Keyword research from your sales calls, not just a tool

Keyword tools answer one question well: how many times did people type this phrase? They are poor at the questions that decide whether a page works. What did the buyer actually mean? What were they afraid of? What did they call the thing before they learned its name?

Your sales team hears those answers every week. Ten recent conversations contain a better starting list than a tool export, because the words are the buyer's own and they are attached to real money. This is how we turn calls into a keyword plan.

Why calls beat a blank tool export

A tool shows you the demand that others already measured. It shows nothing about intent quality, nothing about the objection that kills the deal, and very little about phrases that are too new or too specific to have volume.

Calls and notes give you:

• Buyer vocabulary. A customer says "someone to take over our IT" where your site says "managed services provider". The gap is your ranking problem.
• Objections. "Will we be locked in?" "Do you work in French?" "What if our developer left?" Every one of these is a page or a section.
• Comparison questions. "HubSpot or Salesforce?" "Next.js or WordPress?" These are exactly the shape of query where a clear, structured answer gets cited.
• Qualification signals. What the people who closed said, versus those who went quiet.

This is not instead of tools. It is the order: language first, then volume and competition second.

Step 1: collect the raw material

Pull from the places you already have:

• Notes from the last 10–20 discovery calls
• The first message or "how did you hear about us" field in your CRM
• Support or onboarding emails from new clients
• The questions prospects asked in proposals and RFP clarifications
• Lost-deal notes, which are often the most honest

If you record calls, check that you have the consent and retention practices your counsel is comfortable with. Quebec has specific rules on collecting and keeping personal information, and we are not your lawyers. Notes are enough to start. Strip names and company details before sharing material with anyone outside the team.

Step 2: extract phrases, questions, and objections

Read through and highlight three kinds of text.

Naming phrases. How do buyers describe the problem and the solution? Note exact wording, including French and English variants if your market is bilingual.

Questions. Anything beginning with "how", "can you", "what happens if", "do you". Write them as full questions.

Objections and fears. The thing that almost stopped the deal. These become the headings of your best pages.

Put them in a sheet with columns for the phrase, the language, how many conversations it appeared in, and the stage of the deal where it came up. Frequency matters more than elegance. A phrase heard in six of fifteen calls outranks a clever one heard once.

Step 3: group by the job, not the keyword

Sort the list by what the buyer is trying to do. Typical groups:

• Understand the problem ("why is our site slow after the redesign")
• Compare options ("HubSpot vs Salesforce for a small team")
• Judge a vendor ("how long does a migration take", "do you work in French")
• Prepare to buy ("what should be in a brief", "how much does it cost")

Each group maps to a page type. Comparisons become comparison pages. Vendor-judgement questions often belong on service pages or in a short FAQ. Problem questions become Insights articles. Use the Insights cluster logic from our SEO and GEO piece: one hub and sibling articles, not twenty orphan posts.

Step 4: check the phrases against the tools, honestly

Now bring in a keyword tool or Search Console data. For each cluster, ask:

• Does the phrase, or a close variant, have any measurable demand?
• What kind of page ranks today: a guide, a product page, a forum thread, a government site?
• Is that a page type we can reasonably produce?

The last question matters. If the first page of results is all law firms and government pages, a studio blog post is unlikely to be the right answer. If it is comparison tables from studios and documentation, we have a shot. We look at the page type before we write a word.

Treat tool volumes as rough. They are estimates and they undercount specific, long queries. A zero in the tool is not proof of zero; a big number is not proof of qualified buyers. Neither one is a forecast, and we do not publish volume estimates as promises.

Step 5: write the answer first

For each chosen phrase, write the page so the first paragraph answers the question in a complete, quotable sentence. Then the details. Then the next step.

A claim-complete paragraph names who it applies to, what to do, and what the limits are. "Discovery is a call during Montreal hours, weekdays 9–17 ET, followed by a written proposal" is a paragraph an answer engine can quote. "We offer tailored solutions" is not. This is the same approach we describe for GEO in SEO and GEO in Montreal without a fake map pin.

Use the buyer's phrasing in headings and the first sentence, then your own terminology once you have defined it. If buyers say "someone to run our IT", use that and explain the term "managed service". Don't make them learn your vocabulary before they get an answer.

Step 6: close the loop through the CRM

A page built from calls should be measurable against calls. That is the point of the work, and the reason keyword research and CRM design are the same conversation. We add:

• A hidden source field that records the landing page when a form is submitted
• A "heard about us" question in plain language
• A recurring review, for example monthly, of which pages started conversations

When this is in place, you can see which phrases from calls turned into pages that turned into calls. That feeds the next round of research. It is the argument in site, CRM, and search as one system, and the reason we care about a CRM built around how you sell: the field names should match the words your team actually uses.

A short example

Imagine a Montreal accounting firm. In twelve discovery calls, five prospects ask some version of "can you work with our existing bookkeeper?" Nobody searched that exact phrase in a tool. But the question is the objection that delays decisions.

A page titled "Working with your existing bookkeeper" answers it in the first paragraph, lists how the handoff works, who does what, and the first thirty days. It links to the relevant service and to contact. It is not a keyword-volume play. It is a page that closes deals and sometimes ranks for a long-tail query nobody was tracking.

What we will not do

We will not invent search volumes or promise rankings. We will not scrape and rewrite competitors, and we will not stuff a phrase into a page to look relevant. We do not use rented lists to find prospects either; the point of this work is to understand people who asked you something.

If your sales team has ten calls and no content plan, send a few anonymised notes to hello@aervio.io, or call +1 438-498-3913 from Monday to Friday, 9–17 ET. We are a Montreal studio, and we will tell you which two or three pages to write first.

Questions

Do we need to record sales calls to do this?

No. Notes, email threads, and the first-contact form field in your CRM are enough to start. If you do record calls, tell people and follow your own consent practices and your counsel's advice. Quebec privacy rules apply to how you collect and keep personal information.

What if a phrase from calls has no search volume in the tools?

Low or zero reported volume does not mean nobody searches it. Tools undercount long, specific queries. If sales hears the question repeatedly and a good answer does not exist, a page can still earn clicks and warm leads. Check Search Console after publishing.

How many calls do we need?

Ten to twenty recent conversations are enough to find patterns. You are looking for repeated phrases and objections, not statistical proof.