Belief: “Localisation testing means checking the translations.” Translation is one piece of it. The bigger part is checking that the software’s hidden assumptions, which were usually written by someone with a particular country in mind, do not fail for the people actually using it. Here are the assumptions that most often break in India.
Names do not have two parts
Belief: everyone has a first name and a last name. Many people in India, especially in parts of South India, use a single name, or an initial followed by a given name, such as a father’s or village’s initial in front. A form that insists on a last name forces users to type a dot or repeat the first name, and that garbage then lands on a bank statement or an airline ticket where it must match an ID. Test with a one-word name, a name with initials and spaces, a very long name, and a name with characters like apostrophes and hyphens. Also check whether the name on the screen is the same one printed on the PDF invoice.
Scripts are not just letters
Belief: one character equals one letter, so “name length 50” is simple. In Devanagari, Tamil, Telugu and other Indic scripts, what a reader sees as one syllable can be several Unicode code points joined together: a consonant, a vowel sign, sometimes a virama joining two consonants. A field limited to 20 characters may allow far fewer visible syllables than the user expects, and code that truncates by character count can cut a syllable in half and display broken marks. Programming languages also count differently; Java’s string length, for example, counts UTF-16 units, not what a person sees as a character.
Test by entering real text in the scripts your users use, not placeholder characters. Check that the database stores it correctly (a mis-set character encoding turns it into question marks), that it survives export to CSV and opens correctly in Excel, that sorting is sane, and that fonts on low-end Android phones actually contain the glyphs. A missing glyph shows as an empty box.
Numbers, money and the lakh
Belief: thousands are grouped in threes. In the Indian system the last three digits are grouped, then pairs: 12,34,567 is twelve lakh thirty-four thousand five hundred sixty-seven. Many formatting libraries support this through the English-India locale, but a hard-coded format will print 1,234,567, and a user reading a loan amount in that form pauses to do arithmetic. Check amount fields on screens, in SMS, in PDFs and in exported reports, because each may be produced by different code.
The rupee sign ₹ has its own Unicode character, added in 2010, so check that older fonts and older systems do not render it as a box, and that “Rs.” and “INR” variants are used consistently. Test rounding: GST computed per line item versus on the invoice total can differ by a paisa or two, and your finance team must say which is correct. Words-in-figures output (“Rupees One Lakh Only”) on cheques and invoices deserves its own test cases for values like 100000, 100001 and 99999.99.
Dates, time and the single zone
Indian Standard Time is UTC plus 5 hours 30 minutes, and India does not observe daylight saving. Two consequences. First, a half-hour offset breaks code that assumes whole hours. Second, products that serve India and other countries often convert in both directions, so test a user in IST looking at a deadline set by a server in UTC, and the midnight boundary: an order placed at 11:45 p.m. IST on the 31st is already the 1st in some other time zones and, in UTC, falls on the 31st at 6:15 p.m. Date format also matters. 03/04/2026 is 3 April in India and 4 March in the United States, so unambiguous display (3 Apr 2026) is safer, and parsing must not assume the month comes first.
Addresses and phone numbers
- PIN codes have six digits and do not begin with zero. A validator copied from another country’s postal code rules will reject or accept the wrong things.
- Addresses often have landmarks (“near the temple”), no street number, and varied ordering. Required fields for “street” and “building number” cause drop-offs. Check what the delivery partner’s system really needs.
- Mobile numbers are ten digits, currently starting with 6, 7, 8 or 9, with +91 or a leading 0 in the way users type them. Accept pasted spaces and dashes, normalise on save.
- Dual-SIM phones and shared family numbers mean one phone number does not always equal one person. OTP-based flows should survive a user switching apps to read the message.
Making it a habit
Build a small data set once: five realistic names (single name, initials, long, with apostrophe, in Tamil or Hindi script), five amounts that exercise lakh and crore grouping, a handful of IST edge times, and three awkward addresses. Reuse it across forms, APIs and reports. It costs an afternoon to create and finds defects in almost every product that was designed elsewhere first.
Figures and names in the examples are illustrative.
Leave a Reply