IT

How to scale a SaaS platform without breaking it

What breaks when a startup grows?

Usually six things, and hardly ever all at once. Each one sends recognisable signals before it actually breaks: it pays to read them early.

What breaks as a SaaS platform grows: area, symptom and what to do
AreaSymptomWhat to do
InfrastructureSlowdowns at peak times, errors that only show up under load, releases everyone dreads.Monitor errors and response times on the key flows, keep staging and production apart, release in small reversible steps.
DataQueries slowing as tables grow, reports nobody trusts, backups never tested.Measure queries and add the missing indexes, keep business metrics in one place, rehearse restoring from backup.
TeamOnly one person understands part of the system; newcomers struggle to contribute.Readable code, essential documentation, code review, technical decisions written down with their reasons.
ProcessesEvery release is an event, bugs keep coming back, nobody knows what went live and when.Continuous integration with tests on the critical flows, automated releases, a changelog.
SecurityPermissions that grew by accident, secrets pasted into chat, nobody sure who can see which data.Review roles and permissions, keep secrets in a dedicated manager, log access. Treat GDPR as a product requirement: fines reach €20 million or 4% of worldwide turnover (Regulation (EU) 2016/679, art. 83).
Cloud costsThe bill grows faster than the customer base; services spun up for a test and forgotten.Split costs by service and, where possible, by customer; size resources to real usage; review the bill every month.

Which metrics should you watch as you grow?

The ones that tell you whether customers stay and whether growth pays: recurring revenue (MRR), churn, the cost of acquiring a customer and their lifetime value (CAC and LTV).

Two reference points. According to SaaS Capital (2025), median net revenue retention among private B2B SaaS companies is 102%, with the top quartile at 111% (annual contract values of $25–50k). For David Skok, monthly net revenue churn above 2% is a warning sign: a rule of thumb, not a law.

How we do it on Athleex

Athleex is our own SaaS for personal trainers: we built it and we keep growing it. Its business dashboard shows recurring revenue, churn and customer value; the Churn Radar flags athletes at risk of leaving before they go. Underneath: a Next.js web app on PostgreSQL, hosted on European servers we run ourselves, iOS and Android apps built with Capacitor on the same base, automated CI on GitHub Actions, and nightly database backups with an off-site copy.

Where do you start when scaling?

With the right order. It's the one we follow on our own products.

  1. Measure

    Errors, response times, costs and retention in one place. Without numbers, every architecture choice is an opinion.

  2. Tested backups and restores

    A backup you've never restored is a hope. On Athleex the database is backed up nightly, with an off-site copy.

  3. CI and tests on critical flows

    Login, payments, the flows revenue depends on: tested automatically on every change.

  4. Data and queries

    The fastest-growing tables, the slowest queries, the missing indexes: that's usually where the biggest gains are.

  5. Cloud costs

    Every service with an owner and a cost per customer; resources sized to what's actually used.

  6. Team and processes

    Documentation, code review, releases that need no heroes. It's last only on this list: start early.

A transparent glass tower on a frosted-crystal foundation, with a lilac light running through it

Rewrite from scratch or refactor?

Refactor, nearly always: one piece at a time, starting where the numbers show the pain. A full rewrite freezes the product just as customers are growing.

Refactor when
the system works but slows down, one part is fragile, the code is hard to touch: work on that part, with tests around it.
Consider a rewrite only if
the foundations can't support what you need (security, data, architecture) and every change costs more than rebuilding that part.

If the platform started life as an AI or no-code prototype, we begin with our prototype-to-production service; if it needs rethinking, with SaaS development. A fractional CTO makes the technical calls with you, and the Athleex case study covers our own choices and mistakes. The stop before this one is go-to-market after seed.

Questions about scaling

When does a SaaS platform need to scale?

When the signals come from customers before they show up in the charts: slowdowns, errors under load, releases everyone dreads, costs rising faster than revenue. The best time to prepare is just before that, while growth is predictable and the team still has time to measure.

What is a good churn rate for SaaS?

It depends on market and price, but a rule of thumb from David Skok treats monthly net revenue churn above 2%, roughly 22% a year, as a warning sign. The trend matters more than the number, and so does knowing who leaves and why.

What is net revenue retention, and what's a normal figure?

It's the share of existing customers' revenue still there a year later, counting expansion and churn. According to SaaS Capital (2025), the median among private B2B SaaS companies is 102%, with the top quartile at 111%.

When do you need a CTO, even a fractional one?

When technical decisions start to weigh on the numbers: architecture, cloud costs, hiring, security. If it's not yet time to hire one, a fractional CTO can cover the role in the meantime.

How do you cut cloud costs safely?

Measure first: cost per service and, if you can, per customer. Then switch off what you don't use, size resources to real usage and review the bill every month. Changing architecture or provider comes later, with the numbers in hand.

Further reading

Is your platform growing faster than it can cope?

We start from the numbers you have, and from what's already starting to break.

Let's talk