The first mistake is to treat each page as an isolated competitor. An engine tries to understand WHO is speaking before deciding whether what is said deserves to be shown. It gathers everything it finds about a business: its site, its mentions elsewhere, its profiles, the consistency between what it claims and what others say about it. An excellent page attached to a blurred entity hits a ceiling; a decent page attached to a solid entity goes past it.
So we begin with the entity graph: a structured-data block defined once and emitted on every page, where the organisation, the places, the people and the services point at one another by identifier. It is tedious and invisible to the visitor. It is also what turns a search on your name into a knowledge panel rather than a plain blue link.
The rest follows a strict hierarchy. A slow or badly crawled site has no business publishing; a fast site with no content has nothing to rank; a fast, documented site with no authority stays behind those people cite. We work those three floors in order, and we refuse to write fifty articles on a broken technical base. It shows in the first quote, where the audit always precedes production.