Any SEO (or writer, for that matter) knows that in order for a page to rank in regard to a query, it should reflect some understanding of the important, or "key," words in that general subject. The sine qua non not just of the SEO/SEM game, but writing in general.
Traditionally, this meant populating certain sections of page with those words as a signal to the search engine crawler, among other contextual signals, summarized as "Traditional Input Ranking Factors," below (Table 1).
|Keywords in Title|
|Anchor Text Content|
|Unique Linking Domains|
|Page Loading Time|
This also meant that unscrupulous writers began to game that system with keyword stuffing (and more), and in response, the Search team at Google devised increasingly opaque methods of ranking pages.
The game was on. The Great Speculative Search began.
Eagerly, the SEOs would await the prophet Cutts's descent from the Mountain, bearing the Commandments of Content. Cryptically, he would answer their queries, but never reveal the totality of the truth. Perhaps not even he knew.
We had people who would spend hours and hours discussing the results of various "experiments," rather than learning how to research and write competently and build the information architecture of a webpage around the needs of a human, not machine, audience.
(N.b.: Just because you repeat an arbitrary procedure doesn't make what you're doing an experiment.)
And it was, in a word, asinine.
Beginning in the last couple of years, however, there've been big changes to how quality is perceived.
Ranking factors now have a lot to do with the human behaviors around content, summarized in Table 2.
|Long versus Short Clicks (also known as "bouncing" or "pogo-sticking")|
|How often searchers conducting additional, closely-related queries|
|User engagement on a page and the domain more generally|
|Sharing and amplification rates|
...which is actually a lot like what a reasonable person would devise for themselves.
Google's movement toward these kinds of metrics mean you spend less time worrying over the conclusions of an algorithm, and more time on anticipating what readers' reaction to your content will be.
Which puts you back at Square One. That's where you should have started anyway.