Why localising everything at once fails
Three things go wrong when a business sets out to localise everything. The budget runs out partway, so the edition launches with some sections localised and others still in the source language. Review becomes the bottleneck, because in-country reviewersThe check of translated or adapted content by a native speaker who knows the market and category, who signs off the language before publication. cannot read an estate in the time a launch allows. And every localised page becomes a page that has to be kept current, so the edition's upkeep grows faster than anyone planned.
The search consequence is specific. Google Search Central's guidance on managing multi-regional and multilingual sites, as the international SEO section states it, is that each version of a page should be in one language, navigation and content alike, and that automatically translated text presented as indexable content without review can be treated as auto-generated content under Google's spam policies. So a page is either localised properly or left out of the edition. It is never half-translated, and machine output is never published unreviewed to fill the gaps.
The localisation hub's layer table gives the content lead the decision about which pieces are localised, rewritten or created locally, but no page shows how to make it. The localising websites and campaigns guide lists the core pages every edition needs; the translation and transcreation guide decides how each asset is treated once chosen. This tutorial sits between them: it decides what is chosen.
Before you start
- Know the market and its role. The market entry and prioritisation guide sets which market comes first and what the pilot must prove. A pilot market needs a smaller scope than an established one.
- Have the keyword research. Keyword research for a new market in the target language, with the approved terms and intent for each.
- Have the content inventory or the means to build one. A crawl of the site, the help centre's article list, the email programme's templates, the campaign library and any sales collateral.
- Have usage data from home. Which pages, help articles and emails customers actually use, from your own analytics and support records.
- Know the site structure. The edition's structure is decided by whoever owns international SEO before any page is localised, as the localisation hub advises.
The steps
List every content type the market could see
List content by type, not page by page at first: product or service pages, category pages, pricing, checkout and account pages, delivery, returns and legal pages, help articles, guides and tutorials, blog posts, emails (transactional and marketing), app text, ads and landing pages, sales collateral, video and social content. Under each type, record how many pieces exist and where they live. The list is the estate the market could meet, and the rest of the steps decide what it will meet.
Note the job each piece does
For each type, and for the larger pieces individually, note what job it does for the customer: helps them find the product, explains it, lets them buy, reassures them, serves them after purchase, or builds interest without an immediate purchase. Content with the same job at home may not have it in the new market; a blog post that supports a mature home audience may do nothing for a market where the brand is unknown.
Mark what the market must have to buy and be served
Mark the content without which a customer in the market cannot understand the offer, buy, or get help: product and pricing pages, checkout, delivery, returns, contact, the legal pages and the help articles customers use most. These are must-haves regardless of demand or cost. The localising websites guide's list of core pages is the starting point, and the legal pages need local legal review rather than translation, which that guide also explains.
Add local search demand to each candidate
For each remaining candidate, add what the keyword research found: whether people in the market search for what the piece answers, the Search VolumeThe estimated number of times a query is searched in a period, usually per month. range and the Search IntentThe goal behind a search query: to learn, to find a site, to compare or to act.. Demand often sits in different places from home. The SEO traffic estimator gives a rough range to compare pieces with each other, not a forecast of what any page will earn. A piece with no local demand and no must-have job is a candidate for a later wave or for leaving out.
Label the treatment each would need
Give each candidate its label on the spectrum: translation, localisation or transcreation, as the translation guide sets out, or "create locally" where the source has nothing that fits the market. Note pieces that would need local examples, local proof or local legal input, because those cost more than their length suggests.
Estimate the cost to localise and to keep current
Estimate the effort for each candidate from your own supplier quotes and reviewer capacity: localisation, review and, for pages, the build and the hreflangAnnotation for language/region versions. work. Then add upkeep. Every localised piece has to be updated when the source changes, reviewed again and kept in its alternates set, so a piece that changes often costs far more over a year than one that is stable; the guide to keeping localised content in step with the source shows what that upkeep involves. Content that changes constantly and matters little is the clearest case for leaving out.
Plot value against effort
Put the non-must-have candidates on an impact-effort matrix: value from the job each does and the local demand, effort from the cost to localise and to keep current. High value, low effort goes first; high value, high effort is planned; low value, low effort is a filler for spare capacity; low value, high effort stays out. The plot makes the trade-offs visible to the market owner, who should agree them.
Choose the first wave and what stays out
The first wave is the must-haves plus the high-value, low-effort pieces, sized to what the reviewers can actually review before launch. Later waves are named, with their contents and the trigger that starts them, such as the market passing its gate. Everything else is explicitly out. Writing down what stays out matters as much as the wave itself, because unlisted content tends to be localised later by whoever has spare budget, without a decision. A launch that sends its whole package to the reviewers at once is one of the constraints the diagnosis page on localisation delaying every launch separates from unnamed reviewers, unbriefed suppliers and late local dates.
Decide how excluded content is handled
For each excluded type, decide what the market sees instead. Options include leaving it available only in the source-language edition, reached through a clear language link; not offering it in the market at all; or a short local page that summarises and points to it. What is not an option is mixing source-language pages into the local edition or publishing unreviewed machine translation, because each version should be in one language. Thinly localised pages also risk being read as Thin ContentPages that offer little original value of their own, such as near-empty templates, near-duplicates or copies with only a word or place changed., which helps no one.
Write the scope into the brief and review it at the gate
Record the first wave, the later waves and the exclusions in the localisation brief for the market, with the reason for each decision. The new market launch checklist confirms the must-haves are in place before traffic is bought. Review the scope at each market gate with the results: what local visitors look for and cannot find, which help articles support requests point to, which pages earn search traffic. The content governance guide owns the review cycle the localised pieces then join.
Must have, worth it, later, never: a scoping table
| Bucket | What belongs in it | Typical content | Decision rule |
|---|---|---|---|
| Must have | Content without which the market cannot understand the offer, buy or be served | Product and pricing pages, checkout, delivery, returns, contact, legal pages, the most-used help articles | In the first wave regardless of demand or cost |
| Worth it | Content with clear local demand or a strong job, at reasonable effort | Guides that answer what local searchers ask, category pages, onboarding emails | First wave if reviewers have capacity, otherwise the next wave |
| Later | Content with some value but high effort or uncertain demand | Long guides, video, sales collateral, the wider help centre | Named in a later wave with a trigger |
| Never | Content with little value in the market or upkeep out of proportion to its value | Blog archives, time-bound news, home-market events and promotions | Stays out; handled by the exclusion rule |
The buckets are decided per market. A guide that is "never" for a pilot market may be "worth it" once the market has passed its gate, and the review at each gate moves pieces between buckets with the evidence.
A filament maker scopes its first edition (illustrative)
Vollapix Filament scopes its Market A edition
Vollapix Filament, an invented maker of 3D-printer filament, prepares its first edition for Market A. Its inventory lists 85 product pages, 12 printing guides, 140 help articles, 260 blog posts and 18 email templates.
The must-haves come to the product pages, checkout, delivery and returns pages, contact, the legal pages and the 30 help articles that account for most support visits at home. Keyword research in Market A shows that local searches centre on print settings for particular materials, which 7 of the 12 printing guides answer, and almost nothing matches the blog's topics. The 7 settings guides go into the first wave; the other 5 go into the second.
The blog archive is left out of the edition entirely. A proposal to machine-translate it unreviewed "for search" is turned down, because unreviewed automatic translation published as indexable content conflicts with the search guidance the team follows and would add 260 pages to keep in step. The source-language blog stays reachable from a language link.
The remaining 110 help articles are scheduled for the second wave, triggered by the market passing its first gate. The first wave totals 85 product pages, 7 guides, 30 help articles, 9 transactional emails and the core pages, which the two in-country reviewers can review in the time before launch.
When local demand is added to each candidate piece, the SEO traffic estimator for international markets shows how to enter researched volume and a cautious position so pieces are ranked fairly.
Common mistakes
- Localising everything at once. The budget and the reviewers run out, and the edition launches half done. Scope a first wave.
- Filling gaps with unreviewed machine translation. A page is localised properly or left out of the edition.
- Translating the home priorities. Local demand often sits elsewhere. Use keyword research in the target language.
- Forgetting upkeep in the cost. Every localised piece has to be kept current. Content that changes often and matters little stays out.
- Leaving exclusions unwritten. Unlisted content gets localised later without a decision. Write down what stays out and how it is handled.
- Mixing languages in one edition. Each version stays in one language, navigation and content alike.
- Never revisiting the scope. Review it at each market gate with the market's own evidence.
The marketing localisation hub shows where scoping sits before the language and asset work, and the hreflang guide covers how each localised page joins its alternates once it is published.
