The short answer
An SEO project has six stages: audit and diagnostics, keyword research, structure and landing page design, technical fixes, content with internal linking, and external signals with regular measurement. The first three are done once and produce the plan. The last three run in cycles and don't finish while the site is being promoted.
The order isn't a formality. Keyword research done before an audit is done blind — you don't know which pages are already indexed or which of them compete with each other. Content published on a site with blocked indexing or heavy duplication won't bring traffic no matter how much of it you write.
One boundary up front: this is the shape of a normal commercial site — services or e-commerce. For a news project or an aggregator the emphasis shifts, and on a site that has been in the top for a year half of these stages collapse into maintenance.
The sequence at a glance
Blue marks the stages done once, which produce the plan. Teal marks the ones that repeat month after month.
Stage 1. Audit and diagnostics
The first thing to establish is what is holding the site back right now. Not "what's missing" — what's actively in the way. Those are different things, and the second is usually cheaper to fix.
What gets checked here:
- indexing — how many pages are actually in Google's index and what statuses the rest have in the Search Console page indexing report;
- duplicates and cannibalisation — whether existing pages compete for the same query;
- technical state — robots.txt, sitemap, canonical, redirects, response codes, Core Web Vitals;
- current traffic — which queries already generate impressions and where the site sits on page two, within reach;
- competitors in the SERP — what page types Google actually shows for the target queries.
The deliverable is a prioritised list of problems: what to fix first, what second, and what not to touch at all. Not a thirty-page PDF of crawler output, but decisions.
One honest caveat: an audit by itself produces no traffic. It produces a plan. I've seen plenty of cases where a company paid for an audit, received the document and stopped there — money spent, nothing changed. An audit only makes sense together with implementation.
Stage 2. Keyword research and clustering
Next you collect the queries the site could realistically win and group them into clusters — one cluster per page. I've covered what that means in more detail in the piece on keyword research, and the grouping mechanics in the one on keyword clustering.
The part clients tend to underrate: what matters isn't the length of the list but the split. Five thousand queries in a spreadsheet with no grouping isn't a keyword set, it's raw material. The value appears when you can see that these twelve phrasings are one page, while these two apparently similar ones actually carry different intent and must not share a page, or they'll start taking queries from each other.
In Estonia there's a language layer on top: the same service can be crowded in Estonian search and nearly empty in Russian. The keyword set is built per language, because these are different markets with different competition.
The deliverable is a clustered keyword set with the intent of each cluster stated.
Stage 3. Structure and landing pages
The clusters then have to be mapped onto pages: which existing pages fit, which need to be created, which should be merged or removed. This is design work, and the deliverable is a site map at the level of meaning rather than layout. I described that transition from keyword set to pages in the article on building site structure from keyword research.
Typical decisions at this stage:
| Situation | What to do |
|---|---|
| Cluster exists, page doesn't | Create a new landing page |
| Two pages for one cluster | Merge, 301 the second one |
| Page exists, no cluster fits it | Check whether it's needed at all |
| Cluster is too broad | Split into sub-clusters and separate pages |
The deliverable is a "cluster → URL → status" table that drives everything downstream. It doubles as the roadmap for the coming months: you can see how many pages are needed and in what order to build them.
Stage 4. Technical foundation
This is where the cyclical part starts. The technical fixes from the audit get implemented — and this is where projects stall most often, because the fixes have to land on the site itself, and access plus developer time aren't always available.
What usually falls in scope: indexing and sitemap setup, removing duplicates, cleaning up redirects and canonicals, speed and Core Web Vitals, schema.org markup, and hreflang on multilingual sites. What belongs to that foundation and in what order to fix it is covered in the article on technical SEO optimisation.
An important nuance: the technical stage doesn't produce traffic by itself, it lifts the ceiling. If the site isn't indexed, no amount of content helps. If it is indexed properly, another round of image optimisation buys very little. That's why scope here is set by the audit and not by a wish to "do everything".
The deliverable is verifiable changes on the live site. Not recommendations — implementation.
Stage 5. Content and internal linking
Then the pages from the map get filled in: text, tables, answers to real questions, commercial blocks. In parallel you build the internal linking — new pages need links from existing ones, otherwise they stay orphans and Googlebot reaches them last. How to distribute weight between pages is covered in the piece on internal linking for SEO.
This is also where the conversion side gets decided: a page has to rank and produce an enquiry. Traffic without enquiries is a common outcome, and SEO doesn't cure it — work on the page itself does, which is conversion optimisation.
This stage doesn't end. Every month closes part of the cluster map while older pages that have slipped get refreshed.
Stage 6. External signals and measurement
The last block is what happens off the site, plus checking the result. External links, brand mentions, company profiles. Estonia is a small market with few natural link sources, so this stage runs slower here than in large language segments.
Measurement matters more than it seems. Read it in this order:
- impressions in Search Console — the earliest signal, it moves first;
- query coverage — how many distinct phrasings the site is visible for at all;
- average position on the target clusters;
- clicks and traffic;
- enquiries.
If after two or three months even the first item isn't moving, that's a reason to investigate rather than wait longer. On what happens month by month and when to expect what, I have a separate breakdown: how long until SEO results.
What I see in practice
Three patterns from Estonian projects that repeat more than any others.
The stages get reordered. The most common request at the start is "let's begin with the text, we definitely need it". Sometimes that's true, but the only way to know is an audit. On neglected sites, fixing pages already in the index paid off faster than new ones: an existing page already has impressions, it doesn't have to earn trust from zero.
A stage with no artefact isn't a stage. If at the end of the month you can't point to a document or a change on the site, the work isn't checkable. That's the practical test for a client: not "how many hours" but "what shipped".
Clients want to shorten the first three stages and stretch the last three. The logic is understandable — planning doesn't show up in a report, published pages do. In practice, saving on keyword research and structure costs more later, because you end up rebuilding pages that are already live.
If a provider can't answer which stage they're on and what ships by the end of the month, the work is running without a plan. That isn't always dishonesty, but it's always risk.
How a client can use this
Not to supervise every step, but to ask three questions:
- Which stage are we on? The answer should be short and unambiguous.
- What ships at the end of it? A document, a page, a fix — something checkable.
- What's blocking the next one? Often it turns out the blocker isn't the provider but fixes that were never deployed on the site.
If all three have clear answers, the process is healthy even if traffic hasn't moved yet. If they don't, the problem isn't the timeline — it's the missing plan. Reviewing where your site actually stands and which stage to start from is part of how I run SEO.
