Define audiences, locales and contexts
6 min read
Lesson objectives
- Build an audience, locale and use-context matrix that separates language needs from product and support rules, and identify unsupported generalizations.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
A sponsor asks, 'We already have an Arabic translation. Why do you need three versions?' Do not begin by listing countries or dialects. Ask where users encounter the content, what task they are trying to complete, and what genuinely differs across markets. A locale describes language and regional conventions, including possible number and date formats, but it is neither a user persona nor a summary of a country's laws. Users in one market can differ in expertise, access and preference; users in several markets may understand the same Modern Standard Arabic text. Segmentation therefore rests on the effect of a difference on the task, not on a cosmetic wish to produce more files.
Build a three-axis matrix before choosing wording. Axis one is a defined audience, such as a new customer or an experienced support agent. Axis two is market or interface locale, with text language, direction and data formatting. Axis three is the use moment: an introduction page, payment step, error message or account-recovery instruction. In each cell record the task and any sensitive decision, then evidence that a particular difference affects it. Marketing language may need a tone and comprehension test. An error message needs short practical guidance that matches what the product can do. An expiry date needs an unambiguous format and a clear time zone. A different date format does not imply a different returns policy; that is an operational or legal question for its owner.
Distinguish three decision levels. Translation chooses wording that carries meaning in context. Localization checks whether the interface, data and journey fit a defined audience. A service-policy decision, such as eligibility for a refund or the promised reply time, is not made by a translator and cannot be inferred by an adviser from a country name. The adviser records the possible difference and asks the policy owner for an explicit rule. That rule then becomes a testable product and content requirement. If 'we will reply soon' is used in one market where the team commits to a day and another where it makes no such promise, a more elegant phrase cannot resolve the operational ambiguity.
The interviews themselves have limits. Five interviews cannot estimate how common a misunderstanding is among all users, but they can expose a mechanism worth testing. Design a short task: show text and badge as they appear in the real interface, then ask participants what they can do now and whether a decision has been made. Record their interpretation before leading them with 'Did you understand that you were not approved?' Compare current and proposed versions in balanced order. Do not change the text, colour and badge placement simultaneously if you need to attribute an effect to one element. If the deadline forces several changes at once, label the finding a test of the combined change, not proof about wording alone.
Numbers and direction also matter. Unicode CLDR presents locale conventions for dates, numbers and time zones, but those data do not replace testing a product interface in its actual market and version. For mixed Arabic text and Latin identifiers, W3C guidance describes language and direction as separate metadata for localizable strings. Selecting an Arabic font alone cannot fix a booking code whose order is reversed or whose characters are clipped. Ask the product team for an actual interface example, identify the effect on task completion, and route the rendering defect to the technical owner rather than changing the translation at random.
Report to the sponsor an outcome, not a file count: 'We propose one shared meaning, with a targeted B-market test, a C-market mobile fix, and a product-owned policy decision still open.' That is more precise than 'we need three Arabics' and safer than assuming one formal register solves every journey. State why each variant would exist, its maintenance cost, and the condition that removes the need for it. If appropriate testing shows no material comprehension or functional difference, avoid a variant that will be harder to update. If an important difference appears at a critical step, do not sacrifice clarity for cosmetic uniformity. Locale segmentation should reduce misunderstanding and risk; it is not a production goal in itself.
Lesson summary
Segment locale, audience and task on evidence; do not infer policy from language.