Most keyword research goes wrong in the first ten minutes, when somebody exports two thousand keywords sorted by search volume and starts writing. Volume is the least useful column on that spreadsheet, and treating it as the main one is why so much content gets published, ranks for nothing, and quietly convinces the business that SEO does not work.
This guide covers how to do keyword research properly: what to look for instead of volume, how to read search intent, how to judge whether you can realistically compete, and how to turn the result into a plan rather than a list.
What Keyword Research Is Actually For
The point is not to find popular phrases. It is to find the specific questions your future customers ask, in the words they actually use, at the moment they are close enough to buying that answering them is worth your time.
That reframing changes everything downstream. A phrase with two hundred monthly searches from people ready to purchase is worth more than one with twenty thousand searches from people doing homework. Businesses that chase the second number produce a lot of traffic and very few enquiries, then conclude the traffic was fake.
Good keyword research answers three things. What do people search when they have the problem you solve? Which of those searches can you realistically rank for? And which ones lead to money rather than to a pleasant chart?
Step 1: Start With What You Already Know
Before opening any tool, write down forty phrases by hand. This feels unscientific and produces better results than most exports.
List the exact words customers use when they call you. Not your industry vocabulary, theirs. If clients say website banwana hai rather than web development services, that gap matters, because they will type what they say. List the problems that make somebody start looking. List every service you offer as a separate line. List what people ask before buying: cost, duration, whether it works, what happens if it does not.
These seed terms are the honest foundation of all keyword research that follows. Every tool from here on is expanding and validating this list, not replacing it.
Step 2: Expand With Free Sources First
You do not need a paid subscription for useful keyword research, particularly for a local business.
- Google autocomplete. Type your seed and go through the alphabet after it. These are real, frequent searches.
- People also ask. Every expansion loads more. Twenty minutes here gives you a genuine question list.
- Related searches at the bottom of the results page, which is Google telling you what else the same people looked for.
- Google Search Console, if the site has any history at all. The queries report shows phrases you already appear for, which is the highest-confidence data you will ever get.
- Keyword Planner, free with an ads account, giving ranges and, more usefully, the cost per click.
- Your own inbox and call log. The questions people actually ask you are keyword research nobody else has.
Step 3: Sort Everything by Search Intent
This is the step that separates real keyword research from a spreadsheet. Every phrase belongs to one of four intents, and the intent decides what you build, not the volume.
Informational. Someone learning. What is schema markup, how do Facebook ads work. These want an article. They rarely convert immediately and they build the audience you later sell to.
Commercial. Someone comparing before buying. Best digital marketing course, X versus Y, honest reviews. These are the most commercially valuable content pages you can write, and most businesses skip them entirely because they feel uncomfortable to publish honestly.
Transactional. Someone ready. Course fees, book a demo, buy, near me. These want a service page or a landing page, never a blog post.
Navigational. Someone looking for a specific brand. Nothing to build unless it is yours.
The reliable way to confirm intent is to search the phrase and look at what already ranks. If page one is entirely blog articles, Google has decided the intent is informational, and your product page will not rank there no matter how good it is. Fighting that judgement is the most common wasted effort in SEO.
Step 4: Judge Whether You Can Actually Compete
Difficulty scores in keyword research tools are estimates built mostly from backlinks. They are a starting point rather than an answer, and for local searches they are frequently wrong in both directions.
Do the manual check instead. Search the phrase and look at page one properly. Are the results big national brands and marketplaces, or businesses roughly your size? Are the pages genuinely thorough, or thin and years old? Is there a forum thread or an outdated listicle sitting in position four? Every weak result is an opening.
Local searches are consistently easier than the tools suggest, because national competitors are not really competing for a city-specific query. A phrase that looks difficult nationally is often quite winnable when a city name is attached to it.
Step 5: Read Volume Correctly
Search volume figures are estimates, rounded, averaged across a year, and often wrong for regional or Hinglish queries. Treat them as rough bands rather than numbers.
What matters more is the shape of the demand. Ten phrases at a hundred searches each usually beat one phrase at a thousand, because they carry clearer intent, face less competition, and can be answered on one well-built page between them.
Seasonality matters too. Course enquiries spike at particular points in the year, and building that content in the quiet months so it is already ranking when demand arrives is worth more than any keyword you could pick in peak season.
And where volume is genuinely unknown, cost per click is the better proxy. Advertisers do not bid on phrases that never convert.
Step 6: Group Keywords Into Pages, Not Lists
One page can and should rank for dozens of related phrases. The mistake is building a separate thin page for every variation, which splits your own authority and leaves you competing against yourself.
Group anything that shares the same intent and would be satisfied by the same answer. Digital marketing course fees, course fee structure, how much does the course cost and course price all belong on one page. Somebody searching any of them wants the same thing.
Then give each group one primary phrase, three or four secondary phrases, and a page that genuinely answers the whole cluster. Two hundred keywords usually collapses into fifteen or twenty real pages, which is a plan you can actually execute.
Step 7: Avoid Competing Against Yourself
Keyword cannibalisation is what happens when several of your own pages target the same phrase. Google picks one, often the wrong one, and both rank worse than a single strong page would have.
The rule is one primary phrase, one owner page, permanently. Write it down in a sheet before you publish anything, and check that sheet every time somebody suggests a new article. Most cannibalisation is created by well-meaning content plans that never looked at what already existed.
If you already have overlapping pages, you have three options. Merge them into the stronger one and redirect the weaker. Or differentiate them properly by intent, so one is the guide and the other is the service page. Or delete the weakest if it serves nobody. Doing nothing is the only genuinely bad choice.
| Signal | Where To Find It | Weight |
|---|---|---|
| Search intent | Look at page one | Highest |
| Real competition | Manual review of results | High |
| Cost per click | Keyword Planner | Medium, proxy for value |
| Search volume | Any tool, roughly | Low, often misleading |
| Difficulty score | Paid tools | Low, estimate only |
Step 8: Turn the Research Into a Publishing Plan
A spreadsheet nobody executes is not keyword research, it is procrastination with columns. Convert it into an order of work.
Publish transactional and commercial pages first, because they are closest to revenue and usually least contested. Then build the informational articles that support them, and link those articles to the money pages properly. Then revisit and expand whatever starts showing impressions in Search Console but sits on page two, which is almost always the cheapest traffic available to you.
Set a realistic pace. One genuinely good page a week beats four thin ones, and it beats twenty pages published in a burst and then abandoned.
Keyword Research for a Local Business
If you serve one city, the whole exercise simplifies considerably and gets easier to win.
Attach your city and your localities to every core service phrase. Digital marketing course becomes digital marketing course in Faridabad, then in Sector 81, then near NIT. These local variations have lower volume and dramatically higher intent, because nobody types a city name idly.
Include near me phrases in your planning even though they never show meaningful volume in tools. They are enormous in reality and are answered by your Google Business Profile and your location pages rather than by an article.
Pay attention to how people actually type in India. Hinglish queries, transliterated words and voice searches phrased as full questions are all common and all underrepresented in keyword tools. Your call log is a better source for these than any subscription.
Finally, look at what local competitors rank for rather than what national brands do. A competing institute two sectors away is a far more realistic benchmark than a marketplace with a million backlinks.
Mistakes That Waste Months
Each of these is common, and each one is recoverable only by redoing work you already paid for.
- Sorting by volume and starting at the top. The highest-volume phrase in any industry is almost always the least winnable and the least commercial.
- Ignoring what already ranks. If page one is all comparison articles, your service page cannot win there. Match the format Google has already chosen.
- Building a page per keyword variation, which fragments authority and creates cannibalisation you will spend a quarter untangling.
- Trusting difficulty scores absolutely instead of looking at the actual results with your own eyes.
- Never revisiting the keyword research. Search behaviour shifts, competitors publish, and a plan built eighteen months ago is describing a page one that no longer exists.
- Skipping Search Console in favour of third-party estimates, when Console is reporting what actually happened on your own site.
- Optimising for a phrase the business does not deserve to rank for, because the underlying page adds nothing new. Keyword research cannot fix content that has nothing to say.
How to Know It Worked
Rankings are a lagging and noisy measure, so judge the work on better instruments.
Watch impressions in Search Console before positions. Rising impressions on the phrases you targeted means Google has started considering your page for them, and that usually appears weeks before any position improvement.
Then watch the number of distinct queries bringing traffic. Effective keyword research always shows up as breadth: one page slowly earning visits from thirty related phrases rather than one. That breadth is the signal that you grouped the cluster correctly.
Finally, tie it to enquiries. Traffic that never converts means you targeted informational phrases and expected transactional behaviour from them, which is a keyword research problem rather than a content problem.
Frequently Asked Questions
Do I need a paid keyword research tool?
Not to start, and particularly not for a local business. Search Console, Keyword Planner, autocomplete, people also ask and your own call log will carry you a long way. Paid tools become genuinely useful when you need competitor gap analysis at scale, which is a later problem than most people assume.
How many keywords should one page target?
One primary phrase and three to five closely related secondary ones, all sharing the same intent. A well-built page will end up ranking for dozens more that you never explicitly targeted, and that is the normal, healthy outcome.
How often should I redo keyword research?
A full pass once a year, a light review each quarter, and a quick check whenever you plan a new section of the site. Search Console should be checked monthly regardless, because it surfaces phrases you never thought to research.
Is search volume different for Hindi or Hinglish queries?
Yes, and keyword tools consistently underreport them. If your customers speak Hinglish, they search in it, and the tool will show a fraction of the real demand. Trust your call log and your Search Console queries over the estimate.
Should I target keywords my competitors already rank for?
Often yes, if you can genuinely produce a better answer. A competitor ranking proves the intent is commercial and the phrase is reachable. Copying their page is pointless; answering the same question more completely is not.
What is the difference between short-tail and long-tail keywords?
Short-tail phrases are broad, high volume and rarely convert. Long-tail phrases are specific, lower volume and convert far better. Small businesses should build almost entirely on long-tail, then let the short-tail rankings arrive as a by-product of overall authority.
Done properly, keyword research is less about tools and more about listening: to what customers ask, to what already ranks, and to what your own Search Console is quietly telling you. Get that right and everything downstream, from the content plan to the internal links, becomes obvious. If you would rather learn this on live data for a real business than from a slide deck, that is how the SEO track is taught.
Google explains how it interprets queries in its own how Search works documentationwhich is worth reading before trusting any tool.
A Worked Example: One Afternoon of Keyword Research
Abstract steps are easy to nod along to, so here is the whole process compressed into a single session for a real kind of business: a coaching centre in a mid-sized Indian city.
First thirty minutes. Write forty seed phrases by hand from memory and from the call log. No tools. Include the awkward Hinglish versions and the questions people ask on the phone, because those never appear in an export.
Next thirty minutes. Run each seed through autocomplete and people also ask. The list grows to roughly two hundred. Do not filter yet; filtering while collecting kills the useful oddities.
Next hour. Open Search Console and add every query already producing impressions. This is the highest-confidence keyword research available and it costs nothing. Anything sitting between positions eight and twenty goes on a separate list marked as quick wins, because those pages already exist and only need improving.
Next hour. Tag every phrase with an intent, checking page one for anything ambiguous. Delete the navigational ones. Separate informational from commercial and transactional, because they become different kinds of page.
Final hour. Group everything into clusters that share an intent and an answer. Two hundred phrases becomes about eighteen pages. Assign one primary phrase to each, record it in a sheet so nothing gets targeted twice, and order the list by commercial value rather than by search volume.
That is a complete piece of keyword research, done in half a day, with free tools. The remaining work is writing the pages, which is the part no amount of research shortens.