Engineering

Multi-tenant SaaS: the architecture decisions clients should ask about

May 2026 · 5 min read

When a business evaluates a SaaS product — or a partner to build one — the interesting questions aren't about the feature list. They're about what happens beneath it: how one customer's data is kept away from another's, who can see what, and whether you can reconstruct who did what after the fact. We've made these decisions across four products in production, and the pattern is consistent.

Tenant isolation is a starting decision, not a later one

Every query in a multi-tenant system has to be scoped to the tenant making it. Retrofitting that after launch is painful and risky; designing for it from the first schema is straightforward. In our platforms, every data access is scoped to the customer's own tenant, so cross-tenant leakage isn't something we patch — it's something the architecture prevents.

Role-based access earns its keep in the real world

Real organisations have hierarchies. A firm owner, a senior associate and an associate should not all see and do the same things. Role-based access control (RBAC) with sensible default roles means the software matches how teams actually work, and it means the answer to "who could have changed this?" is bounded and knowable.

Audit trails are how trust survives an incident

Things go wrong in every system eventually. What separates a trustworthy platform is whether it can tell you what happened. Timestamped, attributed logs of status changes and document generation turn "we're not sure" into "here's exactly what occurred and when." For any workflow touching statutory or financial records, that's non-negotiable.

The questions worth asking a build partner

  • How is one tenant's data isolated from another's — at the query level or just the UI?
  • What roles ship by default, and how granular can permissions get?
  • Is there an audit trail, and what does it capture?
  • Is the system API-first, so it integrates rather than traps your data?

If a partner has made these decisions in their own products, they'll answer without hesitating. That's usually the signal you're looking for.

← Back to all posts

Keep reading

Next post

Compliance Tech

How AI document intelligence removes data re-entry from MCA compliance

CA and CS firms lose hours re-keying data from MCA PDFs. Here's how document AI reads incorporation papers, KYC and certificates — with a human check before anything is saved.