The Multilingual Reality Most SME Websites Ignore
Walk into any coffee shop in Toa Payoh or Jurong and you’ll hear English, Mandarin, Malay and Tamil within the same five minutes. Yet the overwhelming majority of Singapore SME websites are built, and stay, in English only. That’s not necessarily wrong — English is the working language of business here — but it does mean a real slice of the customer base, particularly older Mandarin-speaking shoppers and Malay- or Tamil-first households, is quietly reading a site that wasn’t written for them, or not reading it at all.
Until recently, fixing this meant hiring a translation agency, paying per word, and maintaining separate content forever. That cost structure made sense for banks and statutory boards. It never made sense for a 12-person retailer or clinic. AI translation has changed that maths meaningfully, though not as completely as some vendors claim.
Where AI Translation Actually Works Well (and Where It Doesn’t)
We’ve now run this for several clients, and the pattern is consistent:
- Works well: product descriptions, FAQs, delivery and returns policies, appointment booking flows, WhatsApp auto-replies, and structured content with clear, literal meaning.
- Needs a human pass: anything with brand voice, humour, idioms, or persuasive copy — a punchy tagline translated literally by DeepL or Google Cloud Translation often reads flat or slightly odd in Mandarin, even when grammatically correct.
- Don’t automate at all: legal terms, medical or dietary claims, and anything PDPA- or MOH-related where a mistranslated word changes an obligation.
The tools worth knowing: Google Cloud Translation API and DeepL API both plug into WordPress via plugins like TranslatePress or WPML’s AI add-on, and both now handle Simplified Chinese, Malay and Tamil at a genuinely usable quality for informational content. DeepL tends to read more naturally for Mandarin; Google’s coverage of Tamil is currently the stronger of the two.
Building a Practical Localisation Workflow
Rather than translating an entire site on day one, we usually recommend SMEs sequence it like this:
- Start with the highest-traffic, highest-intent pages: your service or product pages, contact page, and booking flow.
- Machine-translate first, using a tool that lets a human editor review and lock in changes (not a one-click auto-translate widget that re-translates live on every page load — those are worse for SEO and worse for consistency).
- Get one bilingual staff member, even someone outside marketing, to spend 30 minutes reviewing the machine draft for tone and obvious errors before it goes live.
- Keep a shared glossary of terms you never want auto-translated — your business name, product names, certification terms — so the AI doesn’t “helpfully” translate them.
The Quality Control Step Most Businesses Skip
The single biggest failure mode we see isn’t bad translation quality — modern engines are good enough for most SME content. It’s publishing a machine translation with zero review and later discovering, from a customer complaint, that a promotion’s terms read differently in Mandarin than in English. Build in a review step, however light, before anything customer-facing goes live in a second language. For anything touching pricing, refunds, or health claims, that review should come from a person, not a second AI check.
Getting Started This Month
You don’t need to localise your whole site to get a return. Pick your single highest-value page — often the booking or enquiry form — translate it into Mandarin first (still the largest second-language segment for most SMEs), measure whether conversion or enquiry volume shifts over a month, and only then decide whether Malay and Tamil versions are worth the same investment for your specific customer base. Localisation done well is a targeted move, not an all-or-nothing rebuild.