Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Software Engineering 6 min read

Multi-tenant SaaS architecture: a buying guide for founders

A practical buying guide to tenancy models, isolation evidence, metering, pricing questions and exit terms before you commission a SaaS platform.

E
EncryptTechBD · 0 comments
Developer working on a laptop beside a smartphone
Photo: Christina Morillo / StockSnap

Multi-tenant SaaS is two purchases in one. The first is a business model: you sell a service to many customers from one product. The second is an architecture: those customers share some combination of application, database and infrastructure. A buying guide that treats the second as a technical detail will produce a platform that is cheap to launch and expensive to operate.

This guide is written for founders and operations leaders who are commissioning a SaaS product, replacing a vendor, or moving an internal system to a subscription model. It focuses on the decisions that are hard to reverse after the first hundred tenants.

What you are actually buying

You are buying four things: a tenant model, an isolation guarantee, an operating model and an exit path. Features can be added later. The tenant model usually cannot. Ask every vendor to describe those four in plain language before you look at a demo.

Requirements to settle before vendor talks

  • Who is a tenant? A company, a branch, a reseller, a department or an individual user. The answer changes reporting, billing and support.
  • Data classification. Which records are personal, financial, health-related or regulated, and which may be stored in a shared database?
  • Isolation expectation. Does the customer require a separate database, separate infrastructure or a named administrator with no shared access?
  • Identity model. Single sign-on, directory sync, role design, contractor access and emergency access.
  • Data residency. Where the data must live, which backups count, and what happens during a failover to another region.
  • Integrations. Which systems must exchange data, in which direction, and how often.
  • Reporting. Which reports are tenant-facing, which are internal, and whether the two can safely share a warehouse.
  • Exit. What a customer receives on termination, in what format, and how long deletion takes.

Tenancy models and their trade-offs

Silo

Each tenant gets a separate database or infrastructure stack. Isolation is easy to explain and blast radius is small. The cost is operational: every migration, backup check and monitoring rule runs many times. Silo fits regulated customers, large accounts and products where a per-tenant restore is a contractual requirement.

Pool

Tenants share a schema and rows carry a tenant identifier. This is efficient for many small accounts and makes product-wide reporting simpler. The risk is a missing filter. Row-level security, enforced at the database layer, and automated tests that attempt cross-tenant reads are not optional. Pool fits self-service products with a large, price-sensitive customer base.

Bridge

A shared application with a separate schema or database per tenant. It preserves a stronger isolation story while keeping one deployment. The complexity moves into migrations and connection routing. Bridge is often the right compromise for a business that starts with small tenants and expects a few large ones.

Whatever the model, ask about noisy neighbours, per-tenant rate limits, per-tenant backup and restore, and the blast radius of a bad migration. A vendor that cannot demonstrate a single-tenant restore is asking you to trust a promise.

A vendor scorecard that exposes real capability

  • Isolation evidence: automated cross-tenant tests, row-level security or separate databases, and a written incident history.
  • Restore evidence: a live demonstration of restoring one tenant without touching others.
  • Identity: SSO, directory sync, role granularity, service accounts and revocation time.
  • Metering: how usage is measured, how limits are enforced and how disputes are resolved.
  • Auditability: tenant-visible audit logs, export format and retention.
  • Deployment: maintenance windows, rollback method and change notification.
  • Data portability: a complete export that a third party can read without the vendor's tool.
  • Security practice: evidence mapped to a recognised standard such as OWASP ASVS, not a logo on a slide.

Score each item from one to five and weight isolation, restore and export most heavily. A vendor that scores well on features but poorly on those three is selling a demo, not a platform.

Pricing questions that reveal architecture

Do not ask only for a monthly price. Ask what unit is metered, what happens when a tenant crosses a limit, whether isolation is priced separately, how much notice is required for a plan change, and whether a restore or migration is billable. Ask how API, storage and AI usage are passed through, and whether those charges appear on the same invoice. A vendor that cannot explain the meter usually cannot explain the architecture behind it.

The pilot

Run one real tenant through one complete workflow for thirty to forty-five days. During the pilot, add a user, remove a user, change a role, export the tenant's data, restore a deleted record, test peak load and raise a support ticket at an inconvenient hour. Reconcile the invoice against the contract. The pilot should test the operating relationship, not just the software.

Contract and exit terms

Confirm who owns the configuration and the data, which subcontractors process it, how breach notification works, what happens on termination for convenience, and what transition assistance is included. Ask for a deletion certificate and a defined export window. If the vendor hosts the product in a country your regulator does not accept, no amount of contract language fixes the architecture.

Common buying mistakes

  • Choosing a tenancy model from a blog post rather than from customer requirements.
  • Accepting a shared-database demo without a cross-tenant access test.
  • Ignoring per-tenant restore until the first customer deletes a year of records.
  • Pricing the platform but not the migration, training and internal administration.
  • Letting the vendor own the data model and the export format.
  • Signing a multi-year term before a single tenant has completed a full billing cycle.

Where this fits in our work

If you need the architecture reviewed or the product built, start with Custom Software and Integrations. The pricing page explains how discovery, build and support are scoped. For the interface layer that tenants will actually integrate with, read the REST API design checklist.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes