Analyze OpenAlex Citation Data without Misleading Rankings

Citation counts are attractive because they are numeric, but ranking papers without context rewards age, large fields, and particular publication types. The OpenAlex Research Scraper exports counts together with the metadata needed for a more responsible comparison.

Collect a defined cohort

Use one topic, a narrow year range, and an explicit cap. Preserve work type and source rather than mixing every record into one leaderboard.

{
  "queries": ["AI drug discovery"],
  "fromPublicationYear": 2022,
  "toPublicationYear": 2026,
  "maxResultsPerQuery": 250
}

Verify that the search population is relevant and not truncated. Store OpenAlex ID, DOI, publication date, work type, source, topics, authors, institutions, citation count, query, and collection time.

Add context before ranking

Compare works within similar publication years and types. Calculate citations per year only as a descriptive rate, noting that partial years and indexing delays distort it. Use medians and distributions instead of reporting only the highest values.

OpenAlex counts can change as sources are added or references are linked. A rising count between snapshots reflects the index as well as scholarly activity. Do not present it as a universal citation total.

Retractions require explicit attention. Keep is_retracted visible and exclude or label such works according to the analysis protocol. Missing DOI or abstract does not imply low quality.

For institutional comparisons, authorship affiliations can be incomplete and multi-institution work should not be counted as exclusively belonging to one organization. Document fractional or full-counting rules. The most defensible output combines quantitative context with source inspection and clearly states query, coverage, time, and the limitations of citation-based inference.