Skip to content

SaaS5 min read

Multi-tenancy is a day-one decision

ARQON EngineeringARQON Singapore

Tenant isolation, permissions and audit logs are cheap to build first and expensive to add later. Here is where to start.

Plenty of SaaS products began as a tool built for one company. As a way to find product-market fit, that's fine. The trouble starts the day the second customer signs. The assumptions that made version one quick to build are exactly the ones that are dangerous to share across customers.

Decide how to separate tenants

There are three common approaches: a database per tenant, a schema per tenant, or shared tables with a tenant ID on every row. For most B2B products we recommend the last, enforced with row-level security in the database, rather than trusting application code never to forget a WHERE clause. When someone does forget a filter, the result should be empty, not another customer's data.

Three questions every buyer asks

Once you sell to companies rather than individuals, the same questions turn up on every security questionnaire. They are far cheaper to answer if the foundations were there from the start.

  • "Can we sign in with our own identity provider?" You answer with SSO (SAML / OIDC).
  • "Can we control who sees what ourselves?" You answer with per-tenant roles and permissions.
  • "Can we see who changed what?" You answer with an append-only audit log.

What can wait

Not everything has to exist on day one. Data residency, custom domains and complex pricing logic can usually wait until a customer will pay for them. The test is simple. If adding it later means touching every query or table, add it now. If it only adds something, wait.

On a reconciliation product for a fintech start-up, building tenant isolation and audit logs first added about three weeks to the schedule. In return, the product passed its first enterprise security review without a design change. Those three weeks paid for themselves many times over.

Filed under SaaS · 2026.02.11

Back to the journal

(→)Keep reading

More from the journal.

AI

· 7 min

Build the eval before the feature

The fastest way to lose three months on an AI project is to ship the demo first and work out whether it's right later.

Integration

· 6 min

Your webhooks will arrive twice

Any integration works on the happy path. Whether you can forget it exists depends on how it behaves on a bad day.

Practice

· 5 min

The handover is the product

If your team can't ship on its own three months after we leave, we've failed, whatever else we delivered.