How we work. The five steps, what each one does and what we need from you.
A process told only from the builder's side reads as if the other party does nothing. This one also says what we need from you at each step, which is what makes the deadline hold.
Discovery
Half an hour on a call to work out what your company does and which part hurts. No three-week audit and no forty-question form: most projects get defined by hearing how one concrete thing is done today and where it breaks. Out of that comes what gets built first, what waits, and what it costs, all three closed.
What we need from you
- Walk us through how the process works today, odd case included: the odd case is the one that breaks the system later
- Tell us who's going to use it and what each of them can do
- Show us the spreadsheets or the system you have today, messy as they are
What the step ends with: The scope in writing, the closed price and the date it goes live.
Design and architecture
The screens get drawn and the data model gets defined. You see them before a line of code exists, which is when changing something costs a conversation: the same fix on something already built costs a week, and that's why this step doesn't get skipped however tempting it is.
What we need from you
- Look at the screens and say what's missing or what doesn't belong, in the words your operation uses
- Confirm the statuses are the ones people actually use, not a tidied-up version of reality
- Sign off, which is what lets the building start
What the step ends with: The approved screens and the data model, which is what everything gets built against.
Development
We build, and every week there's something you can click on. Not a progress report: the system running, as far as it has got. We use Claude Code, Codex and Cursor on Neon and Supabase, which is what makes the timeline what it is.
What we need from you
- Log into what's there each week and try it like it's yours, because it is
- Say something early if it isn't what you expected: week one is cheap, week four isn't
- Start gathering the real data that's going into testing
What the step ends with: The system in an environment you can log into whenever you want, updated every week.
Testing
It gets tested with your data and your people, not just ours. We load real cases —including the ones that go wrong— and someone from your team tries it before it ships. This is the step where the things nobody mentioned on the first call turn up, and turning up here is exactly what it's for.
What we need from you
- Put someone from the team on it for a few days, using it, not watching it
- Send us the real data, including the old, messy cases
- List what fails, which gets fixed before launch
What the step ends with: The system tested with real data and the list of fixes, closed.
Launch and support
It goes live on infrastructure in your name, with the code in a repository you own and the team trained. From there we stay on if you want us to —maintenance monthly or per piece of work— but the system doesn't stop working if you stop hiring us, which is the point.
What we need from you
- Be at the training with the people who'll use it every day
- Take the keys to the infrastructure and the repository
- Decide whether you want maintenance, after it's already running and not before
What the step ends with: The system in production, the code, the keys, the documentation and the team trained.
Does any of these look like what you need?
Half an hour, free, to understand what you need. If we can build it, you leave the call with the scope, the timeline and the price; if we can't, we tell you right there.