Skip to content
Reference 2026-07-01 8 min

Scaling SaaS

The bottlenecks that appear at 1k, 10k and 100k users — and how to plan for them.

By Mohammad Zayed

Overview

Most SaaS fails to scale because of unindexed queries and synchronous heavy work — not the framework. Plan the data layer first.

At 1k users

One app server and a single Postgres instance are plenty. Add indexes for your hot queries.

At 10k users

Move email/exports to workers, add a CDN, and introduce read replicas for reporting.

At 100k users

Introduce caching, shard or partition hot tables, and separate tenants by load. Monitor queue depth and DB connections.

Frequently asked questions

Is this a definitive recommendation?
It's our reasoned default, not a universal rule. The right choice depends on your constraints, team and timeline.
Do you only use these technologies?
No. We choose the stack that fits the problem. These pages explain the defaults we reach for and why.
Can you help us decide?
Yes. Book a strategy call and we'll map your requirements to a stack.

Continue reading

Architecture notesEngineering insightsOur services

Want help applying this?

Book a free strategy call. We'll map your bottlenecks to the right systems and send a clear roadmap — even if we don't work together.

Chat on Telegram

Usually replies within minutes. Chat on Telegram: @northflowstudio

No obligation consultationFounder-led projectsInternational clientsFast responseSecure communication