Lose the Head Term on Purpose

Now the finding I did not expect at all, and the most immediately actionable thing in this book.

Standard practice says target the highest-volume query in your category. In this case that is unambiguous: merging documents is the most-searched task in the entire space, by a wide margin, in every keyword tool.

Here is what the head terms actually produced, against the odd, low-volume ones, over seven months:

Query clusterImpressionsClicksCTRAvg position
audio conversion (unusual)37,6803,4409.13%4.40
colour inversion (unusual)75,7632,7683.65%5.22
editing110,7903,8803.50%9.73
compression25,3721,0504.14%8.13
splitting10,9821321.20%8.58
comparison17,1761430.83%9.31
organising47,0251020.22%8.42
merging (the head term)1,846452.44%13.49

Read the top and the bottom of that chart together.

The single most-searched job in the category returned 45 clicks in seven months, at an average position of 13.49, the middle of page two, where nobody looks.

Eight days later, on a wider export, the same query had fallen to 5 clicks at position 35.76: page four. Nothing was changed on that page. The head term did not just pay badly, it would not stay won, and the site grew 13% in clicks over the same stretch without it. A ranking you cannot hold is not an asset, it is a loan.

Two obscure conversions that most competitors don't offer at all returned 6,208 clicks between them, at positions 4.40 and 5.22, top of page one.

The two "weird" tools out-earned the flagship by a factor of 138.

Now the same clusters, measured by how often they were seen rather than clicked:

Why this happens

It is not mysterious once you see it. Position is the whole game, and position is a function of competition, not of your effort.

On a head term, you are competing against every well-resourced incumbent in the category, all of whom have a decade of domain authority and a page specifically built for that query. Your excellent page lands at position 13. Position 13 converts to approximately nothing, no matter how good the page is, the click-through rate at the bottom of page two rounds to zero.

On an adjacent, low-competition query, the incumbents often have no page at all, because the volume didn't justify one for a company of their size. Your merely-decent page lands at position 4. Position 4 converts at 9%.

Small query × good position beats big query × bad position, and it isn't close. Volume you cannot reach is worth zero.

The Avengers do not beat Thanos by hitting him harder, they lose that fight twice. They win by moving it onto terrain he does not control. Your version is quieter and much less cinematic: you do not out-rank the incumbent on their best query. You go and win eight queries they never bothered to show up for.

The organise line is the clearest warning in this dataset

47,025 impressions. 102 clicks. 0.22% click-through.

That page was seen forty-seven thousand times and clicked a hundred times. It is technically ranking, average position 8.42, but 8.42 means the bottom of page one, below the "People also ask" box, below the featured snippet, below the images, in the region users scroll past on their way to the next page.

Impressions are not traffic.

Mysterio's whole act is spectacle with nothing behind it, enormous, convincing, and made of projection. An impressions chart is the most Mysterio-shaped number in your analytics: it is the biggest figure on the dashboard, it climbs beautifully, and a great deal of it is you being displayed to people who are looking somewhere else. They are the appearance of progress. Chapter 17 goes into this properly, because it is the most common way founders misread their own dashboard.

How to find the queries nobody serves

The advice "find the jobs the incumbents skipped" is useless without a method, so here is the one that actually produced the two winners in that table. It takes an afternoon and needs no paid tool.

1. Mine autocomplete for the shape of demand, not the volume. Type your category noun into the search box followed by each letter of the alphabet, and write down every completion. Do it again with "how to", "convert", "to", and "from" prefixes. You are not looking for the big ones, you already know those. You are looking for the completions that surprise you, because a suggestion exists only if people typed it, and a phrasing you did not expect is a phrasing your competitors also did not expect.

2. Read the incumbent's navigation for absences. Open the biggest competitor's tool list or sitemap and write down what is not there. This is the single highest-yield twenty minutes in the whole exercise. Their omissions are deliberate: a company with a large cost base decided each of those jobs did not clear its revenue bar. That decision is correct for them and wrong for you, and it is published on their own website.

3. Harvest the language people actually use. Search your category on the large question-and-answer communities and forums, and copy the phrasing verbatim, including the clumsy phrasing. People do not search in your product vocabulary. The audio conversion cluster in that table was found this way; nobody in the category was describing it the way the people who wanted it described it.

4. Search for the failure, not the task. People search error messages, symptoms and dead ends: "won't open", "file too large", "corrupted", "removes formatting". These are extremely low-competition, extremely high-intent, and almost nobody builds for them because they do not look like features.

5. Then check supply, one query at a time. This is the gate, and it is where most of your candidates die.

How to judge the weakest result on page one. The rule is to judge the tenth result rather than the first. Here is what you are actually looking for, in order of how much it tells you:

If the tenth result is a well-built, current, task-performing page from a site with real authority, stop. The volume is irrelevant to you, and no amount of effort on your page changes that this quarter.

How small is too small? The instinct is to reject anything under some thousands of monthly searches. Run the arithmetic instead. A query with 500 monthly searches at position 4, call it 9% click-through, is 45 clicks a month. The merging cluster in the table above, the largest query in the entire category, delivered 45 clicks in seven months. One unserved query at position 4 matched the flagship head term's entire seven-month output in its first month.

The practical floor is not a volume number, it is a portfolio rule: if a query would take under two days to serve and the page can rank top-five, build it. Chapter 3's argument is why. You cannot predict which one wins, so what matters is that each attempt is cheap enough that being wrong nine times out of ten still works.

How to apply this

Rank your target queries by (volume ÷ competition), not by volume.

Practically: for each candidate query, open the results and look at the weakest result on page one. Not the strongest, the weakest. If the tenth result is a thin, generic page from a site with no particular authority, you can take that slot. If the tenth result is a well-built page from a major brand, you cannot, and the volume is irrelevant to you.

Deliberately look for the jobs the incumbents skipped. They skipped them for a reason that is correct for them. The volume doesn't justify the engineering for a company with a large cost base. For you, the same volume at position 4 is a real business. Their rational decision is your opening.

Build the head-term page anyway, but budget nothing for it. It costs almost nothing to have, it captures the occasional long-tail variant, and if your authority ever rises it is already positioned. Just do not plan around it or measure yourself against it.

Expect to be wrong about which of your odd bets wins. This chapter and Chapter 3 are the same lesson from two directions: you cannot predict, so bias toward cheap and toward uncontested, and take the outliers when they appear.

The general principle

At zero authority, you cannot win on demand. You can only win on lack of supply.

Find the searches where the supply of good answers is thin. That is a different research task from finding the searches with the most demand, and it is the one that matters when you are starting from nothing.

About this book

Lose the Head Term on Purpose is chapter 4 of How to Grow Your SaaS to 100K Users Without Spending on Ads, a playbook on taking one product from zero to 100K+ users on search alone, with nothing spent on advertising. Every claim in it is checked against the real data, including the three findings that contradicted the author.

Continue reading

Free tools that implement this book