Case study · SEO & GEO · Published with the client's accord
Lagrace: six months of search data from a small joinery firm.
From 44 sessions a year to 29,645 impressions a month. This is the whole record: the method, the monthly numbers, the 240 pages that failed, and what is still not won.
Signed Jérôme Mottas, founder · Published 2026-08-25 · Last updated 2026-08-25 · Download the data (CSV)
The starting point.
Lagrace is a joinery company in Haute-Savoie, France: windows, shutters, gates, pergolas. The work was good and the order book lived on word of mouth. The website, built on Wix, had recorded 44 sessions over the previous year. 23 of them came from search. Google had 24 pages indexed.
44 sessions a year is not a typo. It is what an invisible site looks like, and most small firms with excellent work and no search presence would recognise the number if they dared to look it up.
What we did.
We rebuilt the site and published content in two very different ways, which is what makes the data useful. On one side, 39 articles written one at a time: real questions homeowners ask about windows, shutters and installations, each answered fully, with zero duplicated text between any two of them. On the other side, 240 city and service pages generated from a template, the tactic half the local-SEO industry sells.
We also fixed the Google Business Profile. The profile sat in the category “Aluminum window supplier”, which is not the trade, and it was corrected to “Menuisier”, joiner, what the company actually is, along with its services. And we cleaned up cannibalisation: several pages had drifted into competing for the same queries with near-identical titles, so the titles were rewritten until each page owned one intent.
The numbers.
| Metric | Before | Six months later |
|---|---|---|
| Sessions, total | 44 / year | 29,645 impressions / month |
| Clicks from search | 23 / year | 67 (May) → 497 (July) / month |
| Pages indexed | 24 | 362 |
| Content carrying the clicks | n/a | 39 articles, 77% of all clicks |
The owner reports revenue up more than 150% on the entire previous year, within the first three months after relaunch. We publish that figure as what it is: owner-reported, and deliberately rounded down from the raw estimate. The exact accounts will true it up.
What did not work: the 240 templated pages.
Here is the part the case studies you usually read leave out. The 240 templated city and service pages produced 0.23 clicks per page per month. Measured duplication between them ran from 30 to 61%. They cost build time, they diluted the crawl, and they now sit in the index doing close to nothing.
The 39 hand-written articles, with 0% duplication, hold positions 3 to 9 on their queries and carry 77% of all clicks the site receives. Thirty-nine pages outwork two hundred and forty. We built both, we measured both, and we will not build the templated kind again (we are slightly embarrassed it took a measurement to be sure).
Honest limits.
The local pack is not yet won: on the highest-value local queries, the map results still belong to older, more-reviewed competitors, and that annoys us weekly. The domain is young and its authority is thin, so most rankings sit at positions 3 to 9 rather than 1. Impressions grew faster than clicks, which is what a young domain climbing page one looks like from inside.
And one number in this study is not ours: the revenue figure is the owner's estimate, not an audited account. Everything else comes from Search Console and the site inventory, and it is all in the CSV.
The same discipline, on your problem.
A 30-minute diagnostic: we pressure-test your search visibility, AI rollout or transformation, and give you the first three things we would do. No pitch.