There’s no single correct process for building a SaaS, but there is a defensible order — one that front-loads the cheap decisions that invalidate expensive ones. Most wasted years come from doing these stages in the wrong sequence, not from doing any of them badly.
Here’s the workflow, and the decision each stage actually exists to make.
Stage 1 — Idea
The decision: is this a problem someone has, or a solution you find interesting?
Write the problem in one sentence, from the perspective of the person who has it, without mentioning your solution. If you can’t, you don’t have an idea yet — you have a technology you like.
Two useful filters at this stage:
- Do you have unfair access? A domain you know, an audience you have, a problem you’ve lived. This matters more than the idea’s quality.
- Is anyone paying to solve it badly today? Existing spend — even on spreadsheets and manual labour — is the cheapest evidence of a market.
Time-box it. Ideas are worth days, not weeks.
Stage 2 — Validation
The decision: does anyone want this enough to act?
This is the stage founders skip, and skipping it is the single most expensive mistake available (more on that). What it looks like in practice:
- Read how people describe the problem in their own words — communities, reviews, support forums.
- Talk to ten of them. Ask what they do now, never what they’d hypothetically use.
- Ask for a commitment: a call, a waitlist signup, a design-partner agreement, a pre-payment. Interest is free; commitment is signal.
How to validate a product idea on Reddit covers doing all of this for nothing. Exit criteria: several people described a costly workaround, and at least a few committed to something.
If nobody does, you’ve just saved a year. That’s a successful stage, not a failed one.
Stage 3 — Scope
The decision: what is the smallest thing that solves the problem end to end?
Not the smallest product — the smallest complete one. A tool that does one job fully beats one that does five jobs partially, because a partial job doesn’t get paid for.
- Write down the one workflow a user completes to get value.
- Everything not on that path is v2. Write it down elsewhere so you stop thinking about it.
- Decide what “activated” means now, because you’ll instrument it later and you’ll want the definition to predate the data.
Stage 4 — Tech stack
The decision: what lets this small team ship and change quickly?
The honest guidance is that the stack matters far less than founders enjoy debating. Optimise for:
- What you already know. Learning a framework and finding product-market fit at the same time is two hard problems.
- Boring, well-documented defaults. A monolith, a relational database, a managed host. You can split it later, with better information.
- A billing provider chosen deliberately, because it’s genuinely annoying to change later — merchant of record, fees, and tax handling all follow from it (how to choose).
- Thin boundaries around third-party services, so swapping one doesn’t mean a rewrite.
Resist scale-shaped architecture. Build for ten times today’s usage, not a thousand.
Stage 5 — Design
The decision: what’s the shortest path from landing to value?
At this stage design means flow, not aesthetics:
- Signup with the least friction your model allows. Decide auth methods deliberately — they change conversion more than most features (which sign-up methods to use).
- No empty first screen. Seed data, pre-fill, or make the first action obvious. An empty dashboard is where activated users go to become churned ones.
- One primary action per screen.
- Use a component library. A custom design system before you have customers is procrastination with good taste.
Stage 6 — Marketing stack and distribution groundwork
The decision: what has a long lead time, and therefore needs starting now?
This runs in parallel with building, not after. The channels that compound take months, so starting them at launch guarantees launching into silence:
- Instrument analytics and attribution before traffic arrives, so you can tell later which channel produced which customer.
- Set up transactional and lifecycle email on an authenticated sending domain.
- Start collecting emails with a landing page, from day one.
- Write the first two or three problem-focused pages so search has time to work.
- Show up in two or three communities so launch day isn’t the first time anyone’s heard of you.
- Stand up a referral or affiliate program. It needs partners recruited and tracking installed before it produces anything, which is why deferring it costs a year of compounding. Pay-on-results means it costs nothing while you wait (when to start).
The marketing stack every SaaS actually needs covers what to buy and what to skip.
Stage 7 — Private beta
The decision: does this work without you in the room?
Give it to ten real users and watch. Not a demo — watch them use it unaided. Every place they hesitate is a place your onboarding is wrong.
Exit criteria: new users reach first value without a hand-hold, and a few of them are paying.
Stage 8 — Launch
The decision: how do you convert accumulated goodwill into attention?
Launch is a harvest, not a start. Sequence it across a week — your own audience first, then directories, then communities with format-appropriate posts, then a retrospective. SaaS launch strategy has the full sequence.
Stage 9 — Iterate on one channel
The decision: which single channel gets your attention next?
The launch told you where your buyers actually came from. Commit to that one, plus partners — the exception to the one-channel rule, because it runs on someone else’s effort. The organic channel comparison covers the trade-offs.
The one rule that holds the process together
Each stage exists to make the next one cheaper to be wrong about. Validation is cheap and can invalidate scope. Scope is cheap and can invalidate the stack. That’s why the order matters more than the contents — and why skipping the early stages doesn’t save time, it just moves the discovery of your mistake to the most expensive possible moment.
FAQ
What is the right order to build a SaaS?
Idea, validation, minimum scope, stack, design, then distribution groundwork in parallel with building, followed by private beta, launch, and focused iteration on one channel. The order matters because each stage can invalidate the next.
How long should each stage take?
Idea: days. Validation: two to four weeks. First shippable version: weeks to a few months, not a year. If any stage runs much longer, it’s usually a sign the previous one wasn’t finished.
When should marketing start during a SaaS build?
During development, not after. Search content, community presence, and referral partners all take months to compound, so starting them at launch means launching with no audience.
More product and growth strategy is in the growth & pricing hub.