How to Prepare a Mobile App for Localization

Choose Markets Before Choosing Languages

For each proposed market, confirm:

• The intended audience and business objective.
• The app features, payments, subscriptions, and support that will be available.
• Regulatory, privacy, contractual, or market-specific requirements.
• Product terminology already used in that market.
• Whether the release will be complete, limited, phased, or a pilot.
• Who will review and approve local language.

A language may be spoken across many countries, but one translation is not automatically suitable for every market.

Specify Locales, Not Only Languages

A locale combines a language with a region and, when needed, a script or other configuration. Examples include es-MX, fr-CA, pt-BR, and zh-Hant-TW. Locale decisions can affect terminology, tone, currencies, measurements, legal language, date formats, app-store assets, and user expectations.

Decide Where Translations Can Be Shared

A shared version may be appropriate when:

• Product terminology is understood consistently.
• Functionality is the same across markets.
• No market-specific legal or regulatory language is required.
• Tone and formality align.
• Regional stakeholders agree on the shared version.

A market-specific version may be preferable when terminology, regulation, product positioning, formality, store optimization, or customer expectations differ materially.

Define Locale Fallbacks

Fallback behavior determines what users see when an exact locale match is unavailable. A product might use Spanish for Mexico first, a broader Latin American Spanish resource second, and the source language as the ultimate default. The hierarchy should be intentional and tested rather than left to chance.

Choose a Rollout Model

Externalize All User-Facing Content

Store customer-facing text in localizable resources rather than embedding it directly in source code. Review navigation, buttons, forms, validation, errors, onboarding, empty states, search, permissions, purchases, notifications, widgets, shortcuts, accessibility labels, server messages, and embedded web content.

For technical support beyond this readiness guide, explore Stepes’ Software Internationalization Services.

Use Stable, Meaningful Resource Identifiers

Stable keys connect source text to translations, context, review history, screenshots, and future updates. Prefer identifiers such as checkout.payment.confirm_button over temporary labels such as button_07. Do not reuse one key for unrelated meanings merely because the current English text is identical.

Avoid Sentence Concatenation

Building sentences from separate fragments often fails in languages with different word order, agreement, gender, case, or plural behavior.

"You have " + count + " new messages"

messages.new_count = "You have {count} new messages"

Support Plurals and Grammatical Variations

English commonly distinguishes singular and plural, while other languages may require additional forms. Support cardinal and ordinal plurals, gender or grammatical agreement, select statements, formal and informal variants, and feature-specific variations where the product requires them. Unicode CLDR documents plural categories such as zero, one, two, few, many, and other. See the Unicode plural-rules reference.

Apple String Catalogs support plurals and variations, and Android provides dedicated plural resources. Platform-specific implementation details belong in the dedicated iOS App Localization Guide and Android App Localization Guide.

Use Locale-Aware Formatting

Use platform formatters and reliable locale data for dates, times, time zones, numbers, percentages, currencies, measurements, addresses, phone numbers, names, lists, sorting, and collation. A translated label does not correct an incorrectly formatted amount or date.

Support Unicode Across the Full Product Stack

Check databases, APIs, search, input validation, authentication, user profiles, notifications, analytics, exports, reporting, support tools, and third-party integrations. A mobile interface may display a script correctly while a backend system rejects the same characters.

Review Fonts and Glyph Coverage

Confirm character coverage, script legibility, fallback behavior, weights, styles, diacritics, combining characters, line-height needs, and accessibility scaling with representative content rather than the source language alone.

Design for Text Expansion and Contraction

Review fixed-width buttons, navigation tabs, dialogs, forms, cards, tables, charts, compact headers, landscape layouts, tablets, and accessibility text sizes. There is no universal expansion percentage that guarantees a safe interface. Flexible components and realistic testing are more reliable than one fixed allowance.

Prepare for Right-to-Left Interfaces

Right-to-left support affects layout direction, navigation, icons, progress indicators, carousels, tables, forms, mixed-direction content, numbers, product names, punctuation, gestures, animations, and third-party components. Use direction-aware start and end concepts rather than hard-coded left and right assumptions where the technology permits. See the W3C internationalization techniques.

Define App-Level Language Behavior

Decide whether the app follows the device language, a system-level app-specific setting, an in-app language selector, an account-level preference, or a combination. Define first-launch behavior, persistence, synchronization, restart requirements, active-transaction behavior, and how server-delivered content follows the same choice. Android 13 and later support centralized per-app language preferences. See the Android app-language guidance.

Include External and Third-Party Content

Inventory content from content management systems, remote configuration, APIs, notifications, email, SMS, chatbots, downloadable documents, embedded sites, authentication providers, payments, maps, analytics, consent tools, and support platforms. Assign ownership for every source.

In-App Interface Content

• Menus, navigation, buttons, forms, instructions, validation, and errors.
• Search, filters, tooltips, empty states, onboarding, account settings, and security messages.
• Subscription, cancellation, payment, confirmation, help, and support flows.

Device and Operating-System Content

• Permission descriptions, app name, home-screen labels, notifications, widgets, and shortcuts.
• Voice-assistant content, accessibility labels, deep-link destinations, and platform prompts.

Commercial and Transactional Content

• Subscription tiers, pricing descriptions, in-app purchases, promotions, trials, and referrals.
• Payment confirmations, cancellation guidance, refunds, loyalty, and reward content.

App-Store Content

Prepare the app name, subtitle or short description, full description, keywords where applicable, promotional text, release notes, screenshots, preview videos, in-app product names, and market-specific creative assets. Apple and Google Play both support localized store content and visual assets. Review the App Store Connect localization guidance and Google Play localization guidance.

Media and Supporting Content

• Images containing text, illustrations, video, voiceover, subtitles, captions, and animation.
• Help centers, FAQs, legal notices, privacy policies, marketing pages, email, and SMS.

Preserve Native Resource Structures

iOS resources may include String Catalogs and .xcstrings, .strings, .stringsdict, XLIFF packages, interface content, localized assets, and App Store Connect exports.

Android resources may include XML strings, arrays, plurals, locale-specific values directories, Jetpack Compose resources, app labels, and Google Play exports.

Cross-platform resources may include Flutter ARB, JSON, YAML, JavaScript or TypeScript resources, RESX, PO, Java properties, XLIFF, CSV, or another assessed structured format.

Apple recommends String Catalogs for modern Xcode localization, with support for translations, comments, plurals, and variations. See the Apple String Catalog documentation.

Preserve Contextual Metadata

Mark Nontranslatable Content

Identify brand and product names, file paths, URLs, API values, analytics identifiers, code, markup, keyboard shortcuts, model numbers, legal names, and protected trademarks. “Do not translate” is most useful when accompanied by the reason and usage rule.

Use Spreadsheets Carefully

A spreadsheet can work for a small or manually managed project, but it should preserve keys, file mapping, placeholders, comments, character limits, plural relationships, version history, approval status, and import requirements. Copying interface text into one unstructured column removes the relationships needed for safe reintegration and future updates.

Remove Obsolete Content and Control Source Changes

Remove deprecated features, test strings, development notes, hidden experiments, abandoned keys, internal-only labels, and content no longer included in the release. Then define how new, deleted, changed, and emergency strings will be tracked during localization.

Identify Protected Elements

• Variables, placeholders, format specifiers, and ICU MessageFormat syntax.
• HTML, XML, Markdown, escape sequences, resource references, and newlines.
• URLs, product codes, commands, analytics events, and nontranslatable tokens.

Explain Variables, Not Just Their Syntax

In Welcome, {name}, the translator should know whether the value is a given name, full name, company, display name, or system-generated label. The inserted value may affect punctuation, formality, gender, word order, or grammatical case.

More complex messages such as {user_name} added {count} items to {list_name} may require plural logic, changed sentence order, or additional grammatical handling.

Validate Technical Elements Automatically

Where the format permits, compare source and target content for missing or added variables, changed names, invalid types, unbalanced tags, altered markup, incorrect escaping, missing plural branches, changed numbers, and character-limit violations. Automated checks protect technical integrity, but they do not confirm linguistic quality or in-context suitability.

Explain Why the String Exists

Terms such as Order, Apply, Home, Charge, Book, Current, and Save can carry different meanings depending on the screen, user action, and product domain. The translator needs the function, not only the source word.

Supply Screenshots and Screen References

Show the complete relevant screen, identify the translatable element, include useful surrounding content, represent different states, and map the screenshot reliably to resource keys. A screenshot without key mapping may still leave the linguist guessing.

Share Designs, Prototypes, and User Flows

Figma files, clickable prototypes, design-system components, user-flow diagrams, storyboards, redlines, screen inventories, and responsive behavior help linguists understand meaning and help engineers identify layout risk.

Provide Representative Builds

Depending on security and product maturity, provide TestFlight access, Android test builds, staging environments, test accounts, sample transactions, feature flags, representative user roles, and regional configurations. Build access reveals navigation, timing, and interaction that a file export cannot show.

Document User Intent and Establish a Query Process

Explain what the user is trying to accomplish, what happened before the message, what happens after it, whether the action can be reversed, and whether the content is instructional, promotional, transactional, legal, medical, or safety-related. Manage questions in one shared log with clear owners, response expectations, and reusable decisions.

Build a Product Terminology List

Prioritize product and feature names, navigation, subscriptions, industry terms, regulated language, privacy and security terms, words that remain in English, prohibited wording, abbreviations, and the distinction between customer-facing and internal terminology.

Define Product Voice

Document formality, tone, concision, technical depth, contractions, sentence fragments, capitalization, punctuation, user address, inclusive-language expectations, error-message style, button-label conventions, and use of humor. A marketing banner, payment error, privacy notice, and medical warning may require different guidance.

Coordinate Language Across Channels

Align the app, website, SaaS platform, documentation, help center, support, app-store listing, email, notifications, marketing, and legal content. Users should not see one term in the store, another during onboarding, and a third in customer support.

Reuse Approved Language Assets

Review translation memories, glossaries, previous releases, website translations, software content, documentation, support content, and regional marketing. A Translation Memory can store approved source and target segments for suitable reuse, but context and validation still determine whether a historical translation fits a new use.

Involve Regional Reviewers Early

Define reviewer responsibilities, deadlines, approval authority, and the difference between an error and a stylistic preference. Record approved decisions in the glossary and translation memory so the same question does not return in every release.

Separate Text From Images

Keep text editable, use layered source files, preserve fonts and source assets, place language in overlays rather than flattened graphics, and maintain templates for localized screenshots. This reduces manual recreation and makes future updates easier.

Review Visual Content for Market Suitability

Evaluate icons, gestures, maps, flags, people, clothing, food, currencies, addresses, documents, cultural references, medical or legal imagery, and directional symbols. Base adaptations on real market requirements rather than assumptions or stereotypes.

Prepare Localized Screenshots and App Previews

Store screenshots should reflect the localized product users will install. Avoid pairing translated descriptions with source-language screens unless the product intentionally operates that way. See Apple’s localized screenshot and app-preview guidance.

Inventory Audio and Video

Identify voiceover, spoken prompts, tutorial narration, subtitles, captions, on-screen text, animation timing, audio descriptions, accessibility transcripts, and preview videos. The appropriate method may involve subtitling, dubbing, re-recording, text replacement, or a combination.

Include Accessibility Content

Prepare screen-reader labels, hints, alternative text, voice-control names, form instructions, error descriptions, captions, transcripts, and accessible names for icons and controls. Keep accessibility text connected to the related interface element and user action. See the Android accessibility guidance.

Test Enlarged and Flexible Text

Review dynamic text settings, font scaling, screen-reader focus, multiline labels, small screens, tablets, landscape orientation, split-screen layouts, and adaptive or foldable interfaces where supported. A translation that fits at the default size may fail when users enlarge text.

Run Pseudolocalization

Pseudolocalization replaces source strings with simulated content that can expose hard-coded text, missing resources, expansion, truncation, unsupported characters, encoding problems, fragile sentence structure, and right-to-left issues before real translations are available. See the Android pseudolocale documentation.

Test Expansion and Right-to-Left Behavior

Use expanded pseudo-text for buttons, navigation, dialogs, forms, notifications, tables, cards, charts, compact screens, orientation changes, and accessibility sizes. Use an RTL pseudolocale to review screen direction, icons, carousels, progress indicators, forms, mixed-direction text, product names, charts, and third-party components.

Detect Missing and Hard-Coded Text

Combine static analysis, build-time checks, runtime inspection, resource comparison, screenshots, automated UI tests, and manual flow review. Include states triggered by errors, permission denial, account changes, and unusual transactions—not only the primary happy path.

Validate Placeholders, Resources, and Locale Formatting

Confirm that placeholders are valid, plural branches compile, files pass syntax checks, markup is balanced, escaping is correct, defaults are complete, locale identifiers are valid, and builds load the expected resources. Then change the locale and review dates, times, currencies, numbers, measurements, addresses, names, sorting, search, and input validation.

Select the Content Exchange Model

Options include secure manual exchange, a customer portal, translation management system, repository integration, localization API, automated build pipeline, or a hybrid workflow. The right model depends on release frequency, resource architecture, security, language count, review requirements, and engineering capacity.

Define Approval Stages

Define Review Authority and Security Requirements

Clarify who may request changes, approve terminology, resolve disagreements, accept known limitations, and provide final approval. Document access to unreleased content, repositories, test credentials, personal or regulated data, production environments, file retention, confidentiality, regional restrictions, and approved AI use. Use test or synthetic data when real customer data is not necessary.

Plan AI-Assisted Translation Responsibly

Define eligible content, approved models and environments, data rules, terminology and translation-memory controls, context, automated checks, professional review, specialist validation, and final approval. High-visibility, brand-sensitive, transactional, legal, medical, privacy, safety, and context-dependent strings generally need stronger human oversight than repetitive low-risk content. Stepes’ AI Translation Services provide additional guidance on governed AI + Human Translation Workflows.

Platform toolchains are also adding AI-assisted capabilities. Apple now documents agent-based workflows that can add languages, update String Catalogs, translate strings, and add plural variants. Treat these tools as accelerators rather than substitutes for readiness, security, terminology, context, technical QA, and accountable approval. See Apple’s agent-assisted localization documentation.

Prepare for Continuous Localization

Frequent releases need a defined method for new and modified strings, branches, translation-memory reuse, terminology updates, job creation, review routing, build readiness, quality gates, pull requests, hotfixes, and deprecated content. Explore Stepes’ Software Localization API and the dedicated Continuous Localization for Mobile Apps guide.

Include Every Required Stage

• Locale confirmation, source cutoff, resource export, translation, and linguistic review.
• Regional review, engineering integration, localized build creation, and in-context validation.
• Functional and locale testing, correction, retesting, regression, and final approval.
• Store metadata, screenshots, previews, submission, launch monitoring, and post-release support.

Do not schedule the first localized build immediately before store submission. Testing needs enough time to identify, resolve, integrate, and retest defects.

Coordinate Connected Release Assets

Treat in-app strings, store listings, screenshots, release notes, purchase descriptions, support content, documentation, marketing, launch email, legal notices, and customer-service readiness as one connected market experience.

Choose Simultaneous or Phased Release

A simultaneous launch may be appropriate when markets and announcements are tightly linked and sufficient testing capacity is available. A phased release may be safer when product readiness, regional review, legal requirements, store timing, or QA capacity differ by market.

Reserve Time for Corrections and Post-Launch Ownership

A localized defect may require changes to language, source text, interface design, resource structure, formatting, third-party content, screenshots, or support material. Assign owners for user-reported issues, store feedback, emergency corrections, new features, terminology, translation memory, and future release synchronization.