The steel contractor procurement teams can finally find

01 · The challenge

Eighty projects, three certifications, no way to be found

Masas Trading & Contracting fabricates and erects structural steel, pipe racks, pre-engineered buildings and roofing and cladding systems from its yard at Al Wukair, south of Doha, with a stated capacity of 2,000 tonnes a month. Its website was a one-page placeholder with a phone number, in place since 2023, and underneath it the install had been compromised: Google had indexed more than a thousand spam pages under the domain, so to a search engine masastc.com was somebody else's storefront. Prequalification with main contractors and with Ashghal happened entirely on paper. There was nothing to send a procurement team, and the address they would have searched for was poisoned.

02 · Objectives

What we set out to do

• Publish every service and every cleared project in both languages, with the pairing declared to search engines.
• Match name, address and phone to the Google Business Profile character for character, everywhere they appear.
• Handle every URL search engines knew from the old site: the legitimate ones redirected, the rest gone for good.
• Answer at least fifteen buyer questions in a form that stands alone when quoted, so an assistant can answer from the site.

03 · Discovery

How we found the real problem

The customer is not the public. It is a main contractor's procurement or QA lead qualifying a steel subcontractor, and they ask four things before they look at a photograph: capacity, welding code, steel grades and surface treatment. The old site answered none of them. The head terms, steel fabrication Qatar and its neighbours, are held by firms with decades of history and directory backing, and a live check found Masas absent even for its own brand name. The strongest competitor publishes its capacity and codes but explains nothing and carries no FAQ markup, and no competitor has a page on prequalification at all. Discovery also settled the harder question: which of the eighty projects could be named, which had to be described by scope and scale only, and which could not appear.

“A steel contractor is qualified on four numbers before anyone looks at a photograph. We built the site so those numbers are the first thing a procurement team, or an assistant, finds.”

The approach, in two lines

04 · Strategy

The plan we committed to

Two decisions shaped the build. The first was to ship the content as code. The plan called for a database and a custom bilingual admin; ten days in, we deferred the admin and shipped the content as JSON in the repository, so every edit is a commit and a deploy, there is no database to run or editing surface to secure, and every public page is static HTML. The second was to make clearance a build failure rather than a rule: contract values do not exist in the content schema so they cannot be published, a project not marked as cleared is anonymised automatically, and the build fails if a hidden client name reaches any output. Two smaller ones followed: four launch services rather than seven, with the structure ready for the rest without a URL change, and a permanent 410 rather than a redirect for the old domain's unrelated URLs, because a redirect would have carried the association onto the new site.

How we approach Search & AI Visibility

05 · Design

The decisions behind the screens

Bilingual by construction, not by translation. Arabic is first-class: logical CSS properties only, right-to-left layouts, a per-locale font budget, and a rule that Arabic is never letter-spaced tight. The language switcher lands on the same page in the other language, never on the home page. We retokenised the brand first, a colour and type set drawn from the existing navy logo, so components consume semantic tokens and no page carries a raw hex value. Every page composes from an approved shortlist of sections, and the build fails if a page invents its own. The structure: home, four service pages, thirty project case studies, certifications, an Ashghal prequalification page that maps the checklist to published evidence, seventeen buyer questions, four articles, contact and request-a-quotation. Motion came after launch, six studies reviewed one by one, and visitors who prefer reduced motion or arrive without JavaScript get the finished page instantly.

How we approach Design & Development

06 · Development

What we built it on

Next.js 15 App Router, React 19, TypeScript and Tailwind 4 on Vercel, on the client's own account: 104 statically prerendered pages. Content lives in JSON; a photo pipeline normalises, curates and withholds frames by clearance; a route manifest generated on every build tells the middleware every public URL. Enquiries go by email from a verified sending domain with reply-to set to the prospect, and WhatsApp deep links run throughout. Nineteen build-failing verification gates and forty-five unit tests check address consistency, clearance, structured data against the visible content, reciprocal locales, no-JavaScript content, unique titles, image sizing, accessibility, redirects and the generated llms.txt. DNS moved from the old registrar to Vercel without touching the company's mail. Lighthouse on a mid-range phone: 94 to 95 for performance, 100 for accessibility, best practices and SEO, with zero layout shift.

07 · Visibility

Built to be found from day one

JSON-LD on every page type, Organization, LocalBusiness, Service, FAQPage, Article, CreativeWork and breadcrumbs, each verified against the visible page by a gate. Reciprocal hreflang on all 94 sitemap URLs, with dates from the content rather than the build clock. Every old URL handled: the holding-page paths redirected, more than 1,100 spam URLs and the scripts serving them answering 410, and the penalty check run clean before the cutover. The Google Business Profile address matched character for character, and the www host consolidated onto the apex after Bing chose the wrong one. For AI search: robots.txt names GPTBot, ClaudeBot and PerplexityBot explicitly, a generated llms.txt carries every page, certificate and answered question, IndexNow feeds Bing, which Copilot and ChatGPT browse from, and a script checks that each buyer question's page carries the specifics an assistant needs. Bing indexed the site within a week.

How we approach Search & AI Visibility

08 · Pipeline

Where the enquiries go

The form delivers to the owner's inbox and the prospect's details are never stored on the site. Analytics are first-party, so there is no consent banner in the way. Two events matter: an enquiry submitted, recording which form but no prospect data, and a WhatsApp click, captured by one listener across the whole site. Attribution is by referrer and campaign tag. No CRM: the owner works from WhatsApp and email, and the site respects that.

How we approach Sales Pipeline