How to build a SaaS from scratch: from MVP to first customer
What to build first, what to leave for later, and the technical decisions that shape the product for years. A realistic sequence to the first paying customer.
May 12, 2026 · Santiago Raúl Díaz · 2 min read · Updated on
Building a SaaS from scratch means validating that someone has the problem, building the smallest version that solves it end to end, charging for it, and only then expanding. The hard part isn't coding: it's resisting the urge to add features before you have a paying customer.
Before writing code
The question that prevents most failures: is anyone solving this problem today in an uncomfortable way? If nobody is working around it with spreadsheets, manual processes, or ill-fitting tools, the problem probably doesn't hurt enough for anyone to pay for a solution.
Validation isn't asking whether people like the idea. It's asking how they do it today, how long it takes, and what would happen if they stopped. The answers to those three questions define the product's scope better than any brainstorm.
What goes into the MVP
The criterion is that a user can go through the complete workflow and get the result, however simply. A product that does one whole thing is worth more than one that half-does five.
- The complete main flow, from entry to result.
- Sign-up and login.
- The minimum admin needed to support users without touching the database.
- Some way to charge, even if it's manual at first.
And what can wait: visual customisation, advanced reports, integrations nobody has asked for yet, a mobile app, and any metrics dashboard you won't look at in the first three months.
Architecture decisions that are expensive to change later
| Criterion | Decision | Why it matters |
|---|---|---|
| Multi-tenant or separate instances | Structural | Changing it later means redoing the data model |
| Permissions and roles model | Structural | Adding roles onto a flat design touches the whole system |
| Per-customer data strategy | Structural | Shapes isolation, backups, and compliance |
| Activity logging from day one | Important | Without traceability you can't support or audit |
| Framework choice | Reversible | It can be migrated, even if it costs |
The first three are worth settling before the first line of code. The rest can be corrected along the way.
The pricing model
Define early what grows as the customer uses the product more: users, operation volume, storage. That's your pricing axis, and it should align with the value the customer perceives, not just with your cost.
Charging from the first customer, however little, is the only real validation. An enthusiastic free user doesn't prove the product has a market.
After launch
Launch isn't the end: it's when you start actually learning. Priorities shift from building to watching how it's used, where users get stuck, and what the ones who leave were asking for.
- Instrument the product so you know what's used and what isn't.
- Talk to the users who churn: they say more than the satisfied ones.
- Resist features requested by a single large customer.
- Measure how long a new user takes to get the first useful result.
Validate that the problem exists and hurts, build the simplest complete workflow, settle the three structural architecture questions early, charge from the first customer, and use launch to learn rather than celebrate.
Frequently asked questions
How long does a SaaS MVP take to build?
It depends on scope, but the useful rule runs the other way: set a short deadline and trim the scope to fit it. An MVP that takes a year isn't an MVP.
Should it be multi-tenant from the start?
Generally yes for products aimed at many small customers. Separate instances make sense when each customer requires strong isolation or deep customisation.
When do I start charging?
With the first customer. Charging a little validates the problem; giving it away validates only that it's free.
What do I leave out of the MVP?
Visual customisation, advanced reports, unrequested integrations, a mobile app, and any dashboard you won't look at within three months.