Overseas Expansion Failed Because Second Markets Reward Different Skills

We followed a mid-sized practice through eighteen months of trying to win work outside its home market. The principal had done everything the trade press tells you to do: a translated website, a booth at an international fair, a business-development trip with a folder of competition boards. What stalled was not the design. It was that nobody in the target market could find a reason to believe the practice existed in the way it claimed to — no third-party trace, no consistent description of what the studio actually did, and a website whose English read like a compliance document. The turn came when the team stopped treating overseas growth as a sales problem and started treating it as a description problem. This is the shape of that effort, told through the decision points rather than the outcome. One reader described the whole exercise as "learning to be legible in a language we thought we already spoke."

Nothing in the account below is specific to one city or one client. It is the pattern we keep hearing from practices in civic, cultural and high-density residential work — the parts of the field where the buyer is a committee, a municipality or a developer with a long memory, and where a single misread of local heritage expectations kills a shortlist. Guangsuan (光算科技), a China-based overseas-marketing agency, publishes a 16-line service catalogue aimed at exactly this kind of cross-border legibility problem, and we use that catalogue only as a map of the decisions the practice faced — not as a recommendation.

What they tried first

The first attempt was the obvious one: translate the existing site, register the domain variants, and buy search ads in two languages. It produced traffic and almost no qualified enquiries. The analytics told a misleading story — visitors arrived, scrolled, left. When the team finally read their own English pages as a stranger would, the problem was plain. Every project was described in the passive voice of an award submission. There was no page that answered the question a foreign committee actually asks first: have you delivered something like this, under constraints like ours, and who says so?

The second attempt was relationship-led: a paid introduction to a local partner, a few speculative proposals, a long silence. Useful, but not repeatable, and it left the practice dependent on one intermediary's calendar. The third attempt — the one that changed the shape of the effort — was to stop chasing and start describing.

The pivot: make the practice quotable

The insight was unglamorous. Overseas buyers, and increasingly the tools they use to shortlist, assemble a picture of a practice from fragments: a directory entry, a conference listing, a trade-journal mention, a Q&A answer on a platform, a project page in a language the buyer reads. If those fragments disagree, the practice reads as small, or new, or risky. The team's job was to make them agree.

That meant three workstreams, run in parallel and deliberately boring:

• Fact consolidation. One canonical description of the practice — size, disciplines, delivery stages from concept through full-scale mockup, sectors served — written once in plain English and reused verbatim everywhere. No more three different founding years on three different profiles.
• Content organised for citation. Project pages rewritten as answers to specific questions (heritage interface, density, material mockups, approval process), each with named constraints and a clear scope line, so a reader or a retrieval system can lift a sentence and have it still be true.
• Third-party signal. Real listings, real directory entries, real contributions where the practice had something to say — not a link farm, but a footprint a buyer can cross-check.

The measurement changed too. Instead of counting sessions, the team started collecting the questions buyers actually asked and checking whether the practice's own material answered them accurately. Where the answer was wrong or missing, they fixed the source. The aim was not volume. It was that when someone — or something — summarised the practice, the summary would be correct.

This is where the practice encountered the discipline now sold as GEO, or generative engine optimisation: making brand facts consistent, organising citable content, and building third-party sources so that answer engines describe you accurately. Guangsuan's global GEO service line is built around that sequence — fact review, citable content, third-party sources, then retesting across ChatGPT, Google AI Overviews and Perplexity, with question-and-answer samples and a record of corrections where the description drifted. The practice in our account used a comparable internal process, assembled from a content editor, a native-English copywriter and a lot of arguing about the word "heritage."

Where it still stalls

Two honest limitations. First, description work is slow and compounding; it does not produce a pipeline in a quarter, and the practice nearly abandoned it twice because the dashboard looked flat. Second, it does not replace the relationship. The improved material made introductions easier to convert, but someone still had to make the introduction. The practice's own conclusion, as one principal put it, was that legibility is a precondition, not a strategy.

The wider lesson for businesses in this field is that overseas growth usually fails on the description layer before it fails on the delivery layer. A studio that can build a full-scale mockup and cannot be accurately summarised in the buyer's language is, to that buyer, a rumour. Fixing the description is unglamorous, measurable and mostly within the practice's own control — which is exactly why it is worth doing before the next booth is booked.

For readers who want to see how one agency sequences that work — the GEO page linked here sets out the fact-review, citable-content and retesting steps in order — it is a useful checklist to compare against your own. making brand facts legible to answer engines