If you get this wrong, nothing downstream matters. The best-written article in the world cannot rank for a keyword nobody is searching for.
Most agencies skip the strategy phase entirely. They ask the client what topics they care about, throw those into ChatGPT to generate titles, and start writing. Three months later the client is asking why the blog has 40 new posts and zero new traffic.
The reason is simple. Those 40 posts targeted keywords that either had no search volume, zero commercial intent, or impossible competition. They were never going to rank.
Every client you take on needs a keyword map before you write a single post. The structure is always the same:
| Layer | What it is | Count per client |
|---|---|---|
| Pillars | The 3-5 broad topics that define what the client sells or teaches | 3-5 |
| Clusters | Mid-volume subtopics under each pillar (e.g., "types of X", "X for Y audience") | 10-20 per pillar |
| Posts | Specific long-tail keywords, one per article | 5-20 per cluster |
A well-built map gives you 500-2,000 target keywords ranked by priority. That is your editorial calendar for the next 12-24 months.
Not all keywords are worth writing about. The ones that move the needle have three traits:
The formula most SEO teams use is some version of (Volume × Intent fit) / Difficulty = Priority score. Rank your map by that score and work the top 20% first.
| Vertical | What "good intent" looks like |
|---|---|
| Health-tech | Symptom queries, condition names, "vs" comparisons between treatments or brands |
| Local service (law, dental, real estate) | Geo-modified service terms: "Austin estate planning attorney," "Dallas cosmetic dentist" |
| DTC ecommerce | Category + qualifier: "best running shoes for flat feet," "men's joggers under $50" |
| B2B SaaS | Pain point + solution: "how to manage remote teams," "salesforce alternatives" |
| CPG / regulated | Educational + safety: "is X safe for dogs," "benefits of X for anxiety" |
Here is what the research process actually looks like when you run it manually:
End to end, that is 3-6 hours of hands-on work per client for the initial map. Then 1-2 hours per week to refresh as new opportunities emerge, clients shift focus, or Google updates change the landscape.
At one or two clients this workflow is fine. Painful but doable.
At five clients you are looking at 15-30 hours of keyword research every month. Your Ahrefs subscription starts hitting its credit limits mid-month. You start using "good enough" research instead of thorough research because you do not have the time. Your Google Sheets get messier. A junior team member takes over and the quality of the maps drops. Three months in, you realize two of your clients are targeting overlapping keywords and competing with each other.
The specific trap: most agencies do beautiful keyword strategy for their first one or two content clients, then quietly stop doing it for the rest. They copy structure from old research, ship articles on "close enough" keywords, and wonder why the later clients never hit the same traffic curves as the early ones.
A proper production system treats keyword strategy as the input, not a separate project. You give it the client's URL and a few pillar topics. It pulls the SERP data, runs the intent analysis, scores opportunity, and outputs the 3-layer map in a consistent format every time. Same framework, same scoring, same output format across every client you manage.
When we look at Step 2 (production) next, keep in mind that every shortcut you take in keyword strategy compounds. A weak map produces weak articles even if your writing process is flawless.