Currently we pick the first populated field that has an institute in it. There is never any in doi data, and we try pubmed first, then scopus. There are a few occasions where the scopus one is better than the pubmed one, but we don't ever get it (as we already have the pubmed one). Ideally we'd try pubmed first, then see if we can clean and geocode it, if we can't then try getting the scopus one and do the same. This is a small number of edge cases though.
Currently we pick the first populated field that has an institute in it. There is never any in doi data, and we try pubmed first, then scopus. There are a few occasions where the scopus one is better than the pubmed one, but we don't ever get it (as we already have the pubmed one). Ideally we'd try pubmed first, then see if we can clean and geocode it, if we can't then try getting the scopus one and do the same. This is a small number of edge cases though.