Most keyword research guides are either 500 words of hand-waving or 8,000 words of rehashed theory that never get specific. This one is different. I’m going to walk you through the entire process using a single example niche - home coffee brewing - so you can see exactly what happens at every step, with real numbers.

If you’ve been doing keyword research by typing random phrases into a tool and hoping for the best, this is the fix.

What keyword research actually is (and isn’t)

Keyword research is the process of finding the specific queries your target audience types into Google, then figuring out which ones are worth pursuing. That’s it. It’s not “brainstorming content ideas” or “understanding your audience’s pain points” - those are useful things, but they’re inputs to keyword research, not keyword research itself.

The output of good keyword research is a prioritized list of keywords grouped into clusters, each with volume data, difficulty scores, and intent labels. That list drives your entire content strategy. Skip this step or do it badly and you’ll spend months writing articles nobody searches for.

Start with seed keywords

Every keyword research project starts with seeds - broad terms that define your topic space. You want three to seven of them. Not fifty. Not one. Three to seven gives you enough coverage without drowning in data later.

For our home coffee brewing example, here are five seeds:

  • “home coffee brewing”
  • “pour over coffee”
  • “french press coffee”
  • “coffee grinder”
  • “espresso at home”

Where do seeds come from? Your own head, mostly. Think about the core topics your site covers. If you’re stuck, check what your top three competitors rank for and grab the broadest terms. Google Search Console is another goldmine - look at queries where you’re getting impressions but not clicks.

One mistake I see constantly: starting with seeds that are too narrow. “Best burr grinder under $50 for pour over” is not a seed. That’s a finished keyword. Seeds should be two to three words max.

Expand your seed list into thousands of keywords

Take each seed and run it through your keyword tool of choice - Ahrefs, SEMrush, Google Keyword Planner, whatever you’ve got access to. You want every variation, question, and related term the tool can surface.

Here’s what our coffee seeds generate:

SeedKeywords returned
home coffee brewing2,140
pour over coffee4,870
french press coffee3,620
coffee grinder8,310
espresso at home1,950
Total20,890

That’s nearly 21,000 raw keywords. There’s massive overlap between these lists - “best coffee grinder for pour over” shows up in at least two of them. That’s fine. You’ll deduplicate later.

Expansion methods beyond your main tool

Your keyword tool gives you the bulk, but there are three expansion methods worth adding:

Google Autocomplete. Type each seed into Google and note the suggestions. These are high-frequency real searches that tools sometimes miss. For “pour over coffee,” autocomplete gives you “pour over coffee ratio,” “pour over coffee maker,” and “pour over coffee vs french press.” All useful.

People Also Ask. Click into a few top results for each seed and harvest the PAA boxes. These are questions Google explicitly associates with your topic. For “french press coffee,” PAA surfaces “how long should french press coffee steep” and “is french press coffee bad for cholesterol.” The second one is interesting - it’s a health angle you might not have considered.

Reddit and forums. Search Reddit for your seeds and skim the threads. You’ll find language your audience actually uses, which is often different from what keyword tools suggest. Coffee Reddit talks about “dialing in” their grinder, “channelling” in espresso, and “bloom time” for pour over. These are real terms real people search for.

After expansion, our coffee example sits at roughly 23,400 keywords including the extras from autocomplete and forums.

Filter ruthlessly

23,400 keywords is not a content plan. It’s noise. Filtering is where you turn noise into signal, and the order you filter matters.

Filter 1: Remove duplicates and near-duplicates

Exact duplicates go first. Then catch near-duplicates - “how to brew coffee at home” and “how to make coffee at home” are the same intent. Deduplication typically removes 15-25% of your list. Our coffee list drops from 23,400 to about 18,600.

Filter 2: Keyword difficulty

Set a KD ceiling based on your site’s authority:

  • New site (DR 0-15): KD under 15
  • Growing site (DR 15-35): KD under 30
  • Established site (DR 35+): KD under 45

Let’s say we’re a growing coffee blog at DR 22. We set KD under 30. This cuts our list from 18,600 to around 5,400 keywords. That’s a 71% reduction in one filter. Good. Most of those eliminated keywords were terms like “best coffee maker” (KD 67) and “espresso machine” (KD 72) - big-volume head terms that we have zero chance of ranking for right now.

For a closer look on finding terms you can actually compete for, see how to find low competition keywords.

Filter 3: Search volume floor

Remove anything under 50 monthly searches. Those keywords aren’t worth dedicating a page to - you’ll pick them up as secondary terms on other pages. This trims another 30-40%. We’re down to about 3,400 keywords.

Filter 4: Search intent alignment

This is manual but critical. Scan the remaining list and remove keywords that don’t match what your site delivers. Common mismatches:

  • Product searches when you’re informational. “Buy hario v60” implies a purchase. Unless you sell coffee gear, skip it.
  • Local intent. “Coffee roasters near me” - you’re not a local directory.
  • Brand-specific queries. “Nespresso pod recycling” belongs to Nespresso, not you.
  • Off-topic tangents. “Coffee table dimensions” came in through the “coffee” seed. Obviously irrelevant.

After intent filtering, our list is around 2,100 keywords. From 23,400 to 2,100 - that’s a 91% reduction, which is normal.

Classify intent for what remains

Every keyword on your filtered list needs an intent tag: informational, commercial, or transactional. This determines what type of page you build.

  • Informational: “how to brew pour over coffee,” “french press coffee ratio,” “what grind size for aeropress.” These get blog posts or guides.
  • Commercial investigation: “best pour over coffee maker,” “baratza encore vs fellow ode,” “chemex vs v60.” These get comparison or review content.
  • Transactional: “buy chemex coffee maker,” “hario v60 price.” These get product or landing pages.

For our coffee example, the 2,100 keywords split roughly:

  • Informational: 1,260 (60%)
  • Commercial: 630 (30%)
  • Transactional: 210 (10%)

The informational bucket is where most of your content calendar will live, especially for a newer site. Commercial keywords are your money pages. Transactional keywords only matter if you actually sell something.

Cluster keywords into page-level groups

Here’s where most people stop - they have a filtered list, so they start writing. Bad move. You need to cluster first.

Clustering means grouping keywords that should target the same page. “French press brew time,” “how long to steep french press,” and “french press steeping time” are three keywords, but they’re one article. If you write three separate posts, you’ll cannibalize yourself and rank for none of them.

Our 2,100 keywords cluster down into roughly 280 topic groups. That’s 280 potential pages, not 2,100. Some clusters are fat - 15-20 keywords pointing at one pillar page. Some are thin - two or three closely related terms.

Here’s what a real cluster looks like from our coffee data:

Cluster: Pour Over Coffee Ratio

  • pour over coffee ratio (1,900/mo, KD 22)
  • pour over coffee to water ratio (880/mo, KD 18)
  • how much coffee for pour over (720/mo, KD 14)
  • pour over ratio grams (260/mo, KD 11)
  • v60 coffee ratio (410/mo, KD 19)
  • chemex ratio (390/mo, KD 16)

Total cluster volume: 4,560/mo. The primary keyword alone is 1,900, but the cluster captures over twice that. One well-structured page targets all six terms.

Clustering by hand works for 200 keywords. At 2,100 it’s brutal. Keyword clustering tools handle this in minutes instead of hours, grouping by semantic similarity and giving you a hierarchy - pillar topics, subtopics, and individual article targets.

Prioritize: what to write first

You have 280 clusters. You’re not writing 280 articles next month. You need to pick the first 10-15 and sequence them intelligently.

The prioritization formula

Score each cluster on three factors:

  1. Opportunity = total cluster volume / average cluster KD. Higher is better. A cluster with 4,560 volume and average KD 17 scores 268. A cluster with 800 volume and average KD 28 scores 29. The first one is 9x more attractive.
  2. Topical fit. Does this cluster sit at the core of your site’s expertise, or is it peripheral? Core topics get a multiplier. For our coffee site, “pour over coffee ratio” is dead center. “Coffee and cholesterol” is a tangent.
  3. Funnel position. Mix informational and commercial clusters. A common ratio is 70% informational, 30% commercial for the first batch. The informational content builds topical authority. The commercial content captures people closer to a purchase decision.

Our coffee example: the first 10

After scoring, here are the top 10 clusters to write first:

  1. Pour over coffee ratio (4,560/mo combined, avg KD 17)
  2. French press brew guide (3,890/mo, avg KD 14)
  3. Best pour over coffee maker (3,200/mo, avg KD 26)
  4. Coffee grind size chart (2,870/mo, avg KD 19)
  5. Aeropress brewing method (2,140/mo, avg KD 12)
  6. How to make cold brew at home (4,100/mo, avg KD 24)
  7. Burr grinder vs blade grinder (1,680/mo, avg KD 21)
  8. Best coffee grinder for pour over (1,540/mo, avg KD 23)
  9. French press vs pour over (1,920/mo, avg KD 18)
  10. How to store coffee beans (1,350/mo, avg KD 11)

Total addressable volume from just 10 articles: 27,250 monthly searches. That’s real traffic if you execute well.

Notice the mix - seven informational guides and three commercial comparisons. The informational pieces build authority in Google’s eyes. The commercial pieces target people with purchase intent. Together, they create a topical cluster that reinforces itself.

Build your keyword research into a content roadmap

Those 10 clusters are your first sprint. But you’ve got 270 more waiting. The smart move is to build a content roadmap that sequences everything across three to six months, grouped by topic cluster.

For the coffee site, the roadmap might look like:

  • Month 1: Pour over cluster (ratio guide, best maker, v60 vs chemex)
  • Month 2: French press cluster (brew guide, ratio, grind size, vs pour over)
  • Month 3: Coffee grinder cluster (burr vs blade, best for pour over, grind size chart)
  • Month 4: Cold brew cluster (how to make, ratio, best cold brew maker)

Each month builds out a complete topic cluster before moving to the next. This signals topical depth to Google faster than randomly publishing across categories.

Keyword research for ongoing content

Keyword research isn’t a one-time event. The landscape shifts. New keywords appear as trends emerge. Competitors publish content that changes the difficulty picture. Google updates reshuffle who ranks for what.

Run a refresh every quarter. Re-export your seeds, re-filter, and check for new clusters you missed. Also look for keywords where your existing content has started ranking on page two - those are upgrade opportunities where a content refresh could push you onto page one.

Track your clusters in a spreadsheet or tool with status columns: researched, outlined, drafted, published, ranking. This keeps you honest about progress and prevents the common trap of endlessly researching without publishing.

The full process, summarized

Here’s the keyword research workflow from start to finish:

  1. Pick 3-7 seed keywords that define your topic space
  2. Expand using your keyword tool, autocomplete, PAA, and forums - aim for 10,000-25,000 raw keywords
  3. Filter by removing duplicates, setting KD ceiling, volume floor, and intent mismatches - expect 85-95% reduction
  4. Classify intent as informational, commercial, or transactional
  5. Cluster related keywords into page-level groups
  6. Prioritize by opportunity score, topical fit, and funnel position
  7. Sequence into a monthly content roadmap
  8. Refresh quarterly to catch new opportunities

For our coffee example, that process took 23,400 raw keywords down to 280 clusters, prioritized into a six-month content plan starting with 10 articles targeting 27,250 monthly searches.

That’s keyword research. No mysticism, no secret sauce - just a systematic process that turns a mess of data into a publishing plan you can actually execute. The sites that rank aren’t doing anything magical. They’re doing this, consistently, and then writing content that deserves to rank for the terms they chose.