Nothing Below Position Three Exists

We pulled sixteen months of Google Search Console data for this website — every query, page, and position from January 29 through August 16, 2026 — and froze it as a baseline before starting a site overhaul. The numbers were uncomfortable enough to be useful, so we're publishing what they taught us. If you run a niche B2B software site, yours probably look similar.

The raw shape: 45,912 impressions, 534 clicks. And underneath that, four lessons.

1. The click-through cliff is a wall, not a slope

Bucketing our top 500 queries by average position:

Positions 1–3: 18.89% click-through rate
Positions 4–10: 0.10%
Positions 11–20: 0.14%
Positions 21+: zero clicks on 1,451 impressions

We expected decay. We found a wall. Ninety-three of our queries sat on page two with 3,096 impressions — 45% of all query impressions — and produced not one click. For a niche site, "we rank for it" means nothing; either you're in the top three or you're invisible. That reframes SEO priorities completely: fewer targets, won completely, beats broad page-two coverage that produces literally zero.

2. Your brand is probably your only funnel

One query — our company name — produced 120 of our 128 top-500-query clicks, at a 28.78% click-through rate from position 1.6. The other 477 queries produced eight clicks combined.

The honest reading: search was not introducing us to anyone. It was catching people who already knew the name from somewhere else — a demo, an email, a conversation. That's not a failure exactly, but it means the causality most startups assume ("content brings strangers, strangers become customers") ran backwards for us. The content that creates the demand is whatever makes someone type your name next week. Which changes what a blog is for.

3. Some of your impressions are not about you

1,220 of our query impressions — one in six, on the query set we analyzed — were people looking for a different "leaf": a PDF tool called leafpdf.com, Kubernetes log strings containing leaf_update, a CAD file site. Zero of them clicked, correctly.

If we hadn't filtered those, our impression trend would have looked like growing interest in our product. It was actually growing confusion in Google's matching. Any dashboard built on raw impressions was lying to us; we now judge on clicks and on a junk-filtered query set only. Check your own long tail before celebrating an impressions chart.

4. The demand that exists is comparison demand

The one real pool of non-branded intent in our data: 44 queries naming our competitors — pricing, alternatives, X-vs-Y — totaling 2,431 impressions. Our average position on them: 15.7. Result: one click.

People at the moment of choosing a tool were searching, Google was showing us to them, and we were on page two, which per lesson one means we weren't there at all. Meanwhile two developer tutorials we'd written carried 44% of all our impressions — traffic aimed at an audience adjacent to, but not the same as, our buyer. The gap between where our impressions were and where the buying intent was is the single most actionable thing in the dataset.

What we're doing about it, in public

The baseline is frozen in our repo, the overhaul is running against it, and we'll publish the 90-day delta — including if it's flat. Given lesson two, we're skeptical that search becomes a primary channel for a product this niche; the plan treats it as demand capture, with demand creation happening elsewhere. But now the hypothesis has a number attached, and a date.

The methodology notes: positions are impression-weighted; we make no causal claims off pages with under 100 impressions a week (a lesson its own false alarm taught us — a scary-looking homepage position drop turned out to be noise that predated the change we'd suspected). Sixteen months, dataState: final, one domain property.

Nothing Below Position Three Exists