Skip to main content

Building keyword clusters

A keyword cluster is a tight group of keywords that share search intent. Clusters are how serious paid search operators organize campaigns. This article covers why clusters matter, how to build them manually, and where AI can speed the work.

Why clusters​

The naive approach: dump all your keywords into one campaign with one set of ads.

The serious approach: group keywords by intent, build a separate campaign per cluster, and write ad copy specific to each cluster's intent.

Three reasons the serious approach wins:

One ad cannot speak to multiple intents​

Someone who searches for "alternative to Salesforce" wants to switch. Someone who searches for "CRM software" is exploring. An ad that says "Switch from Salesforce in 5 minutes" works for the first searcher. It confuses the second searcher. An ad that says "Compare top CRM platforms" works for the second searcher. It feels too broad for the first searcher.

You need different ads for different intents. That is only possible when you separate your campaigns by intent.

Mixed campaigns muddle your bids​

High intent keywords are worth more per click than low intent keywords. A click from "alternative to Salesforce" might convert at 5 percent. A click from "CRM software" might convert at 1 percent. The high intent keyword can sustain a higher bid.

When high intent and low intent keywords share a campaign, you must set one bid for everything. You either overbid on the low intent terms and waste money, or underbid on the high intent terms and lose conversions.

Reports are useless without clean separation​

When you cannot separate intent, you cannot optimize. "This campaign has CTR of 3 percent and CPA of $250" is useless if half the campaign is one intent and half is another. The high intent half might be at 6 percent CTR and $80 CPA. The low intent half might be at 1 percent and $400 CPA. You never know.

Tight clusters give you clean reporting for each intent. That is how you decide what to scale and what to kill.

What makes a good cluster​

Three properties:

Single, namable intent​

You should be able to write the cluster's intent in one sentence: "People looking to switch from Salesforce", "People exploring CRM options", "People searching for sales engagement features".

If you cannot write one sentence, the cluster is too broad.

Similar volumes​

Keywords within a cluster should have roughly comparable search volumes. A cluster where one keyword has 100,000 monthly searches and another has 100 is suspicious. The first keyword will dominate. The second will not get clicks.

Same ad copy direction​

If different keywords in the cluster need meaningfully different ads, split the cluster.

How to build clusters manually​

Follow these steps if you build clusters by hand.

Step 1: dump every keyword that might be relevant​

Do not filter yet. Include every keyword that might be relevant.

Step 2: group by intent​

Sort the list. Group by what the searcher is trying to do. Common intent buckets for B2B SaaS:

  • A switch from a specific competitor.
  • A generic switch from any competitor.
  • A search for a category leader.
  • A search for category alternatives at a price point.
  • A specific job to solve.
  • A comparison of features.
  • A broad look at the category.
  • A search for free or freemium options.

A keyword usually fits one of these categories. If a keyword fits two categories, decide which one is dominant.

Step 3: prune within each cluster​

For each cluster:

  • Drop keywords that do not fit the cluster's stated intent.
  • Drop keywords with too low volume to matter.
  • Drop keywords that are duplicates with different word order.

Each cluster ends with 10 to 30 high quality keywords.

Step 4: identify negatives​

Some queries look like they should match your cluster, but they will not. "Free CRM" might overlap with your "Salesforce alternative" cluster, but the searcher wants something free, not a switch. Add it as a negative keyword.

This step often surfaces a few clusters you did not realize you needed, for example a "free tier seekers" cluster. You can target that cluster separately, or ignore it entirely.

Step 5: write the cluster's name​

Name each cluster for what it represents, for example "Alt to Salesforce", "Switching CRMs (generic)", or "CRM for sales engineering". A good name is the working title of the campaign you will build for that cluster.

How to build clusters with AI​

When you cluster keywords by hand, it takes hours, and the returns shrink past a certain point. AI tools cluster keywords by intent in minutes. Hero Marketer's campaign wizard does this:

  1. You provide one or two seed keywords.
  2. The wizard queries Google's keyword planner for related terms.
  3. It clusters the keywords by semantic similarity and intent, and weighs them against your product context.
  4. You see the clusters with metrics and pick the one that matches your campaign.

The AI version is faster than doing this by hand. It is also more objective than most humans. Humans tend to over cluster or under cluster. When they over cluster, they split things that should stay together. When they under cluster, they lump together things that should stay apart.

The AI version has weaknesses:

  • It does not know your competitive landscape beyond what is in your product context. Specific competitor terms might not surface unless your product context mentions competitors.
  • It does not know your strategic priorities. "Free tier seekers" might cluster cleanly. But that cluster might be irrelevant to your goals.

In most situations, AI is the right starting point. You refine the clusters from there.

See Choose keyword clusters and What is a keyword cluster for the Hero Marketer specific approach.

How many clusters​

For a starting B2B SaaS Google Ads program, 3 to 5 clusters cover most of the addressable demand.

Do not build more clusters than you need. Five tight, well managed campaigns produce more measurable signal than fifteen sprawling campaigns with overlapping keywords.

Cluster level reporting​

Once you have clean clusters, your reports become much more actionable:

  • "Cluster A (alternative to Salesforce) is converting at $80 CPA."
  • "Cluster B (sales engineering CRM) is at $250 CPA."
  • "Cluster C (general CRM exploration) is at $600 CPA."

Now the decision is obvious. Scale A. Iterate on B. Kill C, or move it to a content strategy that nurtures top of funnel interest without bidding.

That clarity is what cluster based campaign organization gives you. Even if setup takes a couple of hours, it pays for itself many times over.

Next​