//Product

How to scope an MVP you can actually launch in eight weeks

Most MVPs miss their date because they try to prove five things at once. Here's how to find the one thing yours must prove, and cut the rest. · 6 min read · Loop Craft Engineering Team

You have a date: an investor meeting, a pilot customer, a conference. And you have a feature list that would take a year. The gap between them is where most MVPs fail, not because the team is slow, but because the scope was never really decided.

Find the one thing it must prove

An MVP is an experiment. Before listing features, write one sentence: "We believe that [these users] will [do this] because [reason]." Everything in the first version should help prove or disprove that sentence. Everything else waits.

Cut with three questions

  1. Would a user notice if this was missing on day one? If not, cut it.
  2. Can we do this by hand behind the scenes for the first 50 users? If yes, don't build it yet.
  3. Does this help prove the sentence above? If not, move it to the list for later.

Things that are usually safe to postpone: admin dashboards, complex permissions, multiple payment options, deep analytics, and anything described as "nice for later".

Design the risky parts first

Every MVP has one or two parts nobody is sure about, like a tricky integration, an unusual interaction, or an AI step. Prototype those in the first two weeks, while changes are cheap. Leave the familiar parts, like sign-up and settings, for later.

A realistic eight-week shape

  • Weeks 1–2: agree the sentence, the scope and the success measure; prototype the risky parts; set up the repository, hosting and deployment pipeline.
  • Weeks 3–6: build in weekly loops, with a working demo on a real URL every week and the scope re-checked at every demo.
  • Week 7: test with real users, fix what they trip over, and add monitoring.
  • Week 8: launch to your first group of users, measure, and plan the next loop.

Decide what success looks like before you launch

Pick one or two numbers you'll check after launch, like how many invited users complete the core action, or how many come back in the second week. Without them, it's impossible to know whether the MVP worked or what to build next.

What we'd tell a friend: Ask any team you hire to show you working software in the first two weeks, and make sure the code lives in your own repository from day one. Those two habits prevent most MVP disasters.

Facing this in your business? Talk to an engineer.

Send a short note about your process. We'll reply within one business day with practical next steps, even if that's "you don't need us".

  • Weekly demos
  • No lock-in
  • You own the code
  • 1 business day