The next-market decision, on one page.

Portfolio · en-US · profile · SAAS

Recommendation: launch Chinese next, anchored on China.

Localizing into Chinese adds roughly 1.3B speakers the current portfolio does not reach (16.5% of world population) at medium engineering effort. Composite score 95/100 under the SAAS weighting, 23 points clear of the runner-up (Hindi).

Chinese ranks #1 under all 4 weighting profiles.

What shipping Chinese takes · day one

• Writing systemCJK (new to the portfolio)
• Layout directionLeft-to-right, no mirroring
• Plural arms per counted string1
• Custom webfontRequired, budget subsetting

Risks & watchouts

• New writing system (CJK): font stack, vertical metrics, and truncation rules have no coverage today.
• Needs a bundled webfont: adds load-time weight in exactly the markets where connectivity is weakest. Subset aggressively.

Proposed sequence

Step counts assume the earlier steps have shipped, so the three numbers can be summed. Overlapping languages are not double-counted.

Alternatives considered

• Hindi (India): 71 vs 95. LTR layout · Needs custom webfont
• Portuguese (Brazil): 64 vs 95. Script already supported · LTR layout

What this model does not include

Revenue or ARPU by market, English-proficiency substitution (high-proficiency markets tolerate an English product longer), competitor coverage, regulatory requirements, and content depth beyond UI strings. Scores order candidates. They are not forecasts. Bring your own revenue data to fill that gap, or load per-market usage under Atlas → Your Data to add a demand signal.

Depth: UI at publication quality · Ships UI strings edited to publication quality with one LQA round. Docs and support stay in the source language.

The budget ask quotes the Chinese row plus the one-time setup ($10,350 together). The totals row prices all 3 markets of the proposed sequence, translation only, worked in parallel.

Ongoing upkeep (assumption): ~$1,060/year per market. Assumes 40% of the source words are retranslated each year at the same depth rates, LQA rounds, and PM overhead. Replace the churn rate with your own.

• Per-word rates and throughput are industry-typical 2025 defaults, not a vendor quote.
• The estimate prices each locale with a per-language-pair rate index (Japanese, Nordic, and RTL pairs run above a mainstream pair), so per-locale figures differ. The index is a planning figure without a sourced rate behind it.
• The model assumes no translation-memory leverage: a first launch has no memory, so it prices every word as new. A mature product with a large TM pays for fewer new words.
• Engineering setup (RTL enablement, webfonts, i18n framework) is a separate one-time cost (see the budget estimator). It is not in this per-word figure.
• This depth prices ui at "MT + full post-edit" with 1 LQA round and 15% PM overhead.

Appendix · sources

• CLDR territory & language data · Unicode Consortium · cldr-json main branch (latest published CLDR).
• Internet users (% of population) · World Bank Open Data · Most recent reported year per country (varies by country, year stored per record).
• Translation per-word rates & throughput · Curated (industry-typical defaults) · 2025 defaults, every figure editable in the UI.
• Can't Read, Won't Buy (consumer survey) · CSA Research · 2020 survey of 8,709 consumers across 29 countries.

Full methodology: locale-lab · /atlas/methodology