Case studies

Clients are described by category and scale, never named. That rule protects you as much as it protects them, and it is why there are no logos anywhere on this site.

Four pieces of work

Each one describes what the job was, why it was hard, what we actually did and what shipped. The last one includes the part where the result went away, because a case study that stops at the good week is not evidence.

Crypto casino operator · 9 market domains · 10 languages

A nine-domain market network, 527 pages across 14 builds

The problem: The operator needed a presence in Asian markets where its main domain had little visibility and where most content suppliers cannot work at all. Japan, South Korea, Taiwan, Hong Kong, Thailand, Vietnam, Indonesia, Malaysia, Singapore and the Philippines.
Why it is hard: Ten languages and five writing systems, several of them bilingual builds on one domain, and a separate set of operator facts per market. The failure that kills this kind of project is invisible: text from one market's site leaking into another, or the wrong script appearing in a page nobody on the client side can read.
What we did: Every site written from its own keyword research in its own language, rather than one English plan translated nine ways. A review across the network found one market that had been built as a near-copy of its neighbour; it was rewritten from its own research, and a similarity check now runs on every build so that cannot recur. Each build also goes through automated checks for wrong-script text and for titles over the pixel limit in that language.
What shipped: 527 pages across 14 builds on 9 market domains, each page carrying its own language tags and structured data, each site with its own sitemap, delivered as deployable builds with a handover document per site.
Where it stands: Built, delivered and held at the client's request: every page is set to noindex until they decide which markets to open and in what order. We can show the work. We cannot yet show it ranking, and we are not going to imply otherwise.

CS2 skin marketplace · 298 pages · 2 languages

A catalogue of item pages, written and checked one row at a time

The problem: A marketplace with a large item catalogue and no pages for the individual items people actually search for. Every item needed its own page covering float range, wear conditions, rarity, the case it comes from and its release.
Why it is hard: This is where programmatic content usually fails. One template across hundreds of rows produces pages that all open with the same sentence, and a batch checked as a batch ships the same error in every one of them. For many items the manufacturer published no description at all, so there was no source text to work from.
What we did: One template with per-item facts checked per row, and a structure that flexes with the item instead of forcing every page into the same shape. Where no official description existed, the design section was written from the item itself. Prices are left to the client's own live widget rather than written into the copy, because a price typed into a page is wrong within a day.
What the checking caught: In one batch: ten sentences across eight pages claiming wear shows differently on different parts of an item, when the game bakes one texture per finish. Seventeen pages carrying unsourced demand claims. Five exclusivity claims, three of them contradicted by another page in the same batch. All fixed before delivery, and one error found on a page already with the client was flagged rather than quietly corrected.
What shipped: 298 catalogue pages, including 103 in Brazilian Portuguese built as a separate market rather than a translation pass. One category finished complete at 94 of 94 items. Work continues in batches.

Our own demonstration build · Ontario · not a client project

A casino site researched, written and built end to end, to show the chain

What this is: A self-initiated build using an invented brand, made to show what the full chain looks like in one piece of work. There was no client and no real operator, and we would rather say so than dress it up as an engagement.
What we did: Keyword research for the Ontario market, a page plan, then the copy for every page, then the site itself as hand-written static HTML with one stylesheet, self-hosted fonts and no third-party scripts. Structured data, sitemap and robots file included.
Quality control: The build was gated by a script before anyone looked at it: tag balance, one title and one H1 per page, duplicate titles across the site, title and meta pixel widths, broken internal links, missing image alt text and leftover placeholder text. A measurement that could not be taken was treated as a failure rather than a pass.
What it produced: A complete five-page site plus a 404, packaged with a developer handover document and a technical audit of the build against a 40-point checklist, with every item either answered or marked as not judgeable from a build folder and the reason given.
Why it is here: Because the same chain, run for a real brand, is what most of this site is selling, and a demonstration you can inspect is more use than a claim you cannot. This very site was built with the same tooling.

Affiliate · French market · relaunched domain

Position 3 in three weeks, and the honest part of what happened next

The problem: Take a repurposed domain that already had authority into the French top five for no-deposit-bonus terms, quickly after publish.
What we did: Built the outline, wrote the content and set the link plan. Most of the writing effort went into making the page genuinely better than what was already ranking. The agency behind the project then ran a front-loaded link campaign, heavy in the first two weeks rather than a slow drip.
Result: It reached position 3 in the French results within weeks of publishing, and held there for three weeks.
What happened after: It dropped when the company moved on to other projects and link support stopped. That was a budget decision rather than something breaking, and the page itself did not change.
What we will not claim: That this was content alone. The speed came from three things together: a domain with existing authority, content better than what was ranking, and a front-loaded link push. The result is also self-reported by the client, since we never had analytics access to verify it independently. A case study that hid either of those would be the kind of proof this market is already full of.

Common questions

Why are there no client names or logos?

Because the work is covered by agreements that do not allow it, and because a good share of it is white-label delivery where our name appearing anywhere would defeat the purpose.

It costs us something. A logo wall converts better than a paragraph. We would rather lose that than put a client in an awkward position to win a pitch.

Why are there no traffic graphs?

Because we usually do not have analytics or Search Console access, so any graph we showed would be a number a client told us rather than one we measured. Publishing it as proof would be exactly the kind of unverified claim the rest of this site argues against.

What we can describe precisely is what was delivered, at what scale, in what languages, and what quality control it went through. That is the part we controlled.

Can you give references?

Sometimes, with the client's permission, and it depends which client and what you are asking them. Raise it when a scope is on the table rather than at first contact, so we are asking someone a specific question on your behalf rather than a general one.

Is the work in these case studies typical?

The scale varies a lot. The 298-page catalogue and the nine-domain network are at the large end. Most engagements are smaller: a batch of pages for one market, an audit, a set of briefs.