This piece sits in Polymorph’s series on the seven fundamentals of great products. We’re on the third one: turn ideas into value hypotheses you can invalidate cheaply, before the roadmap becomes a commitment device.
Users rarely pay for features on a spreadsheet. They pay for relief, outcomes, and proof that switching is worth the hassle. Validation is the discipline that keeps that sentence honest.
I first wrote this as the 3rd fundamental on LinkedIn. People don’t wake up wanting “another app with AI and a clean UI.” They wake up wanting hours back from a manual process. If you have to convince buyers they have a problem, you’re probably solving the wrong one, or not enough of a problem to fund change.
You’ve already done the hard listening in customer discovery and behavioural research. Validation asks a sharper question: will they pay, adopt, and sustain the behaviour your economics need?
TL;DR: CB Insights attributes 43% of analysed startup failures to poor product–market fit. Unsustainable unit economics show up in 19%. Validation is how you reduce the odds of building the wrong thing. Atlassian finds only 60% of teams practise regular experimentation, which is the engine most of these methods depend on.
What are you actually validating?
Not “will people like the demo?” Will they pay, adopt, and keep the behaviour your model requires?
Alleviate pain
Time returned, cost removed, risk reduced, errors prevented. Preferably in metrics your buyer already recognises on their dashboard.
Justify the switch
The upside has to exceed migration effort, retraining, and political cost. “Better” loses when change is expensive and the spreadsheet still works.
Nielsen Norman Group’s MVP framing is useful here: a prototype MVP tests a value-proposition hypothesis after you’ve done enough discovery to know the problem and who it hits. You’re not shipping a tiny product for vanity. You’re buying learning about value before you scale engineering.
Six low-risk validation moves (ranked by commitment)
Rank matters. Applause is cheap. Money and signed next steps are not. Work down this list until your riskiest assumption has been hit with something stronger than opinion.
-
Paid pilots, deposits, LOIs
Highest signal. Someone puts budget, procurement, or reputation on the line. Define success criteria before the kickoff: which metric moves, by how much, by when. A pilot without a KPI is a long demo. -
Concierge MVP
Deliver the outcome manually to learn workflow fit before you scale code. It’s slow. It’s also how you find the hidden steps buyers won’t put in a survey. Charge something if you can. Free concierge teaches you effort; paid concierge teaches you value. -
Wizard of Oz / interactive prototypes
NN/g describes Wizard of Oz as a moderated method where the interface looks automated while a human drives the responses. Ideal when the “intelligence” is expensive to build (yes, including AI). You’re testing comprehension and desire before production engineering. Confusion here is cheap. Confusion in production is not. -
Outbound tests with a defined next step
Small, measured outreach. You’re testing message-to-meeting fit, not ego. Count replies that book a workflow review or pilot conversation. Vague “interest” doesn’t count. -
Landing page plus targeted traffic
Measure qualified interest: ICP-shaped visits that take a real next step. Vanity traffic flatters. A crisp CTA (“book a 15-minute workflow review”) tells you whether the promise lands. -
Surveys and friendly demos (use sparingly)
Lowest signal. People overstate interest when saying yes costs nothing. Fine for early language checks. Dangerous as your only evidence.
Gagan Biyani’s Minimum Viable Testing framing (First Round Review) pushes the same idea further: test the riskiest assumptions one by one with the smallest possible test, rather than over-building a “mini product” that still carries login systems, dashboards, and debt. An MVT doesn’t have to look like the eventual car. It has to stress the drivetrain.
What does a sharp value hypothesis look like?
Use a single sentence you can falsify:
For [segment] doing [job], we believe [intervention] will drive [outcome] measured by [metric] within [timebox].
If you can’t name the metric and timebox, you don’t have a hypothesis. You have a mood.
Write the kill/pivot/build rule before you run the test. Weak results without a pre-agreed rule become another quarter of “we need more data.”
One practical habit we push: put the hypothesis on a shared doc with the date, owner, and decision rule. When the test ends, write the result in the same place. No separate “learnings” deck that dies in email. The next sprint should be able to see what you believed, what you saw, and what you decided.
Why does the failure data make validation non-negotiable?
CB Insights tags 43% of studied shutdowns with product–market fit issues. Unsustainable unit economics appear in 19%. Those often sit in the same story: assumed adoption, wrong price, retention that never showed up.
Validation is where you catch economic fantasy early: the spreadsheet that assumes behaviour your second-pass research never saw.
You don’t need a perfect lab. You need a habit of comparing predictions with outcomes. That’s why Atlassian’s experimentation gap matters: without a regular test muscle, “validation” becomes a workshop title on a slide.
Biyani’s MVT point lands for mid-market teams as much as startups. The temptation to “just build a small version” still produces login systems, empty dashboards, and technical debt before anyone has proved the value claim. Test the assumption. Then build the product that assumption earned.
What failure modes show up in “validation theatre”?
- Vanity demos: friendly audiences who will never buy.
- Feature surveys: free yeses that vanish when a PO appears.
- Pilot without success criteria: you “learn” without a decision.
- Fake doors with no follow-up: you measure clicks, then ghost the people who showed up.
- Building the login before the value: weeks of infrastructure, still no proof anyone will switch.
Short scenario A team runs a beautiful prototype tour. Stakeholders nod. Nobody commits to a pilot KPI. Six months later, procurement stalls because ROI was never tied to a number only the product could move. The rebuild cost more than three paid pilots would have.
We’ve seen the opposite work too: a two-week concierge for five accounts, one clear metric, one go/no-go meeting. Ugly. Effective. The roadmap that followed had fewer surprises.
How much validation is enough before production code?
Enough to falsify your riskiest assumption. Usually that’s willingness to pay or adopt, plus clarity on the incumbent workflow you must displace.
If discovery and behavioural research are thin, validation will flop around. Fix the evidence base first. If those are solid and you still only have compliments, you haven’t validated yet. You’ve hosted a tour.
Regulated industries don’t get a free pass. You still test problem urgency, buying process, and evidence standards. Paper workflows, anonymised samples, and Wizard of Oz sessions still teach you what procurement will demand later.
A simple stop rule we like: when two independent commitment signals point the same way (for example a paid pilot and observed workaround abandonment), you have enough to build the next thin slice. When signals conflict, don’t average them. Dig into the contradiction the way you would in behavioural research.
CB Insights reports 43% of analysed startup failures involved poor product–market fit, while 19% cited unsustainable unit economics. Atlassian finds only 60% of teams make experimentation a regular habit. NN/g frames prototype MVPs as tools for testing value-proposition hypotheses after discovery. Put together, those points argue for validation that produces commitment signals and falsifiable metrics before you scale the build.
Where does validation sit in the series?
You validate faster when customer discovery and behavioural research are solid. Next, read competitive analysis for new products, because you’re always competing with the status quo. Overview: seven fundamentals.
If you want validation tied to commercial reality, not slide decks, talk to Polymorph. We work with founders, product leads, and executive teams to validate, build, and improve software that earns its place in the business.
FAQ
What is the minimum validation before we write production code?
Enough to falsify your riskiest assumption: usually willingness to pay or adopt, and clarity on the incumbent workflow you must displace. Commitment signals beat compliments.
Are concierge MVPs only for startups?
No. Enterprises use them to learn workflow fit when integrations and governance dominate the problem. Manual delivery is often cheaper than a wrong integration path.
How do we validate in regulated industries?
Test problem urgency, buying process, and evidence standards. Anonymised workflows, paper prototypes, and Wizard of Oz sessions still reduce risk before you touch production systems.
What metric should we pick?
One your buyer recognises: time, cost, risk, quality. Not your internal feature count.
What should we read next?
Competitive analysis for new products and product differentiation strategy.
