The one-line version
Fifteen days ago we put 500 programmatic pages live for a real client. As I
write this, the site has earned 6 clicks and 89 impressions. We're
publishing that anyway — because the curated-screenshot version of this
industry is exactly the problem, and because those two numbers mean something
different when you read the rest of the file.
The client and the problem
Bishal is a bridal and event makeup artist in Kolkata (muabishalmakeup.in).
His business lives and dies on being found for exactly the queries brides
type before their wedding — by service, by neighborhood, by tradition, by
occasion. One homepage can't rank for that universe; a human-written page per
query would take a year. Programmatic was the honest fit: a real query
space, real service data, a page per distinct intent.
The page plan
Honest disclosure first: we didn't record the exact candidate count while
planning. That's a tracking gap we won't repeat in the Glasshouse log, where
every rejected page gets counted in public from week one. What survived
planning was 500 pages across service × locality × tradition × occasion
intents for the Kolkata market.
Every surviving page had to pass three tests: a distinct demand we could
name (no keyword-enumeration filler), a data backbone specific to that page
(service inclusions, price band, locality, tradition-specific detail), and
an intent no sibling page already claimed. The plan mixed service intents
(bridal makeup, engagement makeup, SFX) with tradition intents
(traditional Bengali bridal) and occasion intents — because "bengali bridal
makeup artist near me" and "bridal makeup cost in kolkata" are different
searchers with different questions.
The review gate
Gate rejections for this project: not tracked — the gate ran inside the
build sessions and we didn't log its kill counts. We're not going to guess
them now; estimates dressed as data are exactly what this industry doesn't
need more of. From the Glasshouse build onward, rejections are counted and
published weekly. What the gate is for, in practice: it kills the pages that would have made this case study a spam
story — template twins with no unique value, near-duplicate intents, pages
whose data backbone was too thin to be worth a URL. The gate is also why we
can publish this file: everything that shipped survived a deliberate
before-publish inspection, and this article's numbers describe what that
process produces, not a promise of what it always produces.
The timeline
- Planning start: not tracked — we never timestamped when research began; the honest anchor is the launch date below
- 500 pages live on the client's own subdomain (
services.muabishalmakeup.in): 31 August 2026 - 432 keywords under rank tracking from day one
- This snapshot: first 15 days (1–15 September), pulled 19 September
The numbers — first 15 days, labeled precisely
Important framing: GSC's "last 3 months" filter only contains 15 days of
data, because the site went live 31 Aug. Reporting this as "3 months"
would be the dishonest kind of framing — in the direction of making us look
worse. The honest window is: two weeks.
Totals: 6 clicks · 89 impressions · 6.7% CTR · 28.9 avg position.
Now the part the totals hide:
- 6.7% CTR is a good number. When Google surfaced these pages, roughly 1 in 15 searchers clicked. Thin pages don't earn that; intent-matched ones do.
- 28.9 average position at day 15 is on schedule. A weeks-old domain with pages already oscillating around page 2–3 is what the early pSEO curve looks like before the compounding phase.
-
The best line in the file:
traditional-bengali-bridal-kolkata— 64 impressions in 15 days, zero clicks, position trending up. That page is knocking on page 1 for its exact target query, two weeks in. This is the page we'll be watching in public. - The query list validates the page plan. "bridal makeup kolkata" (26 impressions), "kolkata bridal makeup package", "kolkata bengali bridal makeup", "bridal makeup cost in kolkata", "bengali bridal makeup artist near me" — these are the long-tail intents the plan predicted, surfacing
(drag the image here in the dev.to editor, then delete this line)
- Only 6 of 500 pages have any impressions yet. That is normal and it is the least-understood fact about pSEO: Google evaluates a fresh long tail for months, not weeks. The other 494 pages aren't failing; they're queueing. We'll find out together — see the pledge below.
(drag the image here in the dev.to editor, then delete this line)
(drag the image here in the dev.to editor, then delete this line)
One real defect the data surfaced, reported here first: our Pages report
shows both http:// and https:// variants of the homepage earning
clicks — a canonical/redirect imperfection from launch week. Not fixed yet
as of this snapshot. It's at the top of the fix list, and the public log
will mark the week it ships. It's exactly the kind of thing this project
exists to show instead of hide.
What underperformed
Everything outside the handful of surfacing pages — which, at day 15, is
most of the site. Impressions on several service pages (engagement makeup:
16, SFX: 12) haven't converted to clicks yet, and the homepage's 5 clicks
came mostly from branded navigation. The current best explanations: the
crawl is still shallow (most pages not yet evaluated), and the pages with
impressions are mostly ranking just below the click zone. Both are testable
over the coming weeks — and they'll stay testable in public.
What we'd do differently
The honest list: we launched all 500 pages at once, which bought speed at
the cost of a cleaner crawl story — in hindsight, an initial wave of the
strongest ~100 pages with the rest rolling weekly would have given Google
(and us) cleaner signal per batch. And the http/https canonical slip
shouldn't have survived launch review; it did, and the data caught it, which
is both the mistake and the argument for the review gate.
What's next — the pledge
That was us building for a client, quietly. Starting this week, we do it
again from zero, in public: a new wedding-vendor programmatic site, every
number published weekly for 12 weeks, flops included.
For 12 weeks we publish the build's real numbers every Thursday — traffic,
indexing, gate rejections, decisions, mistakes. If it flops, you'll watch
it flop and read our autopsy. No edits to the pledge.
Today's snapshot (6 clicks, 89 impressions, one page knocking on page 1)
becomes what we'll call a week-0 baseline — the starting line the weekly
log measures against. Bring popcorn. Bring your page plan for the free
teardown slot, too.
💬 Questions about your own page plan? Chat with our AI assistant now: t.me/quantumcx_growth_bot — instant answers, free trial, no card.
Top comments (0)