52 clicks in 28 days. That's 1.79 per day — and compared to the Week 1 baseline of 45 clicks in 90 days (0.5 per day), it's a genuine 3.5x improvement. I'll take it.
But as soon as I started drilling into which pages moved and why, the clean narrative fell apart. The best-performing page has a 26% CTR driven by queries I can't see. The commitments I made in Week 5 — reorder the automation queue, fix the og:title bug — didn't happen. And the biggest English-language search market in the world is showing us 6,967 impressions per month and handing back 3 clicks.
The click rate tripled. The story is messier than that.
The setup
The automation has been running every day since early April — one calculator page per night, verified benchmarks added, worked examples, internal links. 40+ pages improved over 90 days. Week 5's post identified that 8 of those pages weren't even getting impressions from Google, and committed to four things:
- Reorder the queue by impressions — hourly rate calculator first (880 impressions at the time)
- Use the WACC calculator as a two-week control for whether on-page changes move positions
- Watch the ROI calculator for position recovery
- Fix the duplicate og:title tags on every page — treated as a build blocker
Here's what the 28-day GSC data shows.
What I expected vs. what the data showed
I expected the pages the automation touched in early April to start showing movement. And I expected the four action items from Week 5 to have been executed.
What happened:
| Metric | Expected | Actual |
|---|---|---|
| Early-April pages (PE ratio, Funding, Cap Table) | Some position gain after 4-6 week crawl lag | PE ratio: 29 → 18.7. Funding: position 11.5, 2 clicks. Cap Table: position 9.5, 1 click. |
| Recent improvements (CPM, CTR, ROAS, CPA, EBITDA) | Still too early | Confirmed buried at position 60-90 |
| Queue reordered to hourly rate first | Done | Automation ran CPM, CTR, ROAS, CPA, EBITDA, AdSpend, ConversionRate instead |
| og:title bug fixed | Done | Still 2 og:title tags on every page (curled and confirmed) |
Two of four action items sat idle for another week. The automation kept running the queue it had.
Finding #1: The best-performing page is hiding its traffic from me
The business plan generator showed up in the 28-day data with 5 clicks, 19 impressions, 26.32% CTR, position 12.2. A 26% CTR would be extraordinary for any page. At position 12 — the top of page 2 — it's essentially impossible under normal distributions.
So I pulled the query-level breakdown. Here's what GSC returned:
| Query | Clicks | Impressions | Position |
|---|---|---|---|
| business plan generator | 0 | 2 | 97.0 |
That's it. One query, two impressions, position 97, zero clicks. The remaining 17 impressions and 5 clicks are below GSC's privacy threshold — queries with fewer than 16 impressions that don't show up in the breakdown at all.
The 26% CTR isn't what it sounds like. There isn't a single query where one in four searchers clicks our result. What it means is: this page ranks high (probably top 3) for some specific queries that I can't identify, those high-position queries generate clicks, and meanwhile a cloud of long-tail queries at position 50-97 pulls the aggregate position down to 12.2. The aggregate math hides a bimodal distribution. Again.
The week 1 post had the loan-payment calculator sitting at "position 9.8" that turned out to be a fiction. This is the same trap from the other side — a "26% CTR" that's actually a few invisible top-3 rankings averaged with a long tail of zeros.
The page is working. I just can't see for what. The title Google is using:
AI Business Plan Generator — 7 Sections | Valuefy
The live H1: "Free AI Business Plan Generator — 7 Sections, Instant Output." Those seven-sections hooks are probably matching something specific in the query tail. I won't know until the queries break above the threshold.
Finding #2: PE ratio shows improvements work — on a 5-6 week delay
The PE ratio calculator was last improved on April 2 — 38 days ago. No automation has touched it since. The 7-day GSC position: 18.7. The 28-day average: 21.1. Week 1 baseline: roughly position 29 across the 90-day window.
That's a genuine ~10 position gain. From the back of page 3 to the front of page 2. With 4 clicks and 517 impressions in the 28-day window, it's the second-highest click earner among tool pages.
The query that moved:
| Query | Impressions | Position |
|---|---|---|
| pe ratio calculator | 32 | 25.2 |
| forward pe calculator | 4 | 18.0 |
| p/e ratio calculator | 9 | 35.2 |
The "forward pe calculator" query landing at position 18 is why the 7-day average looks better than the 28-day average — it's a low-impression query that happens to be ranking well right now, and it's skewing the recent average toward a better number than the distribution actually represents.
But the directional trend is real: position moved from ~29 to ~21 over 38 days after a single improvement commit. No follow-up automation. No link building. Just on-page content changes — and then 5-6 weeks of waiting.
The implication for everything else in the queue: CPM, CTR, ROAS, CPA, EBITDA — improved in early May — shouldn't show anything visible until mid-June. We're flying blind on last week's work for another month.
Finding #3: The US is our biggest impression market and worst converter
Country-level 28-day data, sorted by impressions:
| Country | Clicks | Impressions | Avg Position |
|---|---|---|---|
| United States | 3 | 6,967 | 45.6 |
| India | 10 | 734 | 17.4 |
| Canada | 1 | 441 | 53.6 |
| Indonesia | 1 | 166 | 13.5 |
| Pakistan | 4 | 156 | 21.7 |
The US has 9.5x India's impressions and 3x fewer clicks. At position 45.6 average, we're buried on page 4-5 for US searches — deep enough that even 7,000 impressions converts to almost nothing. India's 17.4 average position is page 2, and at that depth the clicks start coming.
This isn't a geo-targeting problem. It's an authority problem expressed geographically. The US markets we're showing up in (finance, SaaS, business calculators) are the most competitive on Earth. We get impressions because the search volumes are so large that even position 45 accumulates visibility. We don't get clicks because nobody scrolls to page 5.
The 3 US clicks in 28 days did come from somewhere — the query data doesn't show which — but at $0 in ad spend and domain authority barely above zero, 3 is what page 4-5 looks like in English at scale.
What I'm going to do about it
- Actually run the hourly rate calculator next — it was first on the Week 5 priority list and still hasn't been touched. This needs a manual queue intervention, not an assumption that the automation will self-correct.
- Don't interpret ROAS, CTR, CPM, or EBITDA results before mid-June — the PE ratio example shows the lag is real. Mark the calendar and check those pages on June 15th.
- Try to surface the business plan generator queries — narrow the GSC query filter by date to find if any query broke the 16-impression threshold in a single week. If not, wait.
- Fix the og:title tags — this is the fourth post mentioning this. At this point the commitment is meaningless until there's a commit that proves it happened.
The uncomfortable lesson
Writing action items in a blog post and then not doing them is a specific kind of self-deception. The items go onto a list, the list gives the impression of progress, and the automation keeps running its queue while the list sits there.
None of Week 5's four action items made it into git. The WACC calculator — which was supposed to be a controlled experiment — is at position 55.8 with 1 click and 653 impressions. The queue wasn't reordered. The og:title fix didn't ship. The ROI calculator is at position 41.1, still drifting.
The click rate did triple, which is the genuine good news this week. But the tripling came from early-April improvements starting to surface — not from anything decided or built this week or last week. The lag is long enough that cause and effect barely feel connected.
The experiment running in the background is working. The promises running in the foreground aren't.
I'll check back in June on the pages improved in May. If the PE ratio pattern holds, ROAS should be somewhere around position 55-65 by then. If it's still at 74, the pattern doesn't generalize and I'll say that.
I'm running these experiments on valuefy.app and writing them up as they happen. If you're building programmatic SEO, watching Claude automate your content queue, or trying to make sense of GSC's privacy threshold noise, I'd be glad to compare notes in the comments.
I also run AImiten, where we build AI tooling for companies. This project is where I pressure-test the ideas before they touch client work.
Top comments (0)