The two major bibliographic databases - Scopus and Web of Science\Web of Knowledge - have one very badly-documented limitation; their update lag. While the majority of papers are added quickly, it can take several months for individual papers to be added.
As an example, I ran a search in Web of Science looking for all papers published by my institute in 2014. In November 2015, it reported 290 papers. In January 2016, 293. In February, 305.
In other words, a few percent of the papers took a year or more to be added to the database, which startled me - I was expecting a lag of perhaps three months. Some of these delayed entries are from relatively obscure journals (specialist Australian publications) or book series, and it seems reasonable that these might take longer, but most are from mainstream titles - one is even from Philosophical Transactions of the Royal Society, as established a journal as it's possible to get.
Scopus is comparable. In November 2015, it reported 340 2014 papers, 356 2013 ones and 351 2012. Running an identical search three months on brings us to 341, 367 (!), and 353 respectively - so a few papers are still being added up to three years after publication. (Unfortunately, Scopus doesn't expose the date items added to the database, so it's not possible to identify which ones were added most recently - it's possible that some were due to journals being added to Scopus for the first time along with their back-issues.)
This is, of course, something of a headache for producing stats. But more practically, it could be a real problem for disseminating science - something being unintentionally missed or delayed from one of the big indexing services can make it much less discoverable and much less likely to be read, built upon, or cited, especially in that important first year.
So, the core question: does anyone know of documentation (or research) into how substantial these delays are; how they're distributed; and whether they're getting better or worse?
(I've spent some hours searching for this but to no avail. I'm wondering if I may just be looking in all the wrong places.)