The gap between a working prototype and a first paying customer is almost never technical. It is a sequence problem: work done in the wrong order produces a polished product nobody has agreed to pay for. This is the order that works, with an exit condition for each stage so you know when to stop.
Stage 1: Idea to Something Real
Target: two weeks. You are not building a product, you are building the one workflow that produces the outcome someone described to you. No signup, no billing, no settings page. If it takes a shared spreadsheet and a script you run by hand, that is the correct first version.
The only thing worth building carefully here is the data model, because it is the one decision that is expensive to change later. Everything else is disposable and should be treated that way.
Exit condition: one person who is not you has produced the outcome using your thing, and can describe what it saved them.
Stage 2: First User
Now add exactly three things: authentication, the ability to use the workflow without you present, and a way for the user to reach you. Magic-link login, one screen, and your email address. That is the whole scope.
Do not add onboarding flows, tours, or empty states with illustrations. With one user, onboarding is a 20-minute call, and that call tells you more than any analytics tool will at this stage.
Instrument two events only: the workflow was started, and the workflow completed. The ratio between them is the only metric that matters until you have money.
Exit condition: that user returns and completes the workflow twice in a week without being prompted.
Stage 3: First Paying Customer
Charge before the product is ready. The price conversation is a research tool: it converts “this is interesting” into a number, and the objections you hear are your roadmap.
Take the payment the crude way first. A payment link and a manually flipped flag in the database is a legitimate stage-three billing system, and it takes an hour instead of a week. Build the subscription integration once you have two or three customers and the manual step becomes annoying.
Price higher than feels comfortable. Underpricing at this stage does not get you customers, it gets you customers who will not fund the work, and raising a published price later costs goodwill you do not have yet.
Exit condition: money has moved, and the customer knows what happens if they want to stop.
Stage 4: First 100 Users
This is where the systems work belongs, and not before. In order: real subscription handling with webhooks and a ledger, rate limits on outbound provider calls, the five alerts that page you, and a background queue for anything that takes more than a second.
Add one operational habit alongside it: a weekly 30-minute review of what broke, what a customer asked for twice, and what you spent manual time on. The third column is your automation backlog, and it is more reliable than any feature request list.
Resist multi-tenancy features, roles and permissions, and an admin panel until a customer’s contract requires them. Every one of those is a permanent maintenance obligation that adds nothing to the workflow people are paying for.
Exit condition: you can onboard a new customer without touching the database, and a deploy on Friday does not scare you.
The Ordered Playbook
- Build one workflow end to end with no auth, no billing, and a data model you thought about for an afternoon.
- Get one non-you person through it and write down, in their words, what it replaced.
- Add magic-link auth, one screen, and a direct line to you. Nothing else.
- Instrument two events: workflow started, workflow completed.
- Ask for money before the product is ready, using a payment link and a manual entitlement flag.
- Price at the number that makes you slightly uncomfortable, and write down every objection you hear.
- At two or three paying customers, replace the manual flag with real subscription webhooks and a ledger.
- Add outbound rate limits, a background queue, and the five alerts that page you.
- Start a weekly 30-minute review: what broke, what was asked twice, where your manual time went.
- Automate only what appeared in that third column two weeks in a row.
Every stage above is ordered so that the cheapest information arrives first. The most common failure is doing stage four work during stage two, which produces a well-engineered system with no customer attached to it.
Get articles like this by email
Production patterns, checklists, and failure stories for people shipping SaaS. No spam.