The problem with 'MVP' as a term is that it's been diluted to mean 'anything smaller than the final vision'. A real MVP is much narrower: one core loop, one type of user, one workflow, one paying customer proof point. Everything else is scope creep dressed as ambition.
Here's the 90-day plan we run with VC-backed founders. It works because every week has a specific artefact — no hand-waving, no 'we're building infrastructure'.
Weeks 1–2: Discovery
Not brand strategy. Not user personas. **One question**: what is the smallest thing a real customer will pay for? Interview 5-8 target users, define the core loop, write it up as a one-page brief. If the brief is more than one page, you haven't decided yet.
Artefact at end of week 2: a one-page product brief, a wireframe of the core loop, and a landing page that describes the MVP.
Weeks 3–4: Design + technical spike
Design the core loop in Figma to a level detailed enough to build from. In parallel, a senior engineer builds throwaway spikes on the technical unknowns — usually 2-3 things: 'can we integrate with X vendor cleanly?', 'does the AI feature actually give useful outputs?', 'is the third-party API reliable enough?'.
Artefact at end of week 4: high-fidelity design of the core loop + risk register with each technical unknown marked green/yellow/red.
Weeks 5–8: Build v0
Two senior engineers, one designer, one PM. Weekly demos on Fridays. The goal is a functional core loop that a friendly customer can actually use — not something polished.
The single most important discipline in these four weeks: **do not add anything to the scope**. Every 'wouldn't it be cool if' idea goes on the v1 list. Founders who cannot enforce this end up 4 months in with a broken v0.
Artefact at end of week 8: internal build, deployed to a staging env, walk-throughable end-to-end.
Weeks 9–10: Design partner rollout
Onboard 3-5 design partners. Not a beta group. Specific individual users who have committed to using the product for real work. Watch them use it. Log everything. Fix the top 5 friction points.
Artefact at end of week 10: 3-5 real users, at least one with recurring usage.
Weeks 11–12: Ship + pricing
Public launch: landing page live, pricing set, Stripe integrated, first paying customer signed. This is the artefact of the whole engagement — not 'a working product' but 'a paying customer'.
Founders who skip pricing until 'we have more data' are the ones who spend another 6 months building. Pricing is a research tool, not a monetisation tool. Set a number, be prepared to change it.
What kills MVPs in 90 days
1. **Scope creep from the founder**. The bar for adding anything to the v0 backlog has to be extreme. 2. **Underestimating deployment work**. Auth, payments, SSO, email — allocate 15-20% of engineering time to plumbing. 3. **Not scoping AI features tightly enough**. LLM features are magic in demos and messy in production; if AI is core to the MVP, treat it as its own workstream with its own evals from day one.
The Origami MVP model
We run this 90-day plan as a fixed-scope engagement. Two senior engineers, one designer, one product lead, one QA. Weekly demos, transparent Jira, direct engineer access. Typical budget $150k–$220k depending on integration complexity. If you want the paying-customer version of an MVP in a quarter, book a scoping call.