How to Do Keyword Research

Good keyword research is not a list of phrases with search volumes next to them. It is a ranked plan that ties every keyword to the money it can make you. Here is the method, from raw data to a prioritized target list.

Start with Google Keyword Planner

Keyword Planner is free, and it is where you get your base dataset, the raw list of real keywords people type around your service. Work it in two passes:

1. Start from the client's primary category, the single best label for what they do (“bathroom remodeling,” “stump removal,” “auto body shop”). Use it as your root keyword to surface the wider set of related searches.
2. Read that first pass to spot the keyword clusters and subclusters around the service. Take the roughly ten biggest and run each one back through Keyword Planner as its own seed. That second pass is what surfaces the full keyword universe, not just the obvious head terms.

Download the combined results, then cut the list down: remove anything irrelevant to the client's actual business, and set aside the pure top-of-funnel and low-intent terms. You are building a list ranked by money, so you want the searches with buying intent. You can still cover the informational topics later for topical authority. They just should not dilute the ranking you are about to build.

One catch with the data itself: Keyword Planner reports search volume in wide ranges (1 to 10, 10 to 100, 100 to 1,000) instead of exact figures. When one keyword could be 11 searches a month or 990, you cannot build a reliable model on top of it. It is great for discovering keywords, weak for measuring them.

Get sharper numbers with DataForSEO

To measure them, upgrade the data. A source like the DataForSEO API returns a single modeled monthly figure for each term instead of Keyword Planner's wide bucket, so you can rank keywords against each other far more precisely. It is still a modeled average built on the same Google Ads data, not a ground-truth count, just far more granular. The workflow is simple: use Keyword Planner to find the keywords, then swap in the tighter numbers to rank them. Now your list is built on sharper data instead of vague ranges, which matters the moment you start attaching dollars to it.

Because you will be doing this across hundreds of keywords, script it. A short Python script running in a free Google Colab notebook can send your whole keyword list to the DataForSEO API and return a volume figure for each term in one pass, which you drop straight back into your sheet, far faster than looking terms up one at a time.

Map every keyword to gross profit

This is the step that turns a keyword list into a strategy: put a dollar figure on each keyword based on what it would earn if you ranked in the top three spots. You need four inputs: the industry-average top-three click-through rate, your website's conversion rate (clicks to leads), your close rate (leads to customers), and your average gross profit per job. Then, per keyword:

monthly volume × top-3 CTR × website conversion rate × close rate × gross profit per job = the profit that keyword earns you at the top of Google

Rank on gross profit, not revenue. It is the money that actually reaches your pocket. A keyword with modest volume but high buying intent can easily out-earn a high-volume informational term that rarely converts. Trim your sheet to the columns that feed this calculation, sort by the result, and the list orders itself by money instead of traffic, the same logic behind forecasting what SEO is worth before you spend a dollar.

Group by intent before you build

A ranked list still isn't a site plan. Group the keywords by intent into clusters, because Google rewrites and ranks intent, not individual phrases. Pricing searches, informational searches, and ready-to-buy searches each want their own page. Send them all to one and you satisfy none, and you compete against yourself. From there, mapping one page to each intent gives you the structure of the whole site.

Refine it with real data after launch

Your first model runs on industry averages, and averages are only a starting point. Once the site is live and tracking conversions in Google Analytics 4, and you have close-rate data coming out of your CRM, feed the real numbers back in. Your actual top-three click-through rate, website conversion rate, and close rate replace the benchmarks, and the profit ranking sharpens, which in turn tells you where to adjust your site structure and which keywords to double down on. Keyword research is not a one-time task; it is a loop that tightens every time you get more real data.

Where AI helps, and where it shouldn't

AI is a huge accelerator for the mechanical parts: expanding a seed list, sorting hundreds of keywords into clusters, drafting first-pass page outlines. Let it do those in seconds. What you should keep in your own hands is judgment: deciding which intents are actually worth targeting, setting the conversion and job-value assumptions behind your revenue estimates, and making the final call on the page map. Use AI to iterate quickly; do the decisions that cost money by hand.