What Makes This Different From A Build
A SaaS product is software strangers pay to use without ever speaking to you. Everything that follows from that sentence is the actual work, and most of it is not the feature you had the idea about.
An internal system has known users, a support route down the corridor and a tolerance for rough edges. A product has none of those. It has to explain itself, take money reliably, survive somebody doing something unexpected at two in the morning, and keep one customer’s data completely away from another’s.
That is the difference between building the thing and building the business around the thing.
The Parts That Are Not The Product
Budget for these explicitly, because they are where SaaS projects overrun.
Tenancy. Keeping customers separated. The decision between shared tables with strict scoping and separate databases per customer shapes everything afterwards and is genuinely hard to reverse.
Billing. Subscriptions, trials, upgrades, downgrades, failed payments, refunds, proration, dunning. Each has a failure path and each failure path is a customer conversation if you get it wrong.
Onboarding. A product nobody can start using alone is a service with a login page. Self-service signup, sensible defaults and a first-run experience are product features, not polish.
Account management. Invitations, roles, removing somebody who has left, transferring ownership when the admin leaves the company.
Support surface. Somewhere to see what a customer sees, without touching their data unnecessarily.
The Decision That Is Hardest To Change
Tenancy, and it is worth deciding deliberately rather than by default.
Shared, with scoping is cheaper to run and to operate, and every query must be correctly scoped forever. One missed check is one customer seeing another’s data, which for a SaaS business is close to fatal.
Separate per customer is stronger isolation and simpler to reason about, and multiplies operational work: migrations, backups and deployments across many databases.
Most products should start shared with scoping enforced structurally rather than remembered per query. Enterprise requirements sometimes force the other answer, and that is a business decision made once.
A Platform We’ve Run
For a legal-marketing business we built the subscription platform where UK law firms see how their websites and search visibility compare with the firms they compete with. We ran the first version for about four years. In 2026 we rebuilt it as a Laravel API with a React front end, and moved every customer’s data across while they carried on logging in. The switchover had no downtime, and to the people using it the new platform looked like a redesign.
What Moves The Price
How many pricing shapes you need. Flat per-seat is straightforward. Usage-based, tiered, metered or hybrid pricing is a build in itself.
Whether customers can self-serve. Signup without a human is a meaningful chunk of work and it is what makes the model scale.
How much isolation you need, per the decision above.
Whether you need an audit trail from day one. For anything selling to regulated buyers, yes, and retrofitting it is expensive.
How many integrations customers expect. Every one is a support surface as well as a build.
Where SaaS Builds Go Wrong
Billing treated as a feature rather than a subsystem. Payments succeed, fail, get disputed and renew asynchronously. Handle that carelessly and you get subscriptions that keep charging after cancellation, or orders with no payment, which are accounting problems as well as customer ones.
Scoping missed on one route. The web app enforces tenancy and the API, the export or the admin tool does not. This is the single most damaging bug this kind of product can carry.
Building for a scale you do not have. Elaborate architecture for a customer base of twelve. The constraint is nearly always finding customers, not serving them.
No plan for the second customer’s awkward request. The first bespoke exception is the moment a product becomes an agency.
Questions To Ask Whoever Builds It
- How is tenancy enforced, and where could it be bypassed? Ask specifically about exports, reports and the API.
- What happens when a payment fails? There should be a described sequence, not a shrug.
- Can a customer cancel and export their own data? Both matter, and the second is what makes the first survivable for your reputation.
- What does the first-run experience look like? If nobody has designed it, adoption will be the problem rather than the features.
- What happens when a customer asks for something the product does not do? The answer defines whether you are building a product or a practice.
A Brief You Can Send Anyone
What it does, in one sentence: [and who pays]
Who the buyer is: [role, company size, how they find you]
What they use instead today: [including spreadsheets and “nothing”]
Pricing model: [per seat, usage, tiers, or undecided]
Self-service or sales-led: [decides how much onboarding work there is]
Isolation requirements: [any customer likely to demand separation]
Integrations they will expect: [the first three]
First milestone: [paying customer, funding, or pilot]
How We Build These
Small, on foundations that survive success. Laravel for the application and its billing logic, React for the interface, PostgreSQL with tenancy scoped structurally rather than by convention, Stripe for payments so card details never touch your infrastructure.
The first release should be capable of taking money from one real customer. Everything after that is informed by what they do, which is worth more than another quarter of guessing.
If you are still testing whether anybody wants it, MVP development is the cheaper starting point. Get in touch if you have the answer already and need the product built.