Your Sales Calls Are Your Keyword Research

Most keyword research sold in India starts in the wrong place. Somebody exports a list from a tool, sorts it by monthly search volume, hands you a spreadsheet with 400 rows, and you are supposed to feel that something has happened. You cannot read it. You cannot act on it. And nobody involved has spoken to one of your customers.

You already have better raw material than any tool will give you. If you have been running your business for three years, you have sat through hundreds of enquiry calls. You know the four objections that come up every time. You know the question people ask right before they say yes, and the question they ask right before they go quiet. That is keyword research. It just has not been written down.

This piece is about turning those conversations into pages. No subscription required.

Why the volume spreadsheet fails

Two problems, and both are structural.

First, search volume figures are estimates. Google's own Keyword Planner shows ranges, not counts, and buckets terms together. Third-party tools model the rest. A term shown as "1,000 a month" might be four hundred, might be three thousand, and might be mostly other businesses in your trade checking their own rankings. Treating that number as a fact and ranking your entire content plan by it means ranking your plan by noise.

Second, and more important: volume tells you how many people searched. It tells you nothing about whether they were buying. "CCTV camera price" gets searched a great deal in Delhi. A large slice of that is people comparing a ₹1,800 unit on an e-commerce site. If you install forty-camera systems for warehouses, that term is not your customer, however impressive the number looks in a report.

The thing to be suspicious of is the monthly "keyword ranking report" — a PDF showing twenty terms and their positions, arriving on the 3rd of every month. It is easy to produce, it looks like work, and it can go up steadily while enquiries stay flat, because nobody checked whether those twenty terms were ever attached to money. If an agency's first deliverable is a keyword list rather than a set of questions, ask who they spoke to.

Harvesting the questions you already get

Here is the exercise. It takes about two hours spread over a fortnight, and it should involve whoever actually answers the phone.

1. Open a shared note. One line per question, written the way the customer said it, not the way you would say it. "Kitna time lagega?" is more useful to you than "project timeline enquiry".
2. Mine what already exists. Scroll your WhatsApp Business chats for the last two months. Go through the enquiry form emails. Read the first three messages of ten sales threads. You are looking for the question, not the answer.
3. Ask the person on the counter. Sales staff, receptionist, service engineer. Ask each one: what do people ask you that you are tired of answering? The exasperation is the signal. Repetition is exactly what you want.
4. Note where the call stalled. If three people in a month asked whether you handle GST invoicing for corporate clients, and two of them sounded relieved when you said yes, that relief is a page.

You are looking for a list of thirty to sixty real questions. Do not filter yet. Do not judge whether anyone searches for them. Write them down.

Sorting by money, not by volume

Now sort. Three buckets, and the order matters.

Bucket one — asked late, by people who are nearly buying. Warranty terms, delivery to a particular pin code, whether you do AMC, minimum order, what happens if the machine fails in month seven, whether you can invoice a company registered outside Delhi. These questions arrive after the person has decided you are plausible. Almost nobody writes pages for them, because the volume looks tiny. They convert best.

Bucket two — asked early, comparison questions. "What's the difference between X and Y?" "Should I get the imported one or the Indian one?" "Is it worth doing this in-house?" These pull in people who have a problem but no vendor yet. They are worth more traffic and less immediate business.

Bucket three — asked out of curiosity. Interesting, sometimes good social material, rarely worth a page of its own. Park them.

Then apply one filter to bucket one and two. For each question, ask: if a stranger read my answer, would they be closer to hiring me? If the honest answer is no, drop it. This filter does more useful work than any difficulty score.

One question, one page

The mechanical part is simpler than people expect.

• Use their words in the page title and the first heading. If people say "three phase stabiliser for lift", the page says that, not "voltage stabilisation solutions for vertical transport". Google's search results snippet typically truncates titles somewhere around 55 to 60 characters on desktop, so put the words that matter first.
• Answer in the first paragraph. Not the fifth. If the answer is "yes, we deliver to Noida Sector 62 within two working days", that sentence goes at the top. Everything after it is supporting detail.
• Then give the detail nobody else gives. The exceptions, the caveats, the case where it does not work. This is the part a competitor with a generic page cannot copy, because they have not done the job.
• One clear next step. A phone number that is a link on mobile, or a short form. Not four options.

Group related questions into one page only when a single visitor would plausibly ask all of them in the same breath. Otherwise separate pages. Two hundred honest words on exactly one question beats fifteen hundred words that circle five.

Where AI answers fit, honestly

You have probably noticed that Google now often shows a generated summary above the results, and that some buyers ask ChatGPT or Gemini for a shortlist before they ever open a search page. What that means for a small business is genuinely unsettled, and anyone telling you they have it figured out is guessing with confidence.

What can be said carefully: these systems answer questions. They are built to summarise, and to summarise they need clear source material that states things plainly — specifications, service areas, terms, prices where you are willing to publish them. A page that answers a specific question in a direct sentence is easier for a machine to lift than a brochure paragraph about your commitment to excellence. That is the same page a human buyer wanted anyway, which is why this approach is a reasonable bet either way. We have written more about how this is shifting on our AI visibility work, and the honest position there is that the fundamentals have not changed as much as the marketing suggests.

What we would not do is rewrite your site for a machine at the expense of the person reading it. Nobody outside those companies can see how the ranking works, the systems change every few months, and building your business around an undocumented black box is how people ended up with keyword-stuffed pages in 2011.

Checking whether it worked

You do not need a tool for this either. Google Search Console is free, and the Queries report shows the actual phrases that brought people to each page — real ones, not estimates. Six weeks after publishing a question page, look at two things: is that page getting impressions at all, and are the queries recognisably the question you answered? If people are arriving on words you did not expect, that is a new page, free of charge.

The other check is offline. Ask your sales staff whether a particular question has stopped coming up on calls, or whether callers now open with "I read on your site that...". That sentence is the whole point of the exercise.

What to do this week

Open a note. Write down every question you were asked in the last five working days, in the customer's words. Mark the three that came from people who were close to buying. Write those three pages, badly if necessary, and publish them. Rough and specific beats polished and vague.

If you would rather someone sat with your sales team, ran the exercise properly and built the pages, that is roughly what our SEO work looks like in the first month — and if the site itself makes publishing painful, our web development team deals with that first. Either way, tell us what people keep asking you and we will tell you what we would build.

Questions we get asked