Keyword Research Without a Subscription
A working method for finding what to write, using free sources for the shape of demand and paid data only where it actually decides something.
Most keyword research advice assumes you have already bought the tool, and most keyword research spending happens before anyone has decided what the research is for.
The method below is ordered by cost. Free sources establish what people are actually asking. Paid data enters only at the point where a number changes a decision, which is later than people think and on a much shorter list.
Start with what you already own
If your site has any traffic at all, Search Console is the highest-value keyword source you have, and it is free. It is also the only one that knows which searches Google already considers your site a plausible answer for.
Sort your queries by impressions and read them as a description of what Google thinks you are about. Two things fall out immediately: terms you rank for and did not target, which tell you what your content actually says, and terms you targeted and do not appear for, which tell you where you were wrong.
Then filter to queries where you sit somewhere on the second page. These are the shortest path to traffic you will find anywhere, because the hard part — being considered relevant at all — has already happened. A better-matched title, a section that answers the question directly, or an internal link from a stronger page is often the whole intervention.
Nobody who has this data and skips it is doing research. They are doing guessing with a subscription.
Then take the shape of demand from free sources
Before paying for volumes, get the language right. The words people use rarely match the words a business uses about itself, and paid tools cannot fix a seed list that is phrased wrong.
Google's autocomplete, its "people also ask" panel and its related searches are all free, all live, and all sourced from actual search behaviour. Type your topic and read what Google offers to complete it. Do it for the variants, and for the question forms.
Your own inputs are just as useful and less used: support tickets, sales call notes, the searches people run inside your site, and the phrasing customers use in their first email. That last one is where the commercially valuable, low-competition language usually lives, because it is how buyers describe the problem before they know the vendor's vocabulary.
At the end of this you should have a messy list of a few hundred phrases and, more importantly, a feel for how your market talks.
Now pay for the numbers that decide something
Here is where paid data earns its place — and where most spending is wasted, because people pull metrics for everything instead of for the decisions they are about to make.
There are only three questions worth buying an answer to:
Is anyone actually searching this? Volume separates a phrase you invented from one people use. You need this across the whole shortlist, and it is the cheapest of the three.
Who is already ranking, and are they beatable? This is the one that saves the most money, because it stops projects before they start. Look at the actual results page. If the first page is entirely major publishers or entirely the exact retailers you cannot outrank, the volume is irrelevant — you are not getting it.
Does this term match the audience I want? Look at what currently ranks. If the page-one results are all aimed at a different customer than yours, Google has already decided what that query means, and a page aimed at your customer will not rank for it no matter how good it is.
Notice that two of the three are answered by looking at a SERP rather than at a metric. Difficulty scores are a model of that same evidence, compressed into a number. They are fine for sorting a long list quickly, and they should never be the last thing you look at.
Where a tool actually helps
The manual version of the above is entirely viable and slow — the slowness is joining things. Search Console has your queries but no volumes. Volumes have no ranking context. SERP results are a tab you open one at a time.
That joining is what a research tool is for, and it is also the whole of what an agent with data access does well: give it your seed list and your positioning, and ask for the terms with meaningful volume, achievable competition, and results pages aimed at your customer.
PandaCrawl does that on your own DataForSEO account, so a research session appears on your bill as what it cost rather than coming out of a monthly allowance. That matters here specifically, because the method above deliberately front-loads the free work. If you have done the first two steps properly, the paid step is a short, well-aimed list — and paying per call rather than per seat is the arrangement that rewards you for having been disciplined.
What to actually do with the output
Research produces a list. A list is not a plan, and the gap between them is where most of this work dies.
Group the survivors by the page that would answer them, not by the keyword. Several phrasings of the same question are one page, not six, and building six is how sites end up competing with themselves. Then order the pages by whether you can say something the current top result does not.
That last question decides more than every metric above it. If the honest answer is "nothing, we would say the same things slightly differently", the term is not an opportunity regardless of what the volume column says.