Keyword research has a reputation problem: it sounds like a dark art performed inside expensive tools, so most bloggers either skip it (writing whatever occurs to them, then wondering why nobody arrives) or drown in it (spreadsheets of ten thousand keywords, none ever written about). The truth sits in the middle. Keyword research is customer research with a search-volume column — a repeatable two-to-three-hour workflow that turns “what should I write?” into a ranked queue of topics real people are already asking about. Here’s that workflow, step by step, with the judgment calls spelled out.
Step 1: Harvest Seeds From Real Conversations (30 minutes)
Before any tool, collect the raw language of your audience. Four sources, ten minutes each:
- Customer conversations: sales calls, support tickets, onboarding questions. The exact phrases customers use — “why do my posts get no reach” — are seed keywords no tool will surface as neatly.
- Communities: the subreddits, Facebook groups, and forums where your audience gathers. Recurring questions are pre-validated demand.
- Competitor tables of contents: what your three closest competitors blog about — noting both what performs (their most-linked posts) and what’s conspicuously missing.
- Your own data: if the site has any history, search console’s query report shows what you already almost rank for — the cheapest wins available anywhere.
Output: 20–40 seed phrases in a list. Messy is fine; the next step expands and disciplines them.
Step 2: Expand the Seeds (30 minutes)
Feed each promising seed into expansion sources and collect the variants: search autocomplete (type the seed plus each letter of the alphabet — crude and remarkably productive), the “People Also Ask” and “Related searches” boxes, and a keyword tool if you have one (free tiers of most tools suffice at this stage). You’re gathering three kinds of variants: question forms (“how often should I post on Instagram”), modifier forms (“Instagram posting schedule for small business”), and comparison forms (“Instagram vs TikTok for business”). A hundred-plus candidates from your best seeds is normal. Dump everything into one sheet with a column for the seed it came from.
Step 3: Score on the Three Axes (45 minutes)
Now discipline the list. For each candidate, estimate three things — precision is neither possible nor required; high/medium/low is enough.
Volume: is anyone searching?
Tool estimates are directionally useful and individually unreliable — treat “10 vs 1,000” as signal and “720 vs 880” as noise. Don’t discard low-volume keywords reflexively: a query with 50 monthly searches from people about to buy outearns one with 5,000 from students. And long-tail queries in aggregate usually outnumber the head terms your competitors fight over.
Difficulty: can you realistically rank?
Ignore abstract difficulty scores; look at the actual page one for the query. Green flags: forum threads, thin listicles, posts several years stale, generic content from sites outside the niche. Red flags: every result is a major publication or a dedicated, recent, comprehensive guide. A useful heuristic for newer sites — if you can’t imagine your article being honestly better than at least three of the current top ten, pick a narrower variant of the query and win that instead.
Business value: would ranking make you money?
The axis everyone underweights. Score each keyword by what the searcher is doing: buying-intent queries (comparisons, “best X for Y”, pricing, alternatives) convert even at tiny volume; problem-intent queries (how do I fix/do X) build audience and feed the email list; curiosity-intent queries (what is X) bring traffic that rarely converts. A blog that only chases volume fills with curiosity traffic and stays broke.
Step 4: Check Intent Before Committing (15 minutes)
For each keyword you’re about to greenlight, search it and look at what page one actually is — because the results define what searchers want, and you must match the format to compete. If the query returns listicles, a heartfelt essay won’t rank regardless of quality. If it returns videos and tools, a text post is fighting the format. And if two keywords on your list return essentially the same results, they are one article, not two — splitting them creates two weak pages competing with each other. Merge them under the stronger phrasing.
Step 5: Build the Queue, Cluster-First (30 minutes)
Don’t rank the survivors as one flat list — group them into topic clusters (a broad pillar topic plus its narrow questions), then sequence with three rules. Finish clusters before starting new ones: topical depth is what modern search rewards; fifteen scattered posts build nothing, fifteen interlinked ones build authority. Front-load business value: a couple of buying-intent pieces early make the blog pay for itself while the long game matures. Interleave difficulty: pair each ambitious keyword with two easy wins so the queue produces visible progress — early rankings, even small ones, are what keep a content habit funded and alive.
The output artifact is simple: a sheet with columns for keyword, cluster, intent, volume tier, difficulty verdict, and status. That sheet is your editorial calendar’s source of truth for the next quarter.
Step 6: Close the Loop Monthly (20 minutes, forever)
Keyword research isn’t an annual ceremony; it’s a monthly glance at two reports. In search console: which queries are you getting impressions for that you haven’t deliberately targeted? Those are keywords the market assigned you — often your easiest next articles. Which posts sit at positions 5–15? Those are refresh candidates where a better title, updated content, and a few internal links buy a page-one jump. Feed both findings back into the queue. After a few cycles, your own performance data becomes a better keyword tool than any subscription.
The Mistakes That Waste the Most Time
- Tool worship: treating volume estimates as facts and difficulty scores as verdicts, instead of checking real page-one results with your own eyes.
- Volume chasing: a library of high-traffic, zero-intent posts is a vanity asset.
- One-keyword-one-post literalism: a good article targets one primary keyword and naturally absorbs dozens of variants; writing separate posts for every phrasing builds a cannibalization farm.
- Research as procrastination: the fourth hour of keyword research returns less than the first hour of writing. Timebox it, ship the queue, adjust monthly.
Run this workflow once a quarter and the blank-page problem disappears: every article starts with proof someone wants it, a realistic shot at ranking, and a reason the business should care. That’s the entire point — not perfect data, but a queue you can trust enough to execute without second-guessing every Tuesday. (For how the queue becomes briefs, drafts, distribution, and compounding traffic, see our complete content marketing playbook.)
