The automation ran every night this week. Seven commits, seven calculator pages polished with verified benchmarks and worked examples. The equity and startup finance cluster: DCF, IRR, NPV, Payback Period, Vesting, Cap Table, Dilution. One per night, 21:15 UTC like clockwork.
When I pulled the GSC data to see what had changed, I got an answer I hadn't seen before.
Five of the seven pages have zero impressions in 28 days. Not a low position. Not privacy-threshold queries hiding behind GSC's anonymization. Empty rows.
The setup
The automation has been running for over 40 consecutive days at this point. If you've been following this series, the arc of the last two weeks matters:
- Last week's cluster was SaaS and marketing calculators — CPM, CTR, ROAS, Churn, Burn Rate, ARR, MRR, and several others. When I checked those pages in GSC, they appeared at near-page-1 positions. Then I drilled into the query-level breakdown and those positions dissolved — they were arithmetic averages across dozens of one-impression long-tail queries scattered between position 2 and position 95. Not real top-10 rankings. A mean of a graveyard.
- This week's cluster was equity and startup finance calculators.
The tools look polished. Specific titles, contextual meta descriptions, worked examples, internal links. Here's what GSC shows for the 28 days covering the improvement window:
| Page | 28d Impressions | 28d Clicks | Avg Position |
|---|---|---|---|
| /tools/dcf-calculator/ | ~12 (query drilldown) | 0 | 57.4 |
| /tools/dilution-calculator/ | 2 | 0 | 72.0 |
| /tools/irr-calculator/ | 0 | 0 | — |
| /tools/npv-calculator/ | 0 | 0 | — |
| /tools/payback-period-calculator/ | 0 | 0 | — |
| /tools/vesting-calculator/ | 0 | 0 | — |
| /tools/cap-table-calculator/ | 0 | 0 | — |
Five pages with no data at all. Two with single-digit impressions. Seven pages improved this week. Combined: 0 clicks across the entire cluster.
Finding #1: The escalation — from fake positions to no positions
Last week, the SaaS calculators had the decency to show up in GSC with positions between 9 and 12. That turned out to be an illusion — GSC's page-level average flattened a distribution of invisible long-tail queries into something that looked like "almost page one." Not useful, but at least it existed.
This week is a step down from that.
The IRR, NPV, Payback Period, Vesting, and Cap Table calculators don't collapse under query-level inspection — they return empty responses. When I query GSC for those specific page URLs, the API comes back with "responseAggregationType": "byPage" and no rows. No queries, no positions, no data at all.
I curled all five pages as Googlebot to confirm they're actually live and rendering. They are. The IRR calculator serves:
<title>Free IRR Calculator with MIRR & Hurdle Rate | Valuefy</title>
<meta name="description" content="Free IRR calculator that computes MIRR and compares against your hurdle rate. Enter cash flows, get IRR, Modified IRR, and a clear go/no-go verdict." data-rh="true">
<h1>IRR Calculator with MIRR & Hurdle Rate Comparison</h1>
Well-formed title, specific description, clean H1. The page is there. Googlebot can render it. Google just hasn't decided it's worth returning for any query above the privacy threshold in 28 days.
The SaaS cluster had fictional rankings. The equity cluster has no rankings at all. Week over week, the automation is running against pages with less and less Google surface area.
Finding #2: The page driving 24% of site traffic was never automated
Over the past 28 days, valuefy.app got 42 clicks total across all pages. One page contributed 10 of them:
/tools/purchase-order-generator/
10 clicks | 1,048 impressions | position 46.1
That's 24% of the site's clicks from a single page — one the automation has never touched. The query breakdown is the clearest I've seen on the entire site: specific terms with real user intent. po maker at position 7.8 (29 impressions, 1 click). purchase order generator at position 22.6 (117 impressions, 2 clicks). free purchase order generator at position 14.9 (16 impressions, 1 click). purchase order maker at position 18.5 (16 impressions, 1 click).
Five distinct clicks from five distinct queries. Actual people searching for a concrete thing, finding the page, clicking it.
The seven equity calculators improved this week combined for 14 impressions and zero clicks.
The automation picks the next page from an internal queue. The purchase-order-generator is apparently not in that queue. The pages getting the least Google attention are — which is either very principled (improving weak pages to bring them into the index) or exactly backwards (spending effort where Google has already signaled no interest). Right now the data doesn't support the principled interpretation.
Finding #3: India clicks 7.5x harder than the USA per impression
In the 28-day country breakdown, the United States generated 6,630 impressions. That's about 72% of all country-attributed impressions for the month. The average position for US traffic is 60.5. The US delivered 2 clicks.
India generated 734 impressions — roughly 8% of the total. Average position 18.5. India delivered 15 clicks.
| Country | Impressions | Clicks | Avg Position | CTR |
|---|---|---|---|---|
| USA | 6,630 | 2 | 60.5 | 0.03% |
| India | 734 | 15 | 18.5 | 2.04% |
The US market generates enormous impression volume because the absolute search volume for finance and calculator terms is enormous there. But we're ranking at position 60 for almost all of those US queries — page 6, where the crawlers operate and humans don't. India gets our pages at positions where a 2% CTR is achievable because those queries are less contested.
This isn't a hreflang problem or a geo-targeting problem. It's the authority gap with a geographic filter on it. Wherever the site happens to rank in the top 20, CTR is normal. Everywhere else, it's 0.03%.
Finding #4: The og:title duplicate is still there
I curled each page as Googlebot to verify meta tags as part of this week's check. Every page — IRR, NPV, DCF, purchase-order-generator, the homepage — returns two og:title tags:
<meta property="og:title" content="Valuefy - Free Business Calculators & Financial Tools">
...
<meta property="og:title" content="Free IRR Calculator with MIRR & Hurdle Rate | Valuefy" data-rh="true">
This was identified in Week 1. Week 3 explained why it wasn't fixed quickly — a deliberate architecture decision with one wrong assumption about React Helmet's behavior during prerendering. Since that post: no change across any route.
I'm not relitigating the decision here. I'm noting it because the automation now ships improved content on pages that carry this conflict. Whatever click-through value a page-specific og:title provides for social sharing is partially undercut by the generic sitewide tag that loads first.
What I'm going to do about it
Gate the automation on GSC existence — before the bot improves a page, check whether that page has any impressions in the last 28 days. Zero impressions means skip and move to the next candidate. Improving pages Google can't see generates no useful signal.
Audit why the equity cluster is absent from GSC — five live, well-structured pages with zero GSC presence. Most likely causes in order of probability: not internally linked from anywhere (so Googlebot hasn't crawled them), missing from the sitemap, or simply not yet indexed despite being crawled. Check all three before writing off the pages entirely.
Protect the purchase-order-generator from the automation queue — it's the only page with consistent, identifiable traffic from real queries. I need to understand what Google likes about it before the automation inadvertently changes it.
Stop improving pages ranked at 50+ — better content can lift rankings, but not from position 60. The constraint at that depth isn't on-page quality. It's authority.
The uncomfortable lesson
The automation moves through clusters in sequence: SaaS calculators last week, equity calculators this week, something else next week. Each cluster has connected worse to actual Google visibility than the last one. SaaS pages had fictional page-1 positions. Equity pages have no positions at all.
There's something odd about running a content improvement loop on pages that are invisible to the index. The pages get genuinely better — more specific titles, real benchmarks, worked examples. Nobody reads them. Google doesn't surface them. The loop runs again tomorrow night at 21:15 UTC.
The positive signal: 15 clicks in the last 7 days against a 28-day average of roughly 10.5 per week. Site-level traffic is trending up slightly. It's just not coming from the pages being improved — it's coming from the purchase-order-generator and a handful of others that the automation has apparently passed over.
I'll check in four weeks whether any of the equity calculators have developed GSC presence — not positions, just existence. Impressions above zero. If they're still empty at that point, improving them was the wrong activity entirely, and the automation needs a different filtering step before it picks the next page.
I'm running these experiments on valuefy.app and writing the findings as I go. If you're building programmatic SEO, debugging an automation that runs on its own priorities, or stuck on the gap between "content improved" and "traffic improved" — I'd be glad to compare notes. Drop a comment or reach out.
I also run AImiten, where we build AI tooling for companies. This side project is where I stress-test ideas before they touch client work.
Top comments (0)