Skip to content
Tutorials Center
My learning

Topics

Lesson 38 of 60 · Manual Testing in Practice

Compatibility and Localization Testing

Test across browsers, devices and operating systems, and across locales, time zones and encodings. Build a compatibility matrix and see Intl formats vary by locale.

Your product will not run only on your laptop, in your language, in your time zone. Real users arrive with old phones and new ones, with five different browsers, with screens from 4 to 40 inches, and with languages that read right to left. Compatibility testing checks that the product works in the range of environments it claims to support. Localization testing checks that it works correctly for a given language and region, including words, dates, numbers, currency and culture.

Both are cheap to postpone and expensive to fix late, because they often reveal layout and data-handling decisions made on day one. This lesson shows how to plan the combinations sensibly (you cannot test them all, as The Seven Testing Principles reminded you) and what to look for.

Two related questions

Compatibility Does it work on the platforms we support?
Localization Does it feel right in each language and region?
1 / 2

Deciding what to support

Start from data, not guesswork. Look at your analytics (or the market you expect) and write down a support policy: which browsers and versions, which operating systems, which device classes, which languages. A policy might say "the latest two versions of each major browser, plus the oldest phone model we sell to".

Then turn it into a compatibility matrix: a table whose rows are test areas and whose columns are environments. You rarely run every cell. Instead:

  • Run your most important flows (sign-in, search, purchase) on the main environments, every release.
  • Run a wider smoke check on the rest.
  • Use risk to choose: features that touch the platform (camera, file upload, payments, printing, animation) deserve more environments than static text.
  • When the number of combinations explodes, use a combinatorial technique, which is the topic of Pairwise and Combinatorial Testing.
Flow Chrome (latest) Safari (latest) Firefox (latest) Android phone iPhone
Sign in full full smoke full full
Checkout full full smoke full full
Profile photo upload full full smoke full full
Help pages smoke smoke smoke smoke smoke

Compatibility checks

For each environment on your list, look for:

  • Layout: broken columns, overlapping text, missing icons, wrong fonts.
  • Features: a control that does nothing, a video that will not play, an upload that fails, a date picker that looks completely different.
  • Performance: a page that is fine on a laptop but crawls on an older phone with a slow connection.
  • Input methods: touch, keyboard, mouse, stylus; zoomed text; high-contrast mode.
  • Backward compatibility: a new version still opens old files, old accounts and old data.
  • Forward and cross-version behavior: an app update or an operating system update does not break it.

Real devices and cloud device farms are the strongest evidence. Browser developer tools can emulate screen sizes and slow networks quickly, but emulation does not reproduce a real rendering engine or hardware (mobile specifics are in Mobile App Testing Basics).

Localization: more than translation

Localization testing is often described with two terms. Internationalization (i18n) is designing the product so that it can be adapted: no hard-coded text, flexible layouts, full Unicode support. Localization (l10n) is the adaptation itself for a language and region. Testers check both: that the design allows it, and that each adaptation is correct.

Things that change with locale

  • Language and text: all visible text is translated, including errors, emails and PDF exports. Mixed-language screens are a defect.
  • Date and time formats: 03/05/2026 is 5 March in some countries and 3 May in others.
  • Numbers and currency: the decimal and thousands separators, the position and symbol of currency, even the digits themselves.
  • Sorting and search: alphabetical order differs between languages ("ä" sorts differently in German and Swedish).
  • Names and addresses: some people have one name, postcodes differ in shape, and phone numbers have varying lengths.
  • Direction: Arabic and Hebrew read right to left, so layouts, icons and arrows must mirror.
  • Culture: colours, icons, gestures, jokes and units (kilometers or miles, Celsius or Fahrenheit) carry different meanings.

Seeing locale differences

The first stand-in uses the standard Intl formatters to show how one number, one amount and one moment display in six locales, and one instant in four time zones. This is a stand-in for what a localized product should do; the exact characters (for example the kind of space used as a thousands separator) depend on the platform's locale data, so your console may print slightly different spaces or digits.

JS One value, many locales and time zones

Press Run (or Ctrl + Enter) to see the console output here.

No output.

Notice that the same instant, 23:30 UTC on 5 March, is already 6 March in Tokyo and Auckland. A "daily report" that groups by date can therefore put one event on different days for different users. Time zone and daylight-saving bugs are common enough to deserve test cases of their own:

  • Events near midnight, and at the end of a month or year.
  • A user whose device time zone differs from the server's.
  • The days when clocks change (a day with 23 or 25 hours, and a local time that does not exist or happens twice).
  • Stored times: are they saved in UTC and converted for display?

Use pseudo-localization early: replace text with accented characters and add about 30% extra length ("[Ŝàvé çhàñgéš !!!]"). Untranslated strings stand out immediately, and layouts that cannot grow are exposed before real translations arrive.

Encodings and special characters

Text is stored as bytes, and the encoding decides how characters map to bytes. UTF-8 is the common standard, but data can still be mangled by one component that assumes something else: a database, a CSV export, an email, an old API.

Typical symptoms are "mojibake" (café turning into café), question marks, empty boxes, or a field that rejects a name like "Zoë" or "O'Brien". Test with names and text containing accents, apostrophes, non-Latin scripts and emoji, and check that they survive the whole round trip: entered, saved, displayed, exported and emailed.

A related trap is length. "Number of characters" and "number of bytes" differ, and so does the idea of a character itself. The second stand-in measures both, shows the classic mojibake, and tests text expansion: translations are often longer than English, so a button sized for one word may overflow. A button sized for ten characters fits the English and German labels below, but not the Finnish or Russian ones.

JS Characters, bytes and text expansion

Press Run (or Ctrl + Enter) to see the console output here.

No output.

The emoji is one visible character but two UTF-16 units and four UTF-8 bytes, so a database column limited to bytes, or code that counts units, may reject input the user thinks is short. (The Latin-1 line depends on how your platform names that encoding; the effect is the point.)

A localization test checklist

  • Switch the language, then walk through every screen, not just the home page. Emails, error messages, validation texts and PDFs often stay in the default language.
  • Check text fit: truncated labels, overlapping elements, broken wrapping.
  • Check formats: dates, times, numbers, currency, phone numbers, addresses, units.
  • Try right-to-left languages: mirrored layout, correctly placed punctuation and numbers.
  • Use characters beyond ASCII in every text field, and in file names and search.
  • Test with another device region and time zone, not only the language.
  • Verify legal text, tax and currency rules for each market with someone who knows them. A tester cannot guess these.

Check your understanding

Quick check · 5 questions

of 5 answered You scored / 5 · perfect!

  1. 1. What is the main goal of a compatibility matrix?

    A matrix makes the test coverage of environments visible and lets you decide where to go deep and where a smoke check is enough.
  2. 2. A report shows an order placed at 23:30 UTC on March 5 as March 6 for a user in Tokyo. Is that a defect?

    The same instant falls on different calendar dates in different zones. The question is whether the product converts and groups consistently with what the user expects.
  3. 3. A German label fits a button, but the Finnish one is cut off. Which test idea finds this earliest?

    Pseudo-localization adds extra length and accents early, revealing layouts that cannot grow, before real translations exist.
  4. 4. A user saves the name 'Zoë' and later sees 'Zoë'. What is the most likely cause?

    UTF-8 bytes read as another encoding such as Latin-1 produce this kind of mojibake.
  5. 5. Which pair describes internationalization and localization correctly?

    i18n is the groundwork (no hard-coded text, Unicode support, flexible layout); l10n is the actual adaptation to a locale.

Summary

  • Compatibility testing checks supported browsers, devices and operating systems; a matrix and risk decide where to go deep.
  • Real devices give the strongest evidence; emulators are fast but not identical.
  • Localization covers language, formats, sorting, direction, culture and units, not translation alone.
  • Time zones and daylight-saving changes create date bugs around midnight and clock changes.
  • Test encodings with accents, non-Latin scripts and emoji across the entire round trip.
  • Use pseudo-localization to find hard-coded text and layouts that cannot grow.

Next up: API and Database Checks by Hand.