Inconsistent Hybrid OA Rates When Re-Running the Same Analysis

Recently, a colleague contacted me with an interesting question.They were re-running an analysis in Web of Science that had first run about a year ago, and the results were not consistent.Specifically, the rates of Hybrid articles were coming back twice as high as they did before.Did I happen to know a reason for this discrepancy offhand?I didn't, but my interest was piqued. What Can We Expect?I know that the overall number of articles returning from the same search will tend to increase over time.Article databases continually index more content as time goes on, so some increase should be natural.As expected, my colleague saw the absolute numbers were higher for the current run, but the percentage of Hybrid articles was increasing even beyond this.Getting a percentage to go up when the denominator is also rising signals the numerator is increasing faster.Something was going on.Open Access status classifications for DOI-level articles (Gold, Green, Hybrid, Bronze, and Closed) come from the service Unpaywall.In my opinion, Unpaywall has emerged as the default standard (rightly or wrongly) for the OA classifications used in bibliometric analyses.The algorithm classifies an article's status based on a number of criteria, like the journal it appears in, any license found, and more[1].I have previously written about some of the challenges, inconsistencies, and changes in Unpaywall's data [2].Now it seems we may have another.