Do you really need to localize before your first overseas customer?

The first overseas sale does not settle the sequence

You do not need a fully localized product to acquire the first customer outside your home market. You need it before language friction costs more than removing it, and those are not the same moment.

Figma is the useful counterexample because its history breaks the supposed sequence cleanly. The company said roughly 80% of weekly active users were outside the United States “from day one,” while the product still operated primarily in English ¹. In its 2025 S-1, Figma reported that approximately 85% of monthly active users were outside the US and 53% of revenue was international ².

Localization followed that demand. Japanese arrived first, after research showed that an English-only experience was holding back adoption in a market where Figma already had customers. Spanish, Korean and Brazilian Portuguese followed later ^1,3,4. Language was not the gate to the first user. It became the gate between early adopters and the wider market.

The opposite pattern is real too. Canva described supporting 104 languages in 2021, with continuous translation built into product delivery ⁵. Spotify added 36 languages in one expansion, moving from 26 to 62 and changing more than 30 backend services to do it ⁶. Slack spent close to a year preparing around 20,000 strings across 2,000 files for its first major rollout ⁷.

These companies do not establish one correct order. They establish two boundaries. A differentiated professional product can acquire international users in English. A product serving many markets at high release velocity eventually has to treat localization as part of the delivery system.

Internationalization and localization are two decisions

The question becomes easier when the word localize is split in two. Internationalization is the engineering work that makes more than one locale possible. Localization is the adaptation of the product for a particular language and market.

An internationalized product keeps user-facing strings outside source code, formats dates, numbers and currencies by locale, supports plural rules and safe placeholders, allows text to expand, carries Unicode correctly, and does not assume every interface reads from left to right. Android’s own guidance begins with the same architectural move: take strings out of code and put them in resources ⁸.

That work does not commit a startup to buying five translations. It preserves the option to buy one without rewriting the product first. Software localization is wider than translation: the locale identifier, runtime grammar, layout and interchange file have already made decisions before a translator sees the text.

What an English-only product actually costs

There is no universal revenue penalty for staying in English. The cost depends on the audience, the product, the market, the acquisition channel and how much trust the transaction demands. What can be measured is where the friction appears.

CSA Research surveyed 8,709 consumers in 29 countries in 2020. It reported that 76% preferred buying products with information in their own language, 40% said they would not buy from sites in another language, and 75% said local-language customer care made them more likely to buy from the same brand again ⁹. These are stated preferences, not a randomized conversion test, and the distinction matters. They are evidence of friction, not a promise of uplift.

Product case studies show the same pattern in behavior, although they do not isolate localization as neatly as an experiment would. Google Play reported that SayHi’s addition of 13 languages coincided with a 120% increase in install rate, roughly 20% more revenue and around 50% more reviews in those languages ¹⁰. Karrot found that around 30% of its North American users had a non-English device language; buyers offered translation were 2.4 times as likely to begin a seller chat ¹¹.

Karrot is the more instructive example. It did not translate every surface before learning anything. It removed one language bottleneck from the transaction. That is often what the first localization investment should look like.

The hardest loss to see is the person who never became a user. An English-only product selects for people willing to use English, so a survey of existing customers systematically understates the barrier. Figma already had Japanese users when research found that English was still constraining wider adoption ¹.

Waiting also has an engineering price. Slack’s year-long retrofit is not an argument for translating on day one. It is an argument against letting day one hard-code an English-only architecture ⁷.

Consumer, B2B, mobile and regulated are different answers

B2B SaaS and developer tools can usually wait longer. Their early buyers are more likely to work in English, tolerate a rough edge, buy through a high-touch process and accept language as the price of a differentiated product. Localization becomes more important as the buyer widens from a technical champion to procurement, operations, support and company-wide adoption.

Consumer products should move earlier. Search, onboarding, trust, checkout, engagement, reviews and support are all self-service. The language problem compounds across the funnel, and paid acquisition makes every avoidable abandonment more expensive.

Mobile offers a cheap demand test. Apple and Google both support localized product-page or store-listing assets ^12,13. A translated listing can test whether a locale improves discovery and store conversion before the entire application follows. On the web, the equivalent test is a landing page, pricing page, signup and the first useful workflow.

Regulated and safety-sensitive products do not get to use a growth threshold. The EU Medical Device Regulation requires specified information in the official language or languages chosen by the member state where a device is made available, and the Consumer Rights Directive permits national language requirements for contractual information ^17,18. The details are jurisdiction- and product-specific; our closer look at where localization law actually reaches explains why the required artifact may be a label, instruction, contract or employee-facing interface rather than every product string.

Localize the bottleneck before the whole product

Full product coverage is a poor first experiment because it makes the largest bet before the company has measured the market. Translate the surface most likely to be suppressing the next event in the funnel.

Measure the translated cohort against the baseline: store conversion, signup, activation, checkout, retention and support rate. A target locale with organic users but no improvement after its highest-friction surface is translated has not yet earned a full-product rollout. One with a clear lift has.

This also prevents the local-language front door and English-only room behind it from becoming an accidental bait and switch. The experiment has to include enough of the journey for a person to reach the value the acquisition copy promised.

Continuous localization earns its place through repetition

Continuous localization is usually the best way to keep translations synchronized once changes happen often enough. It connects source control and the translation system so changed strings travel out and completed translations return without a person exporting, emailing and merging files each release ^14,15.

source strings change- translation, review and automated QA- completed locales returned- UI tests and pseudolocalization- release

It earns that machinery when several languages are active, releases happen weekly or daily, strings change frequently, multiple teams contribute copy, and localized releases need to stay close to the source release. For one market, rare releases and a few hundred strings, the manual loop may still be cheaper.

Automation improves synchronization, reuse, context, QA and release latency. It does not make a poor sentence good. Slack described a 97% language-quality pass target while separating linguistic issues from functional bugs such as missing strings, broken variables and clipped text ¹⁶. Canva combined its automation with professional translation, contextual information and coverage checks ⁵.

Translation is the visible cost, not the whole cost

Translation is priced visibly, which makes it easy to mistake for the total. Smartling gives a rough range of $0.15–$0.30 per source word for professional translate-and-edit workflows; lower-cost per-word services are also available and vary by language pair and tier ^19,20.

Machine translation can make raw character processing very cheap ²¹. For a small UI corpus, the expensive parts are more often context, review, integration and checking the result in the interface.

A translation-management system adds another recurring line item. Public plans range from entry-level offerings to much larger team and enterprise tiers; Crowdin, Transifex, Lokalise and Phrase all publish their current plan structures ^22,23,24,25. Those pages change, so they belong in a purchasing comparison, not in a permanent claim that one platform costs a fixed amount.

The more important correction is temporal. A locale is not bought once. New strings, changed flows, support copy and regression testing make it a subscription to your own release cadence. That is why localization billing follows churn, not the size of the first export.

A market should earn the next stage of localization

A credible plan has no magic threshold such as 1,000 international users. It lets evidence move a market from one stage to the next.

The sequence keeps the product ready without pretending every country is already a market. It also makes retreat possible. A translated landing page can be retired cheaply; five fully supported locales, contracts and a support promise cannot.

The threshold is economic, not a user count

The decision should be made against contribution, not top-line market size. Count additional gross profit from acquisition and retention, support savings and avoided risk; subtract the complete annual cost of maintaining the locale.

For a quick screening calculation, divide annual localization cost by the target market’s annual contribution before localization.

That is why localization gets easier to justify after a market proves itself: the required improvement becomes smaller. It is also why generic industry ROI claims are not enough. As our review of which locales actually pay back found, a survey preference or download increase is not the same as a causal estimate of profit for your product.

Internationalize before demand; localize before friction wins

Serious global products localize meaningful parts of the experience. The evidence does not say they all do it before the first customer, or that translating every surface at once is the sensible first move.

A technical B2B company can internationalize, observe where overseas users gather, translate the highest-friction path, and expand after measuring the result. A consumer company should usually bring localized acquisition and the critical transaction path forward. A regulated product begins with the legal language requirement rather than the growth model.

So the answer remains no, with one condition. Do not localize everywhere merely because the world is international. Localize where language is becoming a measurable constraint on growth, trust, compliance or operating efficiency—and build the software early enough that crossing that threshold is an operational decision, not a rewrite.

References

1. 1.Figma, 2022 Expanding Figma's international presence Figma Blog, 22 June 2022
2. 2.Figma, 2025 Form S-1 registration statement US Securities and Exchange Commission, 1 July 2025
3. 3.Figma, 2025 Figma in Spanish: How localization helped us scale Figma Blog
4. 4.Figma, 2025 Designing Figma's Korean localization Figma Blog
5. 5.Canva Engineering, 2021 How to design in every language at once Canva Engineering Blog, 7 July 2021
6. 6.Spotify Engineering, 2022 Scaling translations at Spotify Spotify Engineering, 30 September 2022
7. 7.Slack Engineering, 2017/2020 Localizing Slack Slack Engineering
8. 8.Android Developers Localize your app Android developer guides
9. 9.CSA Research, 2020 Can't Read, Won't Buy — B2C Survey of 8,709 consumers in 29 countries
10. 10.Google Play SayHi increases install rate by 120% with app localization Google Play Console case study
11. 11.Android Developers, 2026 How Karrot increased sales with AI translation Android Developers Blog
12. 12.Apple Developer Product page optimization App Store developer guidance
13. 13.Google Play Console Help Translate and localize your app Google Play Console Help
14. 14.Crowdin Continuous localization: What it is and how it works Crowdin Blog
15. 15.Crowdin Add localization to GitHub workflows with Crowdin Action Crowdin Blog
16. 16.Slack, 2021 How we launch languages at Slack Slack Blog, 11 February 2021
17. 17.European Parliament and Council, 2017 Regulation (EU) 2017/745 on medical devices Official Journal of the European Union, EUR-Lex
18. 18.European Parliament and Council, 2011 Directive 2011/83/EU on consumer rights Consolidated text, EUR-Lex
19. 19.Smartling Translation cost and turnaround times Smartling Help Center
20. 20.Gengo Translation services and pricing Gengo
21. 21.Google Cloud Cloud Translation pricing Google Cloud pricing
22. 22.Crowdin Pricing Crowdin
23. 23.Transifex Pricing Transifex
24. 24.Lokalise Pricing Lokalise
25. 25.Phrase Pricing Phrase