The MVP was an event. Now it is a loop.
For a decade the minimum viable product was treated as a milestone. You scoped the smallest thing that could test demand, shipped it, called the venture validated or not, and moved on to real building.
That framing made sense when shipping was expensive. It makes less sense now.
When a working release costs days instead of quarters, the question stops being "did we build the MVP?" and becomes "how fast can we run the next honest test?" The MVP did not disappear. It turned into a standing loop: build, release against a written hypothesis, read the evidence, update the thesis, repeat.
Ventures that run this as a loop behave differently from ventures that treat it as a phase. They ship smaller. Arguments come from evidence sooner. Weak ideas die before they accumulate headcount.
Start with a thesis doc, not a roadmap
Every venture in a loop keeps one document above all others: the thesis.
A usable thesis doc fits on a page. It states:
- who the customer is and what job they are hiring for
- why now, meaning the shift that makes this possible today
- the mechanism: how the product creates value and captures some of it
- the riskiest assumption currently load-bearing
- what evidence would change our mind
The last two lines matter most. A thesis without a named risky assumption is a pitch. A thesis without falsification conditions is a belief.
The doc is versioned. When evidence arrives, you edit the thesis and date the edit. Over months this becomes the venture's real history, more honest than any deck.
Rule of thumb: if the team cannot state the current riskiest assumption from memory, the loop has drifted back into feature shipping.
Every release is an experiment
In a loop, nothing ships just to ship. Each release carries a hypothesis written before the work starts:
> We believe [change] will produce [observable behavior] among [specific users]. We will know within [timeframe], because we instrumented [signal].
Small format, big discipline. Writing the hypothesis first does three things. It forces you to define success before you are emotionally invested in the output. It tells you what to instrument. And it makes the post-release conversation short: did the predicted behavior show up or not?
Then close the experiment with a readout, even three sentences: hypothesis, result, thesis updated or not. Experiments that never get read out are theater. The readout is where learning actually enters the system.
Two failure modes to watch:
- The unfalsifiable release. The hypothesis predicts nothing observable. Ship it if you must, but call it maintenance, not an experiment.
- The zombie experiment. Results arrived weeks ago; nobody wrote the readout. The loop stalls until someone closes it.
Measure learning velocity, not just shipping velocity
Most teams measure output: releases, story points, features shipped. In a loop, the metric that matters is how fast beliefs get tested and updated.
Practical instrumentation:
- cycle time per experiment, from hypothesis written to readout written
- share of releases that carried a written hypothesis
- thesis revision rate: how often the doc changes because of evidence rather than marketing
- assumption age: how long the current riskiest assumption has gone untested
None of this needs a data team. A table at the bottom of the thesis doc, updated weekly, is enough at studio scale.
What you are watching for is drift. If cycle time stretches, experiments are getting heavier than the team admits. If hypothesis coverage drops toward zero, you are building on faith again. If the thesis never changes despite months of releases, either the venture is boringly correct or nobody is reading the evidence honestly. Both deserve a conversation.
Studios run loops, not projects
For a venture studio the unit of management shifts again. You are not running one loop. You are running a portfolio of loops at different speeds and confidence levels.
Portfolio practice that holds up:
- One board showing every venture's current experiment: hypothesis, owner, due date, status.
- Weekly review keyed to evidence, not activity. Each venture reports its latest readout, not its busyness.
- Capital follows confidence. Resourcing decisions reference the evidence ledger: which assumptions survived contact with users, which collapsed.
- Shared instruments. A common hypothesis template and readout format lets you compare learning speed across ventures without pretending they are comparable businesses.
The studio's edge is not building faster than founders. It is refusing to let any single venture hide from its own evidence for more than a week or two.
When the loop proves the venture should die
A healthy loop occasionally produces the answer nobody wanted: the thesis is wrong in a way that matters.
Signals that the loop is telling you to stop:
- The core mechanism failed repeatedly under fair tests, and each rescue required changing the customer, the job, or the economics to make it work.
- Learning velocity collapsed. Experiments take longer, results get softer, and the team argues about process instead of evidence.
- The riskiest assumption has been rewritten so many times it no longer resembles the original thesis. Pivots compound; past some point, continuity is fiction.
- Nobody can say what would have to be true for this to work, or everything would have to be true.
Killing well is an operator skill. Write the decision down with the evidence that drove it. Recycle what has value: code, audience insight, distribution learnings, people. Announce it plainly. A studio that kills cleanly earns the trust to fund boldly.
The opposite failure is quieter and more expensive: a venture that never dies because it never runs a decisive test. Slow loops do not protect fragile theses. They embalm them.
The sharper frame
The MVP answered one question once. The loop answers better questions continuously: what do we now believe, what changed our mind, and what will we test next?
Build the thesis doc. Attach a hypothesis to every release. Close every experiment with a readout. Manage the portfolio by evidence. When the loop says stop, stop.
That is what testing a venture continuously actually looks like. Less ceremony than the old launch playbook, far more honesty per week.
If your team is shipping faster than it is learning, the missing piece may not be more output. It may be a loop with enough structure to make the evidence impossible to ignore. If you want help building that loop, book a discovery call.