Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

International Marketing › Troubleshooting

Localisation Delaying Every Launch

Why every launch waits on localisation: unnamed reviewers, everything localised at once, suppliers briefed without terms and local dates compiled late.
Troubleshootingintermediate19 min readUpdated

First, map one launch's localisation path and the time each step waited

Take the most recent launch that slipped and rebuild its localisation path from the calendar, the supplier emails and the review threads: the date the master asset was final, the date each market's brief went to the supplier, the date each job came back, the date each in-country reviewer received it, the date they returned it, each further round, the sign-offs, and the date the edition went live. For every step, split the elapsed time into working and waiting. The theory of constraints says any chain that produces an output is limited by one step, that work piles up in front of that step, and that time or money gained anywhere else changes nothing at the end of the chain; its own example is a content line stuck in review, which hiring another writer does not speed up. The constraint is where the work waited longest and where the queue grows when more launches are added, and in most localisation pipelines that is a reviewer's inbox rather than a translator's desk.

Two readings save time before the table. First, decide whether the wait is in localisation at all. If the master asset was final late, or the brief was still changing while the suppliers worked, the constraint is upstream, and the campaign workflow from intake to launch owns it: building before the brief is agreed, reviewers seeing work too late, a queue in front of one gate or a person whose calendar decides every launch date. Second, decide whether the wait is production or review. The supplier's on-time delivery against the agreed date, from the business's own records, and the reviewer's log of changes per job tell you which: a supplier that delivers on time into a reviewer who returns it weeks later with the same term corrections has a review constraint fed by a briefing fault, not a supplier fault. A pipeline in one language that is slow for the same reasons belongs to the diagnosis page on content production that is too slow, which measures the waits before changing anything; this page is for the version of that problem that crosses a language and a border.

The framework's focusing steps then order the fixes. Identify the constraint from the map. Exploit it before spending: get more from the reviewer's hours by sending work they can approve rather than correct, and by sending less of it. Subordinate the other steps to its pace, which means briefing no more than review can clear and dating launches from the reviewer's capacity rather than the supplier's. Elevate it only then, with a second named reviewer or agreed hours. And repeat, because once review clears, the limit moves to the brief, the supplier or the calendar. If the localised output arrives on time and reads wrong, that is a different page: the diagnosis of translated copy that reads wrong covers it, and if the launch lands on its date and the campaign is flat in some markets, the page on a global campaign underperforming in some markets does. The international marketing troubleshooting collection lists all three beside the market questions.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
The longest wait on the map is between a job arriving with a market and the market's comments coming back, and nobody can say who is reading itWhether a reviewer is named per market and content type, whether their time is agreed with their manager and recorded, and whether a turnaround was ever written downIn-country reviewers unnamed, unbriefed or reviewing in their spare time with no agreed turnaroundName one reviewer per market and content type, agree and record their hours, brief them on the brand, termbase and style guide, and separate the language, commercial and compliance sign-offs
Every launch sends the complete package, release notes, help articles, campaign and all, and the reviewers run out of time before the dateThe scope of the last launch against what a customer in the market needed to buy and be served, and whether a later wave was ever namedEverything localised at once instead of a first wave the reviewers can actually reviewSort the package into must have, worth it, later and never, size the first wave to the reviewers' capacity, name the later waves with a trigger and write down what stays out
Each round comes back with the same corrections to product names and terms, and the reviewer's log shows repeat changesWhat travelled with the job: the brief, the termbase and translation memory, the style guide, screenshots or staging links and the review routeJobs sent without the brief, termbase and style guide, so every round comes back with the same correctionsSend every job with the full pack, have reviewers correct the termbase as well as the asset, and keep the master files with the business
Urgent lines wait behind copy refreshes, and a reviewer reading a batch finds a price or safety change inside itThe change log for the source: whether each change records the editions it touches and a risk triage, and whether urgent changes travel as separate jobsChanges to the source not triaged by risk, so a typo and a safety line wait in the same queueLink every source page to its editions, log each change with the editions it touches, send price, claim, legal and safety changes to every edition at once and batch the rest on an agreed cycle
A market's edition lands in its own peak trading weeks, its reviewer's holiday or a quiet period nobody at the centre knew aboutWhere the market's dates came from, whether lead times were worked backwards from the business's own past jobs, and whether the clash check was runLocal dates, holidays and freezes compiled at head office after the timeline was setAsk each market owner for its moments, freezes and reviewer availability before the timeline, work lead times backwards from past jobs and run the clash check
Markets return versions that change the idea, the identity or a cleared claim, and the central team sends them round againThe master brief: whether each element is marked fixed, open or adapt with approval, and whether the toolkit carries editable assets and an adaptation guideA master brief that never marked what was fixed, so markets relocalised the idea itselfMark every element, build the toolkit, brief all markets together, collect local plans in one format and name one accountable person per decision per market

Fix, verify and prevent each cause

Reviewers were never named or given time

The translation and transcreation guide describes in-country review as the step where a native speaker who knows the market reads the asset in its final context and confirms it works, distinct from the translator's own check against the source; the glossary entry defines it the same way. The guide then names the fault directly: market colleagues in sales or support are often asked to review in their spare time, which is why review becomes the bottleneck. Reviewers should be named per market and per content type, briefed on the brand, the termbase and the style guide, and given a turnaround that the launch plan respects; if internal reviewers are used, their time should be agreed with their managers and recorded as part of their role. Sign-off has three parts, language, commercial and compliance, each explicit, recorded and owned by one person, and the rollout tutorial puts those three names into its responsibility chart with a deadline in each person's own clock.

Fix: for each market, name the reviewer for each content type, agree their hours with their manager and write the agreement down, brief them once on the brand, the termbase and the style guide, and record who gives each of the three sign-offs.

Verify: the next job's review comes back within the recorded turnaround, the sign-off record carries one name per sign-off, and the map of the next launch shows the review wait shorter than the production wait.

Prevent: budget reviewer hours as part of each market's role rather than as a favour, carry each reviewer's holidays on the calendar, and use the same reviewer to judge supplier test pieces, so the person who will read the work has chosen who produces it.

Everything was localised at once

The tutorial on deciding which content to localise first names three things that go wrong when a business sets out to localise everything: the budget runs out partway, review becomes the bottleneck because in-country reviewers cannot read an estate in the time a launch allows, and every localised page becomes a page to keep current. A launch package has the same shape as an estate. Release notes, help articles, product pages, emails and a campaign sent together are more than any reviewer can read before the date, and the tutorial's scoping table sorts them: must have, the content without which a customer in the market cannot understand the offer, buy or be served; worth it, where there is clear local demand at reasonable effort; later, named with a trigger; and never, with the exclusion handled deliberately. The tutorial also keeps each version of a page in one language, navigation and content alike, so a half-localised edition is not a way of launching early.

Fix: inventory the launch package, mark the must-haves, add local demand to the rest, estimate the cost to localise and to keep current, and size the first wave to what the named reviewers can read before the date; write the later waves with their triggers and the exclusions with how each is handled, into the localisation brief.

Verify: the first wave is reviewed in context before launch, nothing in the edition is half-translated or unreviewed machine output, and the second wave starts on its trigger rather than on spare budget.

Prevent: make scope a field in every launch brief, review it at each market gate with what local visitors look for and cannot find, and treat a package with no later wave as unscoped.

Jobs went out without the brief, termbase and style guide

The guide to choosing and managing translation suppliers lists what every commission carries whoever the supplier is: the brief, the termbase and translation memory with the instruction to use approved terms and return new ones as proposals, the language's style guide, context in the form of screenshots or staging links, and the review route. A job without the pack, it says, produces work that has to be corrected in review, and the correction is then charged to the reviewer's time, which is usually the scarcest resource in localisation. The guide's own supplier measures show the fault in the reviewer's log: a rising count of term corrections usually means the termbase is not being used, and the same correction twice suggests a missing rule or a supplier not reading the reference files. The transcreation guide adds that reviewers should correct the termbase when they change a term, not only the asset, or the same correction will be needed on every future job.

Fix: assemble the pack once per market and language, send the current termbase and memory with every job, label each asset translation, localisation or transcreation so the supplier knows what is being bought, ask the supplier to say in writing where machine translation is used and who post-edits it, and route every reviewer correction into the termbase.

Verify: the reviewer's log for the next two jobs shows term corrections falling and no repeat corrections, and queries from the supplier arrive before delivery rather than corrections after it.

Prevent: choose suppliers with paid test pieces that carry the same pack and are judged blind by the in-country reviewer, review suppliers on a regular cycle against the log, and keep the master termbase, memory and style guide in the business's control so a change of supplier starts from the same files.

Source changes were not triaged by risk

The guide to keeping localised content in step starts from a record that links every source page to its editions and a change log with one extra field, which editions carry this page, so that a change to a page in three editions creates three pending updates. It then triages each change by what a stale edition would tell its market: price, offer, claim, legal and safety changes travel to every edition at once as one job with the same brief and deadline; product facts go in the next batch or at once if customers would act on them; copy refreshes are batched on a regular cycle; a slip travels only if the edition has the same slip. Treating every change as urgent, the guide warns, overwhelms the reviewers, and urgent changes then queue behind refreshes. Its example is exactly the launch fault: a safety warning sent without the urgent flag and queued behind a copy refresh. Google's pages on localized versions and multi-regional and multilingual sites, as the hreflang guide states them, ask alternates to be reciprocal and to point at live pages, which is why a page retired in one edition leaves every edition's alternates in the same release rather than waiting its turn in the queue.

Fix: build the record, log every source change with the editions it touches and its triage, send urgent changes as separate jobs to every edition together, batch the refreshes on a cycle the market owner and reviewers agree, and put each approved change into the termbase and memory so the old line does not come back.

Verify: the drift check finds no edition behind on a price, claim, legal or safety line, the review queue holds only batched refreshes, and the map of the next launch shows no urgent change waiting behind a refresh.

Prevent: run the drift check on a cycle with one named owner, treat anything found behind on a high-risk line as urgent from that moment, and name in a responsibility chart who triages, who briefs and who reviews for each edition.

Local dates and freezes were compiled at head office

The tutorial on building a marketing calendar for several markets asks each market owner for the holidays, cultural moments, shopping events, school terms and category moments that matter there, with the dates and a line on why, and says plainly not to compile that list at head office from general sources, because the market knows which moments its customers notice. It works localisation and review lead times backwards from each live date using the business's own past jobs rather than a general allowance, marks freezes and quiet periods per market, including periods when the local team or partner is unavailable, and runs a clash check for a global moment that lands in a market's freeze and for a localisation milestone that falls on a market's holiday, when the reviewer is away. That backwards step, it notes, is the one that most often moves a date, because it shows that a market's edition has to be briefed earlier than the central plan assumed.

Fix: before the next timeline is set, collect each market's moments, freezes and reviewer availability with the source recorded beside each date, date the brief, translation, review and sign-off milestones backwards from each market's live date with that market's own lead times, run the clash check, and record any market that launches later than the others and why.

Verify: every launch item carries localisation milestones with owners in the market's clock, the next edition lands outside the market's freeze, and the reviewer is in the office when the job arrives.

Prevent: review the calendar with every market owner each quarter, record the lead times from the jobs just finished, and track localised editions that went live with the global launch as the calendar's own measure.

The master brief never said what was fixed

The rollout tutorial lists the ways a campaign loses its shape between markets, and the first is a brief that does not say what is fixed: markets guess, and the guesses differ, with one keeping a headline that does not work in its language and another changing the idea because it could. Its first step marks every element of the master brief fixed, open or adapt with approval, with the idea, the proposition, the identity and the cleared claims usually fixed and the headlines, casting, offers, channels and timing usually open. The toolkit then carries editable master assets, an adaptation guide with examples, the claims list, the copy deck marked translate or transcreate and the timeline, and is tested by asking one market owner to read it cold and list their questions. All markets are briefed together, each returns a local plan in a fixed format, and a responsibility chart names one accountable person per decision per market.

Fix: rewrite the master brief with the three-way marking, build and test the toolkit, run one recorded briefing, collect local plans, decide exception requests quickly and set a launch order that lets the most prepared market find problems in the master assets once.

Verify: local plans show adaptation only on open elements, a cold read of the toolkit raises no questions the brief should have answered, and no market's version comes back for changing a fixed element.

Prevent: hold one review after each campaign on which fixed elements caused trouble and which toolkit gaps cost time, and change the next brief accordingly, so the marking improves with every rollout.

Prevention: named reviewers with budgeted time, a scoped first wave and triage before the queue

Each cause above is a decision made for the reviewer by default, and the prevention is to make the reviewer's hours the measure the whole pipeline is paced to. Start with the people. Every market has a named reviewer per content type with agreed, recorded hours and a stated turnaround, and the three sign-offs, language, commercial and compliance, each have one owner. Their holidays and quiet periods sit on the calendar beside the market's freezes. When the map shows review still limiting the throughput after the pack, the scope and the triage are fixed, the answer is a second named reviewer with hours of their own, which is the framework's elevate step taken after the exploit step rather than instead of it.

Then pace the work to those hours. A launch package is scoped before it is briefed, into a first wave the reviewers can read before the date and later waves with triggers, and the scope is written into the localisation brief alongside the market, the approach for each asset, the tone, the terms, the keywords, the local checks and the named reviewers, so the brief travels with every job and nothing arrives at a supplier without it. The termbase and translation memory go out current and come back with proposals, and every correction a reviewer makes goes into the termbase so it is made once. Source changes are logged with the editions they touch and triaged before they join any queue, with urgent lines travelling together and refreshes batched, and the drift check runs on its cycle with its owner.

Then set the dates from the markets. The calendar carries each market's own moments and freezes from the market owner, with the source recorded, and every launch item's localisation milestones are dated backwards from the market's live date with lead times from the business's own past jobs. The clash check runs before the timeline is published, and a market that cannot be ready launches later in a recorded sequence rather than half-localised on the shared date. A campaign that travels carries a master brief with every element marked fixed, open or adapt with approval, a tested toolkit and one accountable person per decision per market.

Finally, measure the chain on one output. The theory of constraints asks for a single measure that judges every change, and for localisation it is whether localised editions went live with the global launch or whether items launched on their planned date, read at the quarterly calendar review with the lead times from the jobs just finished. Local measures, words delivered or jobs sent, reward activity at steps that do not set the total. When the measure moves, map the next launch again, because the limit will have moved with it, usually to the brief, the supplier or the calendar. The marketing localisation hub shows where scoping, suppliers, review and upkeep sit in one workflow, and the campaign workflow guide's gates catch the upstream faults, a brief still changing or a master asset final late, that no localisation fix can recover.

A cycling computer maker finds the queue was one reviewer's evenings (illustrative)

Illustrative example — not a real result

Pettifogle Cycle Computers

Pettifogle Cycle Computers, an invented maker of cycling computers, launches firmware features in 4 markets and the last 5 launches each slipped by 3 to 6 weeks at localisation. The operations lead maps the most recent one. The whole package, 2 release notes, 14 help articles, 6 product page updates and a campaign with 9 assets, went to the suppliers on the same day. Market B's supplier delivered in 8 working days; Market B's review took 31, because it was done by 2 sales staff in their evenings with no agreed time. Market D's reviewer returned the same 11 term corrections for the third launch running, because the supplier had never been sent a termbase. A price line changed in the source during the job and waited 19 days inside a batch of copy refreshes. Market C's edition went live on the shared date, which fell in the second week of its summer freeze, so its campaign ran to nobody.

The fix starts at the constraint. Market B's review is moved to one named reviewer with 6 hours a week agreed with her manager and a 5-working-day turnaround, and Market D's reviewer gets the same arrangement. The next launch is scoped: 2 release notes, the 5 help articles that answer most support contacts, the product page updates and 3 campaign assets go in the first wave, 9 help articles and the remaining assets in a second wave triggered by the first wave going live. Every job now carries the brief, a termbase of 340 terms and the translation memory, and reviewers correct the termbase as they go; Market D's term corrections fall to 2 on the first job and none on the second. Source changes are logged with the editions they touch and triaged, so the next price change reaches all 4 editions as one job in 3 days. The 4 market owners supply their moments and freezes before the timeline, and Market C's launch is dated 2 weeks after the others. The next launch lands on its date in 3 markets and 2 weeks later in Market C, as recorded, and the operations lead maps it again: the longest wait is now 4 days at the supplier for Market A, which is where the next quarter's attention goes.

Common mistakes

  • Adding a supplier when the constraint is review. More files arrive at the same reviewer; the launch slips by the same amount at a higher cost.
  • Reviewing in spare time. Review by whoever has a free evening becomes review by nobody. Name reviewers and record their hours.
  • Sending the whole package. A first wave the reviewers can read before the date launches; an estate does not.
  • Jobs without the pack. Without the brief, termbase and style guide, the supplier guesses and the reviewer corrects, on every job.
  • Corrections that never reach the termbase. The same term comes back wrong on the next launch.
  • Treating every source change as urgent, or none. Urgent lines queue behind refreshes, or refreshes masquerade as urgent. Triage by what a stale edition would tell its market.
  • Compiling the markets' dates at head office. The list misses the freeze the market would have flagged, and the reviewer's holiday with it.
  • A brief that does not say what is fixed. Markets relocalise the idea and the central team sends it round again.
  • Not mapping the next launch. Once review clears, the limit moves. Map again before investing further.

Frequently asked questions

Should we add another translation supplier to speed things up?

Not until you have mapped where the last launch actually waited. Capacity added at any step other than the constraint changes nothing at the end of the chain; it only moves work faster into the queue in front of the limit. In most localisation pipelines the limit is review, so a second supplier delivers more files to a reviewer who already cannot clear the first supplier's, and the launch slips by the same amount with a larger invoice. Get more from the constraint first: send every job with the brief, termbase and style guide so the reviewer corrects less, scope the first wave to what the reviewer can read, and triage source changes so refreshes stop queuing in front of urgent lines. Only if review still limits the throughput does a second named reviewer, with agreed and recorded time, add capacity where it counts.

Can we launch the market on the source-language version and localise afterwards?

Not by mixing languages into the market's edition. Each version of a page stays in one language, navigation and content alike, and a half-localised edition is neither a launch nor a scope decision. The honest options are the ones the scoping tutorial sets out for excluded content: keep the piece available in the source-language edition behind a clear language link, leave it out of the market for now, or publish a short reviewed local page that summarises and points to it. For a campaign, the rollout tutorial's answer is a launch order rather than a shared date: a market whose first wave is not yet reviewed launches when it is, and the record says why. What never works is publishing unreviewed machine output to fill the gap, because the reviewer then has to read it anyway, later, under more pressure.

How will we know the constraint has moved once review is fixed?

Map the next launch the same way and look for the new queue. Relieving one step moves the limit somewhere else, so once review clears on time the wait usually appears at the brief, at a supplier, or at the calendar, and a team that keeps investing in review after it has stopped being the limit spends effort for no gain. Judge every change on one output measure for the whole chain, such as localised editions live with the global launch or items launched on their planned date, rather than on local measures like words delivered or jobs sent. The quarterly calendar review with the market owners is the natural place to read it, with the lead times from the jobs just finished recorded beside it.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Managing multi-regional and multilingual sites — Google Search Central
  2. Tell Google about localized versions of your page — Google Search Central

Related