How to do keyword research for clients

Keyword research for a client starts with their commercial brief, not with a tool. Get the services they want to sell, the margin on each, the geography they serve and the work they refuse, then research only the demand that maps to those. Expand and score in one batch pass, filter for intent before volume, check winnability against their current authority, and deliver a page plan with a build order rather than a keyword export.

Why client research is a different job from in-house research

An in-house SEO researches one site deeply and lives with the answer for a year. An agency researches a new business every time a contract is signed, then defends the conclusions to someone who has never heard the phrase "search intent" and is measuring you against a monthly invoice.

That changes what good looks like. In-house research can be exhaustive because there is time. Client research has to be fast enough to fit inside a scoping budget, defensible enough to survive a challenge in a meeting, and packaged well enough that an account manager can present it without a strategist sitting beside them. A brilliant analysis that takes three days and arrives as a 4,000-row spreadsheet fails on all three counts.

Step one: take the commercial brief before opening a tool

The single most expensive mistake in client SEO is researching the wrong service line beautifully. Before any expansion, get answers to five things:

• Which services do they actually want more of? Not which they offer. A firm may list twelve services and want volume in three.
• What is the margin on each? Ranking for the low-margin loss leader is a real risk, and clients rarely volunteer this unless asked directly.
• What geography counts? National, a metro, a service radius. This decides whether local modifiers belong in the research at all.
• Which leads do they turn away? The queries behind unwanted work should be excluded on purpose, not accidentally ranked for.
• Who do they lose deals to? Every named competitor becomes a seed, and usually the highest-intent one on the list.

Twenty minutes on those five questions saves a month of building pages the client cannot monetize.

Step two: seed from services, not from the industry

Industry-level seeds produce impressive volumes and useless pages. "Commercial real estate" has enormous demand and almost none of it belongs to a fifteen-person brokerage in Denver. Service-level seeds produce the terms a buyer types the week they are ready to hire: the service name, the problem it fixes, the qualifier that describes their situation, and the competitor they are comparing against.

Build the seed list from four sources: the client service pages, the words their sales team hears on calls, the competitor set from the brief, and the questions their support inbox answers repeatedly. That last source is consistently the best and consistently ignored.

Step three: batch the mechanical work across accounts

Expansion, volume pulls and first-pass difficulty scoring are identical operations regardless of which logo they belong to. Run them together. The analyst hours you save there are the hours that let you take another retainer without another hire, which is the actual economics of an agency.

What must never be batched is the conclusion. Shared workflow, separate judgment. The moment one client template starts deciding another client strategy, you are selling the same plan twice and both clients can tell. Run the enrichment in one pass with bulk keyword research, then review each account alone against its own brief.

Step four: filter on intent, then on winnability

Sort every surviving phrase by what the searcher wants, not by how many of them there are. For most service businesses the hierarchy is simple: people naming the service with a buying modifier, people naming a competitor, people describing the problem urgently, then everyone else. Research-only queries that will never convert for this client get cut, however large the number next to them.

Then check what actually ranks for the survivors. A term the client cannot realistically beat within the contract term belongs in a later phase with an honest timeline attached, not in month one where it will make you look slow. Being explicit about this in the deliverable buys enormous credibility, because clients have usually been promised the head term by a previous agency and watched it not happen.

Step five: cluster into pages, and name the page type

Group everything a single page could satisfy, then label each cluster with what it becomes: a service page, a location page, a comparison page, or an article. That labeling step is what turns research into something a client can quote, schedule and approve.

It also prevents the most common structural failure in client sites, where four near-identical service pages compete with each other and none of them ranks. If two clusters would need the same page to satisfy them, they were one cluster. The keyword clustering tool handles the grouping, and the page-type call is where your expertise earns the retainer.

Step six: package it as a decision, not a data dump

A client-ready deliverable leads with the build order and the reasoning, and puts the keyword tables behind that. Five sections cover it: priority clusters with volume and difficulty, the page assigned to each, what the site already ranks for, the sequenced build order, and which competitors own which clusters today.

Raw keyword exports do not survive a client meeting. Nobody approves a spreadsheet. They approve "these four pages, in this order, because your two closest competitors own this space and this is the one you can take first". The full agency workflow, including what each account type needs, is laid out on the keyword research for agencies page.

Adapting the process per account type

The workflow stays constant; the surviving keyword shapes change. Local service clients live on service plus city pairs and the map pack, which makes local keyword research the core of the plan. Ecommerce clients need category terms separated cleanly from product and research terms so the page types do not fight. B2B and SaaS clients have small volumes where comparison, alternative and pricing queries carry the entire pipeline. New sites and rebuilds are difficulty-dominated, and anything above the client authority ceiling belongs in a later quarter.

Reporting the research after the plan ships

Research is not finished when the plan is approved, because the client will ask what changed. Report against the clusters you defined rather than a generic traffic line, so movement is legible: this cluster went from no coverage to position 12, this one from 14 to 6, this page now ranks for nine of its eleven target phrases. Clients who can see cluster-level progress renew, because the work maps to something they recognize.

The same clustered plan also does double duty for distribution. Once an article exists for a cluster, the angles inside it are what you reshape into the client social and newsletter versions, which is usually the cheapest add-on an agency can attach to an SEO retainer.

The short version

Brief first, tools second. Seed from services and competitors, not from the industry. Batch the mechanical work and never batch the judgment. Score intent before volume and winnability before ambition. Cluster into named page types, and hand over a build order with reasons attached. Do that consistently and the research stops being the slow part of onboarding a client and becomes the reason they trust the rest of the engagement. Try it on a live account with the free Keywordpilot demo: one client seed returns the demand around it, scored for intent and difficulty and already grouped into the pages worth proposing.