Keyword research

The five free sources, in the order you should use them

Nobody needs a subscription to start. They need a method. Work these in order — each one narrows what you look at in the next.

Budget three hours for a first pass on a small site. Boring, and the most valuable three hours in the plan.

1. Google Search Console. The only source built from your own data. Performance report, Queries tab, six months, sorted by impressions. Every row is a query a real person typed. Start here, always.
2. Google autocomplete. Type your core term and cycle the alphabet after it, then prefix it with "how", "why", "cost of", "best". Drawn from live query logs, so more current than any tool's database.
3. People Also Ask. Expand the box twice on each main term. Google is showing you how it clusters sub-questions — a free topic map.
4. Google Keyword Planner. Free with any Ads account. Without live spend it reports wide ranges rather than exact numbers — annoying, but honest. Use it for relative scale and new phrasings, not precision.
5. Competitor navigation. Open the four competitors you actually lose deals to. Their services menu and blog categories are their keyword strategy, already published.

Why the volume number lies, and lies harder in India

Every tool's volume figure is a model, not a measurement — built from clickstream panels, Keyword Planner ranges and extrapolation. Two reputable tools give different numbers for the same query on the same day.

In the Indian market the error gets wider for reasons that are structural, not fixable:

• Language fragmentation. The same intent gets typed in English, Hinglish and regional scripts. Each variant is counted separately, so each looks small and the real total is invisible.
• Spelling variance. "Jewellery" and "jewelry", "optimisation" and "optimization", every transliteration of a city name. Demand splits across all of them.
• Panel coverage. Clickstream panels underrepresent Indian users, so Indian estimates carry more modelling and less observation.
• Rounding to zero. Anything under roughly ten searches a month gets reported as 0 or 10. That's where most high-intent long-tail queries live.

Score by intent and beatability, not by volume

Once you have a raw list — a few hundred rows is plenty — score each row against three questions. Anything scoring 1 across the board gets deleted, not parked.

• Volume is a tiebreaker only. When two queries score identically, take the bigger one. That's the whole role volume should play.
• Check the SERP before scoring question two. Ninety seconds in incognito beats any difficulty metric — see search intent.

Group into one page per intent

This step separates research from a list, and most people skip it. You aren't building one page per keyword. You're building one page per question a searcher is asking, and each will rank for dozens of phrasings.

"seo agency cost", "how much does an seo agency charge" and "seo agency pricing india" are one page. Publishing three is how you create keyword cannibalisation and end up with three pages at position 14 instead of one at position 4.

Sort by meaning, not by string. Two queries satisfied by the same page belong together. Two needing different formats — a definition versus a pricing table — are separate pages.

The one-sheet output, and the publishing order

The deliverable is one sheet with a row per planned page — not a 4,000-row keyword export, which gets abandoned in a shared drive within a fortnight.

Six columns: target page, primary query, variants it should cover, intent, the format the SERP demands, the URL. Then order the rows — commercial pages first, each cluster's pillar before its supporting pages, and refreshes ahead of anything new, because an existing URL already carries whatever authority it built.

That order is your publishing calendar. Our keyword research work ends with exactly this sheet, and it's where an SEO retainer starts — ours from ₹75,000/mo, smaller sites from ₹40,000.