There's a meaningful difference between a website and a scalable online business. A website tells people you exist. A scalable online business can serve 10 customers and 10,000 customers without needing you to manually intervene each time someone buys, books, or registers.
Building the latter — a proper SaaS platform, a marketplace, or a subscription business — requires thinking about technology differently from the start. This post breaks down what that actually looks like, specifically in the Southeast Asian context where mobile penetration is extreme, payments are fragmented, and the local market has distinct needs that global platforms often ignore.
Most successful SaaS products start as a monolith — a single application where all the code runs together. Shopify, GitHub, and Basecamp all started this way. The "microservices first" approach is tempting but almost always wrong for early-stage products: it adds operational complexity before you've even figured out if the product has traction.
For a Southeast Asian SaaS product at early stage, a well-structured monolith with a clean REST or GraphQL API is the right starting architecture. You get one codebase to maintain, simpler deployment, and faster iteration speed. When you genuinely have scaling bottlenecks — real ones, not hypothetical ones — you extract services then.
Most SaaS products in Southeast Asia are well-served by PostgreSQL — a reliable, open-source relational database that scales further than most products will ever reach before they're large enough to afford dedicated infrastructure engineers. The common mistake is premature optimisation: choosing MongoDB or a distributed database because it sounds modern, then spending months fighting data consistency problems that PostgreSQL handles by default.
For multi-tenant SaaS (one product serving multiple companies), the key decision is how to separate tenants' data. The options — separate databases per tenant, shared database with tenant_id column, or separate schemas — each have trade-offs. For small-to-medium SaaS platforms targeting Thai or Indian markets, a shared database with row-level tenant isolation is usually the right balance of simplicity and security.
Hosting your application on a server in the US or Europe adds 150–300ms of latency for Southeast Asian users. That's not catastrophic, but it's noticeable — especially on slower mobile connections. For a product primarily serving Thai, Malaysian, or Indonesian users, hosting in Singapore (AWS ap-southeast-1, Google asia-southeast1, or providers like DigitalOcean Singapore) cuts latency to under 30ms.
Unlike in Western markets where Stripe covers most scenarios, Southeast Asian payments require a portfolio approach. In Thailand specifically:
A SaaS platform targeting Thai SMEs should support at minimum PromptPay and credit card via 2C2P or Omise. The Stripe-only approach that works in Western markets will lose you a significant percentage of Thai customers.
Building a product that truly serves Thai users means more than Google Translate. Thai script has specific rendering requirements — line-height needs to be larger than for Latin characters (1.8–2.0 vs 1.5), and certain fonts that look fine in English render Thai characters poorly. Date formats, number formats, and address formats in Thailand differ from Western conventions. If your product handles invoices or contracts, Thai tax invoice requirements (tax ID format, required fields) must be built in.
For a SaaS product with monthly or annual subscriptions, the billing logic is often underestimated. You need: pro-rata calculations when users upgrade mid-cycle, dunning (handling failed payment retries gracefully without immediately cutting off a customer), annual vs monthly pricing logic, discount codes and promotions, and tax calculations.
Rather than building this from scratch, most SaaS products in Southeast Asia use Paddle or Chargebee as a subscription billing layer on top of their payment gateway. For Thai-focused SaaS, Chargebee supports Thai Baht and integrates with Omise and 2C2P.
If you're building a platform where multiple businesses share the same infrastructure (a POS system used by many restaurants, a hotel PMS used by many properties), data isolation is non-negotiable. A bug that exposes Business A's data to Business B's users doesn't just create a support ticket — it ends your product. Row-level security in PostgreSQL, enforced at the database level rather than just the application layer, is the safest approach.
If your SaaS has a mobile component (and for Southeast Asian B2C products, it probably should), your API design needs to account for unreliable network conditions. This means: offline-first data caching where possible, optimistic UI updates that sync when connectivity returns, pagination over large data sets (don't return 10,000 records in one response), and image compression and lazy loading. Thai 4G is generally good in cities but rural connectivity is still patchy — designing for degraded network conditions pays off.
Early-stage SaaS products can run fine on managed services — Railway, Render, Fly.io, or a simple Coolify/VPS setup — maintained by a full-stack developer. You typically don't need a dedicated DevOps engineer until you're handling sensitive data requiring compliance (healthcare, financial), running at a scale where infrastructure cost optimisation is significant (usually $5,000+/month in cloud spend), or deploying multiple times per day across many services.
Below that threshold, managed infrastructure is almost always cheaper and faster than hiring or building your own.
We've built SaaS platforms for the Thai and Indian markets. Let's talk about your architecture and what the right tech stack looks like for your product.
Book a Technical Consultation →