Before You Build a Prototype, Can You Answer This One Question: “What Does It Need to Prove?”

[thomashellby]

I’ve been thinking about why so many early-stage prototypes end up as expensive demos that don’t teach the team anything. The pattern is usually the same: a founder has a solid idea, hires a team, and jumps straight into screens and features. Three months later there’s a polished product that looks great but doesn’t answer the questions that mattered.

My take is that most of those failures start before a single line of code, with the brief. Or more accurately, with the lack of one.

A prototype is a learning tool, not a mini product

The most useful question before any build is: what do we need this prototype to prove? Maybe it’s that users understand the core value in under a minute, that a booking flow can be completed without help, or that a risky technical integration is feasible. Each of those leads to a very different prototype. A clickable design and a working backend demo are not interchangeable, and treating them as the same thing is where budgets quietly disappear.

Where I see briefs go wrong

• Describing the solution without the problem. A paragraph about an “AI-powered ecosystem” tells a team nothing. “Customers can’t compare local providers without calling five of them” tells them exactly what to build.

• A feature list instead of priorities. If everything is a must-have, nothing is. Splitting features into must-have, should-have, and could-have gives everyone a way to make trade-offs when time or budget gets tight.

• Never defining “simulated vs. working.” A checkout screen that demonstrates the experience is very different from one that actually processes payments. If this isn’t written down, expect a disagreement at delivery.

• No acceptance criteria. “Prototype delivered” means different things to different people. Something like “a user completes the main booking journey on mobile without assistance” is something you can actually verify.

• Ignoring ownership. Who owns the designs, source code, and assets once the work is done? It’s boring, but it’s much easier to settle before work starts than after.

The unglamorous parts that save money

• Name the primary user group specifically. “Small restaurant owners managing online orders” beats “anyone online.”
• Use synthetic data early on unless you truly need real data, which also sidesteps security headaches.
• Decide who has final say on feedback. A lot of project delays come from unclear approvals, not from development.
• Define what the prototype will not prove. A clickable demo can show a workflow is understandable. It can’t show people will pay, or that the system handles thousands of users.

Choosing a partner

If you’re working with outside help, it’s worth comparing more than price. Look at scope, how feedback and revision rounds are handled, what’s actually delivered, and what happens when requirements change. A cheaper quote that leaves major responsibilities unclear often costs more later. Founders who don’t have in-house product or engineering experience often lean on a prototype development company for startups to help define scope, user flows, and technical priorities before committing to a build. Teams like Triple Minds, which cover development, design, and AI consulting together, are an example of the one-stop model some founders prefer over coordinating separate vendors, though the right fit depends on your project and how much structure you already have.

Either way, a clear brief makes any partner more effective, because they’re estimating real work instead of guessing.

Let’s discuss
I’d like to hear from people who’ve been through this:

• Have you built a prototype that taught you nothing, and what do you think went wrong?
• How do you decide what a prototype should not include?
• Do you prefer a clickable design for early validation, or a working technical build, and what made you choose it?
• Founders: did writing a brief change how your project went? Developers: what’s the one thing you wish clients put in theirs?

Real examples welcome, especially the messy ones. Those tend to be the most useful.