Interesting discussion topic ....
I've been doing a lot of work with document management both at home and at my day job.
At home, I've been using my ScanSnap with Mariner Paperless 2.0 (
https://www.marinersoftware.com/products/paperless/) since it was a version 1.0 product, initially sold by another company.
I always hear good things about DEVONThink, but I recall installing a trial once and finding it was "overkill" for my needs. For starters, I've really never been that "sold" on OCR. I find that inevitably, I've got items to scan in that contain somewhat poor quality text, or a combination of printed text and handwriting, or simply a font that's really small -- and OCR does a terrible job on them.
So whenever one of these packages touts powerful "search" functionality relying on OCR? I automatically ignore it as largely worthless. (Even our $14,000+ corporate doc management package for Windows can't OCR worth a darn. When you view the OCR'd version of a given document it scanned in and saved, it typically thinks various 0's are 8's and vice-versa, makes a few 4's into 9's, or mistakes a random capital R for a capital A, etc. etc. Winds up making it useless when you want to search for a given invoice, RMA or order number.) There's no substitute for manually entering relevant search terms or keywords as you scan in the documents.
I will say Mariner Paperless doesn't feel completely "polished". If you expect a program that feels flawless and absolutely rock solid, it's not quite there. But I find it's very usable and reliable, and well worth the fairly inexpensive asking price. The biggest issue I've seen with it is really just a minor glitch. Sometimes, right after I scan in a new document, it shows its preview/thumbnail as a blank sheet of paper. But as soon as I close the program and re-launch it, everything's fine. Another minor quibble I have is the way it tries to auto-fill document names based on previous entries. Their auto-fill is far from intelligent. It seems to constantly offer the first match alphabetically, vs. offering the most often or most recently used similar name.
(So for example, if you scanned in a document a year ago and titled it "Johnson Corporation utility bill payment"? You might be scanning in a document each week called "Johnson Meyer, Inc. weekly chart of expenses" and EVERY time, it auto-fills the former name when you begin keying in its title, because the C in Corporation comes alphabetically before the M in Meyer.)