Product and system design
You know what you want to build. This is where it becomes buildable.
The product decided, the screens drawn, the data model and the architecture written down, and a build plan with real estimates — including the parts you should not build. Yours to hand to us, to your team, or to an AI tool that will finally have something to follow.
Already written it down? Attach the PDF or the Markdown file. A spec you already have is worth more than any form.
- How it starts
- A written brief or a file you already have, then a call. Both free.
- Who does the work
- Nikola, plus a named collaborator where production design is involved — named before they start, never a bench.
- What you get
- Artefacts you own and can hand to any team, including one that is not us.
- Payment
- Split across agreed milestones. Nothing hourly by surprise.
How a project works
Why this is a separate piece of work
Every expensive mistake in software is made before anyone writes code
Not because people are careless. Because these decisions look like details until the moment they are load-bearing, and by then they are in the schema, in customers' habits, and in six months of code.
- The data model decided by the first screen
- Whatever the first feature needed became the shape of everything. The second feature fits badly, the fourth needs a migration nobody has time for, and the fifth is quietly dropped.
- Permissions bolted on afterwards
- Who can see what is a design decision, not a feature. Added later it becomes a predicate copy-pasted into forty queries, and the one place it was forgotten is the incident.
- Scope that only grows
- Without a written definition of what the product is, every reasonable request is in scope. The launch date moves twice and then stops being discussed.
- An AI tool with nothing to follow
- Lovable, Cursor and Claude Code build fast and build generically, because nobody gave them a system. The same tool with a design system and a data model produces your product instead of everyone's.
- Integrations designed for the happy path
- What the payment provider does when it is down, what the webhook does when it arrives twice, what the import does with a bad row. These are decisions, and they are made either now or during an outage.
- Estimates that were never estimates
- A number produced without a design is a hope. The point of this work is that the plan at the end has hours attached to things somebody has actually thought through.
What it costs
Priced by how big the thing is, not by how long we can make it take
Three sizes, because that is how these projects actually arrive. Pick the row you recognise — if you are between two, say so in the brief and we will tell you which one it is before you commit to anything.
One thing, done properly
$5,000–9,000
15–25 hours
A single-purpose tool, one kind of user, a handful of screens. An internal tool, a calculator, a booking flow, a first MVP with one job.
- The product one-pager: user, the one job, the metric it moves
- The flows that carry the risk, wireframed
- Data model and the API shape
- A build plan with real estimates, and what to leave out of v1
A real product
$12,000–25,000
35–70 hours
Several kinds of user, permissions that matter, payments, a third-party integration or two. Most funded startups are here.
- Everything above, across the whole product rather than one flow
- Roles and permissions designed before they are a schema migration
- Architecture decision record: what was chosen, what was rejected, why
- Integration design, including what happens when the other side is down
- Screens drawn to a design system your team or your AI tool can follow
- A sequenced build plan — what ships first and what it unblocks
A platform
$30,000–60,000
85–170 hours
Multi-tenant, an existing system to migrate from, compliance in the room, or a team that will build it who need to agree with each other first.
- Everything above, plus the parts that only appear at this size
- Tenancy and data isolation, decided and written down
- Migration path from whatever exists today, with a rollback
- Security and compliance posture — the answers, before the questionnaire
- Interfaces between teams, so two of them can build in parallel
- Workshops with your engineers, because a plan they did not argue with is a plan they will not follow
These are typical ranges, not a price list. They are $300–$400 an hour — the rate on this site — multiplied by the hours a project of that size has actually taken us. The number for yours comes after the call, in writing, once we have read your brief.
What you end up holding
Artefacts, not a presentation
Everything here is something a person can act on. There is no deck, and nothing is delivered in a format that requires us to be in the room to explain it.
- The product one-pager: who it is for, the one job, the metric it has to move
- User flows for the paths that carry the risk, with the states nobody designs — empty, loading, error, permission-denied
- Wireframes or screens, depending on the size, drawn to a design system rather than one at a time
- The data model: tables, relationships, and what each one refuses
- API shape — the endpoints, what they return, and what they do when they fail
- An architecture decision record: what was chosen, what was rejected, and why, so the next engineer does not relitigate it
- Security and access model, written before it is a migration
- A sequenced build plan with hours, and an explicit list of what to leave out of v1
- Everything in a repository or a Figma file you own from the first day
How it runs
Two to six weeks, depending on the size
01
Send the brief
A few sentences, or the spec you already wrote, or screenshots of what exists today. Free, and it is what the call is based on.
02
A call
Forty-five minutes on what you are building and why. You leave knowing which size band this is and what the number is likely to be, whether or not you continue.
03
A written proposal
Scope, deliverables, milestones and one price. Agreed before anything starts, and nothing hourly appears later.
04
The work, in the open
You see it as it is made rather than at a reveal. Decisions get argued while they are still cheap to change.
05
Handover
A working session with whoever builds it — your team, us, or a tool. A plan nobody argued with is a plan nobody follows.
06
Build, if you want
We can build it, or not. The design is complete either way and is deliberately written so that a team who has never met us can use it.
When this is the wrong thing to buy
- You have one broken thing in a working app. That is /fix, it is cheaper, and the guide is probably already on this site.
- You want a designer to make existing screens prettier. That is production design, and a good freelancer will do it better and for less.
- You need the decisions argued every week rather than once. That is the monthly arrangement, not a project.
- You already know exactly what to build and want it built. Skip this and go straight to a build conversation — paying to be told what you already decided is a waste.
- You want a fixed price for the build before the design exists. Nobody honest can give you that, and anyone who does is pricing in the risk.
Why take this from us
Because we then had to live inside the decisions
Golux has built and still operates fifteen products of its own — an AI planning platform, a portfolio monitor, monitoring SaaS, a booking marketplace, an agriculture platform. Every design decision in them came back as a support ticket, a confused user, or a schema migration on a Sunday. That is a different teacher from a portfolio, and it is why the questions here are about the decision rather than the drawing.
- 15
- products designed, built and still operated by this team
- 15+
- years shipping production software
- 208
- AI-built apps audited, and what they got wrong published
- 1
- person accountable, collaborators named before they start
Questions
Product and system design, in detail
- Can you design something that does not exist yet?
- That is most of this work, and it is the cheapest moment to do it. Nothing has been built, so nothing has to be unbuilt — the decisions are still free to change.
- We have an app already. Does this still apply?
- Yes, and it is usually a redesign of the parts that are blocking you rather than the whole thing. Run the free audit first; it often names the actual constraint, and sometimes the honest answer is that you need one fix rather than a design engagement.
- Do you build what you design?
- Often, but it is deliberately not a condition. The deliverables are written so a team that has never met us can build from them, which is also the test of whether they are any good.
- What if our engineers disagree with the architecture?
- Good — that is what the workshops in the larger bands are for. A plan the team did not argue with is a plan they will route around. The decision record exists so the disagreement happens once, in writing, rather than every sprint.
- Will this work with Lovable, Cursor or Claude Code?
- It is one of the best uses of it. Those tools are fast and generic because nobody gives them a system to follow; a data model, a design system and a written build sequence turn them into something that produces your product rather than a template.
- Why a range and not a price?
- Because we have not seen your brief. The ranges are the hourly rate on this site multiplied by hours projects of that size have actually taken, and they are published so you can decide whether to start the conversation. Your number comes after the call, in writing.
- Who owns the output?
- You do, from day one — your repository, your Figma, your documents. There is no licence and nothing that requires us to stay.
- How fast can you start?
- Usually within a week or two. We take a small number of projects at a time so the work stays fast; if there is no room this quarter you will be told rather than queued.
Related
Product design consulting
The same thinking as a monthly arrangement rather than a project.
Product design & UX
The practice behind this: research, prototypes, design systems.
Brand design
The identity around the product — logo, colour, type, tokens.
Product partner
If you want product, design and engineering as one ongoing arrangement.
Free production audit
Already have an app? Start here. Sixty seconds, no account.
Fix one problem
One broken thing rather than a design engagement.
Send it over
Tell us what you want to build
A few sentences is enough to start. If you have already written a spec, a PRD, a deck or even a long message to yourself at 2am, attach it — it will tell us more than any set of form fields.
One person reads every brief, usually the same day. You get an answer on which size band this is and what it would take, whether or not you continue.
