts as much as one in a high-citing field.

When it helps. To compare journals across fields with different citation habits: 1.0 is the level of the journal’s own field.

Definition

For each article FWCI takes the ratio of its actual citations to the expected citations for its OpenAlex subfield and publication year; the journal’s FWCI is the mean of these per-article ratios. A value of 1.0 means the journal is cited at the field average; 2.0 means twice the field average.

Expected baseline. Mean cumulative citations per paper across the platform corpus, computed per (subfield, publication year).
Window. Articles published in the last five years ([year−5, year−1]); citations are counted cumulatively up to the calculation year.
FWCI is contextual / display-only — it is not one of the ranking weights.

Example

Example. A journal has three articles in the calculation window, all in the same subfield. Each contributes the ratio of its actual citations to the expected ones:

Article A — 30 actual against 10 expected → 3.00
Article B — 0 against 12 → 0.00
Article C — 6 against 8 → 0.75

FWCI = (3.00 + 0.00 + 0.75) / 3 = 3.75 / 3 = 1.25. Every article enters the mean with equal weight no matter how many citations it collected: article A, with its ratio of 3.00, counts here for exactly as much as article B with none. The same three articles give an RCR of 1.20 — that gap is precisely the difference between a mean of ratios and a ratio of sums.

RCR

What it shows. The Relative Citation Ratio (RCR) is the field-normalised companion to FWCI, aggregated across the journal rather than averaged per article.

When it helps. Alongside FWCI: RCR measures the journal as a whole and depends less on individual papers with few expected citations.

Definition

RCR divides the sum of the journal’s actual citations by the sum of the expected citations over the same five-year window and the same per-(subfield, year) baseline — a ratio of sums, where FWCI is a mean of ratios. The two therefore weight large and small articles differently and are reported as distinct metrics. RCR is declared in the target Final Score configuration but does not yet take part in the calculation in force (see the weights table below).

Example

Example. The same three articles as in the FWCI worked example: actual citations 30, 0, and 6; expected 10, 12, and 8.

RCR = (30 + 0 + 6) / (10 + 12 + 8) = 36 / 30 = 1.20.

Article A, with its 30 citations, all but determines the result, whereas in FWCI it was only one third of the mean. Hence the same journal has an RCR of 1.20 and an FWCI of 1.25: one value cannot stand in for the other.

Quartiles

What it shows. A quartile (Q1–Q4) shows which quarter of the subject subfield distribution a journal occupies.

When it helps. For a quick read of a journal’s standing in its subfield — for example, when a requirement says “a Q1–Q2 journal”.

How to read it. Q1 denotes the highest citation impact in its niche; Q4 the lowest. Quartiles across different subject subfields cannot be directly compared without adjustment.

Assignment Rules

Quartiles are computed within the subject subfield per the OpenAlex taxonomy (254 subfields).
Q1 — top 25 % of journals in the subfield; Q4 — the bottom quartile.
If a journal belongs to multiple subfields, a separate quartile is stored for each.
The platform’s overall ranking uses the quartile for the journal’s primary (dominant) subfield (primary_topic.subfield in OpenAlex).

Example

Example. The journal’s subfield holds 40 titles ordered by final score. Each quartile takes 10 journals: Q1 — positions 1–10, Q2 — 11–20, Q3 — 21–30, Q4 — 31–40. A journal in position 12 is assigned Q2.

The same journal also belongs to a second subfield, which holds 80 titles and where it sits in position 15: there the Q1 boundary falls at position 20, so the quartile comes out as Q1. The platform’s overall ranking shows Q2 — the quartile for the primary subfield; Q1 remains an advisory value for the second subfield.

Influence in the citation network

KZ-SJR

What it shows. KZ-SJR is a Kazakhstan-adapted algorithm analogous to the Scimago Journal Rank (SJR). It accounts not only for the number of citations but also for the “quality” of the citing source.

When it helps. To judge a journal’s standing by who cites it — when a plain citation count could be misleading.

How to read it. KZ-SJR is robust against citation inflation through low-significance journals and more accurately reflects a publication’s authority than a raw citation count.

Calculation Principle

Citation matrix. A matrix of inter-journal citations for the calculation period is constructed.
Source prestige. The algorithm introduces no separate quartile pre-weight: a citing journal’s contribution is proportional to its own rank, so a higher-ranked source weighs more on its own. A pre-weight would count prestige twice, since the iteration already carries it.
Iterative computation (PageRank-like algorithm). Scores are iteratively recalculated until convergence.
Local impact coefficient. A citation where both the citing and the cited journal are Kazakhstani is multiplied by an adjustment coefficient in the range 1.10–1.25 (default value 1.15); foreign citations count at full weight 1.0 and are never discarded, and a journal’s self-citation is not amplified. The coefficient raises the weight of domestic citations relative to a journal’s other citations, accounting for the specifics of the Kazakhstani scientific landscape without understating the significance of regional publications.
Manipulation protection. The algorithm detects suspicious patterns:
- anomalously high share of mutual citations between two journals;
- sudden spike in self-citations;
- tight closed clusters with disproportionately high intra-group citation rates;
- abrupt citation jumps without a commensurate increase in publication activity;
- mass publications without verified peer-review practice.
Suspicious links are not removed but down-weighted in the ranking calculation. Critical cases are referred for manual expert review. Decisions are recorded in the methodology.
Normalisation. Final values are normalised to a 0–100 scale and distributed into quartiles within the subject subfield.

Caveats and example

Important caveats:

The local impact coefficient is a configurable methodology parameter, not a fixed constant.
The coefficient can only be changed through the methodology versioning mechanism, with full change history preserved.
Thresholds for anomaly detection are fixed in the methodology and versioned.

Example. Journal A received 100 citations, 60 from Q1-journals and 40 from Q3-journals. Journal B received 100 citations, all from Q3-journals. Journal A will have a higher KZ-SJR because the quality of its citing sources is higher.

Eigenfactor and AIS

What it shows. Eigenfactor measures a journal’s influence on the 5-year journal-to-journal citation graph, with self-citations excluded. It is a PageRank-style steady-state score scaled so the sum across all journals equals 100 — each journal’s score is its share of total network influence.

When it helps. Eigenfactor shows a journal’s overall weight in the citation network; AIS compares the average paper regardless of journal size.

How to read it. Eigenfactor reflects raw network influence including journal size. AIS removes the size effect: it answers “is the average article in this journal more or less influential than the Platform corpus average?”

Formula

EF = steady-state of P (5y, no self-cites), rescaled so Σ EF = 100

AIS = 0.01 · EF article_share

Variables and example

Variables:

P — row-stochastic journal-to-journal citation matrix over the 5-year window [Y−4, Y]; the diagonal is zeroed (no self-citation mass).
EF — share of total network influence, in percent (Σ over all journals = 100).
article_share — the journal’s fraction of all articles published in the same 5-year window.
AIS (Article Influence Score) — average-impact reference value is 1.0.

Example. Journal A has EF = 0.6 (i.e. 0.6 % of the network) and 2 % of the corpus’s articles in the window. AIS = (0.01 · 0.6) / 0.02 = 0.30 — below the average-impact threshold of 1.0. The network here is the Platform corpus, that is Kazakhstani work: a domestic comparison base, not a world one.

Productivity

h-index

What it shows. The h-index (Hirsch index) characterises both the productivity and citation impact of a researcher, journal, or organisation simultaneously.

The h5-index is calculated identically but restricted to articles from the last 5 years.

When it helps. To gauge the steady citation record of a journal, researcher or organisation without one or two highly cited papers skewing it.

How to read it. The h-index is robust against single highly-cited outliers (unlike IF) and enables comparison between journals with different publication volumes. Cross-discipline h-index comparisons require normalisation.

Definition

The h-index equals the maximum number h such that h publications have each been cited at least h times.

Example

Example. Sort a journal’s articles by descending citation count:

Article	Citations
1	45
2	38
3	22
4	20
5	19
6	4

Articles 1–5 have each been cited ≥ 5 times, while article 6 has only 4 citations, fewer than its rank of 6. The threshold cannot go higher, so the h-index = 5.

Publications

What it shows. The Np (number of publications) indicator is used in the university ranking and characterises the publication activity of an organisation.

When it helps. To gauge a university’s publishing activity — together with citation measures, not instead of them.

How to read it. A high Np indicates active publication activity but should be considered alongside citation impact indicators (Nc, h-index) and quality indicators (Q1 share).

What is counted: peer-reviewed articles in journals indexed by Crossref and OpenAlex.

Normalisation. In a university’s final score, Np is scaled to [0, 1] against the maximum among Kazakhstani universities; it is not divided by headcount. Size normalisation is carried by a separate indicator — citations per staff member: it enters the final score with a weight of 8 and is additionally published as a standalone sortable view of the ranking.

Subject rankings: separate rankings are calculated per subject field, enabling universities to be compared in the context of their research profile.

Example

Example. A university published 1,200 articles over the calculation period, while the highest figure among Kazakhstani universities is 3,000. Np_norm = 1,200 / 3,000 = 0.40. With a publication-count weight of 8 % and the remaining eight indicators populated, Np contributes 0.40 × 8 = 3.2 points out of 100 to the university’s final score.

Headcount plays no part in this calculation: 1,200 articles yield the same normalised value at a university of 500 staff and at one of 5,000.

Openness and collaboration

Open Access share

What it shows. The Open Access (OA) share is the fraction of the journal’s recent articles that are openly available.

When it helps. When it matters that readers can access the paper without a subscription — or to find open access journals with the licence you need.

How to read it. A high OA share indicates broad accessibility. closed and unknown statuses do not count as OA.

Licence type and DOAJ indexing. The open-access licence type (CC BY, CC BY-SA, CC BY-NC, CC BY-ND) and the journal’s DOAJ-indexing flag come from DOAJ (the Directory of Open Access Journals) — the source for the licence filter. These are journal-level attributes, distinct from the OA share above, which is computed from the open-access status of individual articles (Unpaywall). The DOAJ Seal was retired by DOAJ on 8 April 2025, so the journal’s presence in the directory is shown instead.

International Collaboration

What it shows. The international co-authorship indicator measures the share of publications that include authors from foreign organisations.

When it helps. To see how well a journal or university is connected to international research networks — for example, when looking for co-authors abroad.

How to read it. A high international-collaboration score signals integration into global scientific networks. For Kazakhstani universities this indicator forms part of the institutional ranking.

For journals (intl_coauthor_share): the same ratio is computed over the journal’s articles in the most recent 5 years; an author is “foreign” when their affiliation’s country differs from the journal’s country. Authors with no resolvable country in the window are dropped from both numerator and denominator. Source: Crossref/OpenAlex affiliations.

Inter-University Collaboration

What it shows. These two advisory indicators describe a Kazakhstani university’s domestic co-authorship with other Kazakhstani universities. They are shown for transparency and analysis only and are not inputs to the institutional ranking (final_score).

Inter-university collaboration is the share of the university’s publications with at least one co-author affiliated with a different Kazakhstani university during the year. Partner universities is the count of distinct other Kazakhstani universities that co-authored those publications — a measure of collaboration breadth.

For journals (inter_university_coauthor_share): the share is computed over the journal’s articles in the most recent 5 years; an article counts as inter-university when its authors are affiliated with two or more different Kazakhstani universities (active affiliation in the article’s year, non-estimated). Source: Crossref/OpenAlex affiliations. Advisory — not part of the ranking.

When it helps. To see how widely a university collaborates with other Kazakhstani universities; it does not affect the ranking position.

How to read it. These indicators highlight a university’s role in domestic research networks. They are analytical only and do not affect the official ranking.

Grant Activity

What it shows. Grant activity is a composite indicator used to rank universities: it captures an organisation’s ability to attract research funding. It is not a single number but an aggregate of several components.

When it helps. To gauge how well a university attracts research funding — for example, when choosing where to do project work.

How to read it. In the university weight configuration in force, grant activity carries a weight of 6 out of 100 — so it affects a university’s ranking position directly. A university with none of the indicator’s components filled in is dropped from the weighted sum together with that weight rather than scored zero on it: missing data does not depress the score.

Components

Number of active grants — grants running during the reporting period.
Number of won grants — grants awarded during the reporting period.
Total funding volume — in tenge; foreign-currency grants are converted at the rate on the date.

Degradation with incomplete data

If the funding amount is unavailable, only the grant counter is used.
If none of the components is filled in, the indicator does not take part in the ranking calculation and is not substituted with zero.

Sources (in priority order): the official grant registry (where access is available); data imported from the university via interface / file / API; manual verification by an administrator.

Example

Example. A university has all three components filled in: 8 active grants, 6 won, and 120 million tenge of funding. The maxima among Kazakhstani universities are 40, 20, and 300 million tenge.

Funding — 120 of 300 million tenge → 0.40, weight 0.5
Won grants — 6 of 20 → 0.30, weight 0.3
Active grants — 8 of 40 → 0.20, weight 0.2

Indicator = 0.5 × 0.40 + 0.3 × 0.30 + 0.2 × 0.20 = 0.33.

If the funding amount is not recorded for that same university, degradation applies: the two counters remain and their weights are renormalised from 0.3 and 0.2 to 0.6 and 0.4. Indicator = 0.6 × 0.30 + 0.4 × 0.20 = 0.26 — lower, but still a measured value rather than a substituted zero.

The score

Final Score

What it shows. A journal’s Final Score is a weighted sum of normalised metrics that determines its position in the platform’s overall ranking.

When it helps. For an overall comparison within one subfield; the weights table below shows what the score is made of.

How to read it. The Final Score is normalised to a 0–100 scale. Scores should only be compared within the same subject subfield.

University final score

A university’s final score is the weighted sum of its nine normalised indicators; it determines the position in the overall university ranking (the default sort). The weight configuration in force (versioned together with the calculation methodology):

Average KZ-SJR — 22%
h-index — 18%
Share of Q1 publications — 13%
Number of citations — 12%
Number of publications — 8%
Average Impact Factor — 8%
Citations per staff member — 8%
Grant activity — 6%
International co-authorship — 5%

Normalisation. Each indicator is scaled to [0, 1] against the maximum across all Kazakhstani universities; the result is scaled to 0–100. An indicator a university has no data for does not participate (the score is not diluted with zero). Normalisation is global — across all Kazakhstani universities at once. Size normalisation is carried by a separate weighted indicator (citations per staff member), while separate per-subject-area rankings are computed as an advisory layer on top of the main recomputation and do not enter the final score.

Formula

Score = w_IF · IF_norm + w_CS · CS_norm + w_h · h_norm + w_KZ · KZ-SJR_norm + w_Q · Q_norm + …

Final-score weights. The table below is the target weight configuration. The methodology version in force may apply a subset of it and distribute weight among its own indicators in its own way: an indicator enters the calculation once it is populated across the whole corpus, and the set changes by putting a new methodology version into effect, not on its own.

Which indicators enter the calculation today, and with what shares, is shown by the “What makes a journal’s score” bar at the top of the page: it is read from the methodology version in force and changes with it. The shares are not derived from the target percentages — the table and the bar are comparable by composition only, never by magnitude.

Metric	Weight in the target configuration
KZ-SJR	20 %
Impact Factor	25 %
h-index	15 %
CiteScore	15 %
RCR	15 %
Quartile	10 %

Caveats and example

Important caveats:

The values above are the starting configuration; experts can change any weight via the Management Console.
The system supports separate weight profiles by subject field (medicine, humanities, engineering, etc.).
Normalisation is performed within a subject subfield per the OpenAlex taxonomy (254 subfields), not globally across all journals.
A weight profile can be set at any of the three top levels of the classifier — domain, field or subfield. A journal takes its weights from the first level that has a usable one: its primary subfield first, then that subfield’s field, then the domain; if none of the three qualifies, the general default configuration applies.
Each weight configuration has a unique version and effective date; the version used is recorded at each scoring run.

Example. Suppose the methodology version in force applies three indicators with shares of KZ-SJR 50 %, impact factor 30 % and h-index 20 % — for another version, substitute the shares from the bar at the top of the page. Within its subfield a journal was assigned the normalised values KZ-SJR_norm = 0.60 and IF_norm = 0.90, and it has no h-index.

Score = (0.60 × 50 + 0.90 × 30) / (50 + 30) × 100 = (30 + 27) / 80 × 100 = 57 / 80 × 100 = 71.25.

The divisor here is 80, not 100: it is the sum of the shares of the indicators the journal has. The weight of a missing indicator is dropped along with its share, so the score is not diluted with a zero — a journal without an h-index is not penalised against one that has it. Had the journal h_norm = 0.50, the score would be (30 + 27 + 10) / 100 × 100 = 67.0.

Data Methodology

Platform data is collected from open scientific sources: Crossref, OpenAlex, ORCID, Unpaywall, DOAJ, and other open registries. Every data record carries a confidence score (0–100) reflecting the degree to which the information has been verified. The confidence score is factored into metric calculations and displayed in the interface.

Update schedule:

Scheduled ranking recalculation — once per day during night-time hours.
Manual recalculation — initiated by an administrator after changing weights or coefficients.
During recalculation, public pages continue displaying previous values; new values become public atomically upon completion.
Each recalculation records the version of weights and methodology used.

Frequently Asked Questions

Why does a journal have different quartiles in different sources? Because the quartile depends on which set of journals is used for comparison and which metric underlies the ranking. RankScience calculates quartiles within OpenAlex subfields using the KZ-SJR formula; this may differ from Scopus or Web of Science quartiles.

Why does the Impact Factor on the platform differ from JCR? JCR is calculated using Clarivate Web of Science’s proprietary database. RankScience uses open data from Crossref and OpenAlex. Differences in citation bases can produce divergent values.

What does the “Final Score” mean and can it be compared with global rankings? The RankScience Final Score is an internal platform indicator normalised within the Kazakhstani corpus. Direct comparison with QS, THE, or SCImago scores is not appropriate without additional adjustments.

How often is the data updated? Rankings are recalculated once per day. Source data from external providers is refreshed on a schedule that depends on the API limits of the specific provider.

What makes a journal's score

Citation impact

Impact Factor

CiteScore

Citations

Immediacy Index

Cited Half-Life

Field-normalised

FWCI

RCR

Quartiles

Influence in the citation network

KZ-SJR

Eigenfactor and AIS

Productivity

h-index

Publications

Openness and collaboration

Open Access share

International Collaboration

Inter-University Collaboration

Grant Activity

The score

Final Score

University final score

Data Methodology

Frequently Asked Questions

Further Reading

Compute a metric from your own numbers

Worked example