Start building
Ospyr Ventures learning libraryResearch file / published guide

Founder operations / validation

Validate the problembefore the product.

Software can make a process faster. First find out whether the problem is frequent, costly and important enough for someone to change behaviour.

Written by
Ospyr Ventures editorial team
Published
14 November 2026
Link check
14 November 2026
Reading time
11 minute read

Ospyr Ventures has an interest in the topics covered. The article does not claim Ospyr client results or a named founder testimonial. The named editorial reviewer is pending. See our editorial standards and corrections process.

In brief / 40 to 60 words

The direct answer

Before building software, verify that a specific group repeatedly faces the problem, how they solve it today, what that costs them, who can approve a change, and whether a simpler service or prototype produces value. Test behaviour, not compliments. Keep an evidence log and define what result would justify building, changing the idea or stopping.

How we would approach it

We would look for repeated behaviour before writing specifications. Ospyr would interview the people with the problem, observe their workaround and test a simpler service or prototype. Software becomes justified when the evidence shows what must be repeated at scale.

See how Ospyr Ventures builds

What does a real problem look like in evidence?

Name one group of users and one recent situation in which the problem occurred. Ask them to reconstruct the last occurrence: what triggered it, what they did, who else was involved, what it cost in time or money, and what happened when they did nothing. A statement that the idea is interesting is not the same as a recurring problem. A concrete workaround is often more useful evidence than a feature request.

Keep the interview grounded in past behaviour. Do not lead with a demonstration or ask whether someone would use an imagined app. When a respondent says the problem is urgent, ask what they have already tried and what budget or authority exists to change it. Record exact language and the source of each claim so later design decisions can be traced back to people, not memory.

What are people using instead of the proposed software?

Map the current workflow, including spreadsheets, messages, staff time, competing products and decisions to accept the problem. The alternative to your software may be a person, a process or no action at all. Find the point where the cost becomes visible to the person who can approve spending. A user who feels friction may not be the buyer; a buyer who signs the contract may not use the product.

Price the workaround carefully. Count hours, errors, lost sales or delays only where records support them. A large hypothetical total assembled from assumptions is not customer evidence. If people tolerate the problem and have no reason to change, adding features will not create urgency.

AssumptionSmall testStronger signalWeak signal
Problem recursAsk for recent examples and recordsRepeated dated casesGeneral agreement that it sounds painful
A buyer wants changeRequest a paid pilot or procurement stepNamed budget owner and next actionPraise without a commitment
Workflow can improveDeliver the outcome manuallyUser returns or pays for another cycleOne guided demonstration
Software is neededCompare manual and prototype flowsRepeated value blocked by manual capacityFounder prefers to code

What can you test without building the full system?

Offer the outcome through a manual service, a simple form, a clickable prototype or a limited pilot. Choose the smallest test that exposes the riskiest assumption. If the risk is whether a customer trusts the result, a polished dashboard is unnecessary. If the risk is whether a user can complete a complex task unaided, a static sales page cannot answer it. A prototype must resemble the decision you need to study.

Tell participants what is manual and what is simulated. Do not present a human-run workflow as automated. Record the time and cost to deliver each result, the errors, and whether the user comes back without persuasion. Those numbers help decide whether software would remove a real constraint or merely hide an expensive service behind an interface.

When is there enough evidence to start building?

Write the decision rule before interpreting results. For example, a defined number of eligible customers complete a manual pilot, a proportion repeat the use, an identifiable buyer takes a purchasing step, and the unit cost has a plausible path down with software. The numbers must fit the market and the cost of a wrong decision; no universal interview count makes an idea valid. Include a condition that would make you change or stop the concept.

Then build only the part that removes a proven constraint. Keep the evidence log linked to each proposed feature: who experienced the problem, how often, how the current workaround failed, and what the small test showed. If the team cannot connect a feature to that evidence, defer it. After release, continue checking actual use and payment rather than treating the first build as validation's end.

  • Named user and buyer
  • Recent problem and current workaround
  • Observed value from a manual or prototype test
  • Evidence of a purchase or committed next step
  • Written build, change and stop criteria

A useful next step

Interview five eligible people about a recent occurrence, map their current workaround, and run one small test before specifying features.

Terms worth knowing

Workaround
The method a person uses today to handle a problem, even when it is slow or imperfect.
Prototype
A limited representation of a proposed product used to test a specific question.
Validation
Evidence that a defined customer has a problem and will take meaningful action for a solution.

Watch, read and test.

YC lesson

How to Get Startup Ideas

Watch for how founders locate problems worth investigating.

Open resource
YC essay

Do Things That Don't Scale

Read why manual delivery can expose what a product must eventually automate.

Open resource
Design method

Simulating intelligence

See how a prototype can test a difficult product assumption.

Open resource