What should you validate before making a significant product investment?
A practical look at the evidence worth gathering before committing significant time, budget and resources.
A company is considering a meaningful investment: a new feature, a redesign, an AI capability, a move into a new market, a change to a core workflow, the next major version of the product. The team has opinions. Some of them have data attached. The budget is real — and so is the opportunity cost of spending it wrongly.
What should be validated before committing?
Not everything. Not every assumption deserves research — research applied indiscriminately is just delay with a respectable name. The objective is narrower: identify the assumptions that could materially affect the product decision, and examine those, in proportion to what is at stake.
Start with a different question
Before asking “how should we build it?”, ask:
That question turns a plan into a set of assumptions — about the problem, the people who have it, the context it lives in, and the value of solving it. Assumptions can be examined. Five questions do most of the work.
01 — Is the problem real?
What evidence demonstrates that the problem actually occurs? Not that it could occur, or that it appears in a strategy deck — that it happens, to identifiable people, in identifiable situations. Observed behaviour, recurring support requests, the workarounds people have built for themselves: these are the signals that a problem exists in the world rather than in a slide.
02 — Does it matter enough?
Frequency alone does not establish value. A problem can be constant and trivial, or rare and severe. The question is significance: what does the problem cost the people or organisations experiencing it — in time, money, risk or missed opportunity — and what do they do about it today?
Existing workarounds are informative. People who have built their own solution are telling you the problem is worth something. People who live comfortably with it may be telling you the opposite.
03 — Are we solving the right part of it?
A team can correctly identify a problem and still invest in the wrong intervention. The problem is real; the chosen slice of it is not where the value sits. It is worth asking which part of the problem, if solved, would change the outcome most — and whether the proposed investment is aimed at that part, or merely at the part that is easiest to build.
04 — Will this work in the real context?
A solution that performs beautifully in a demo can fail in the environment it has to live in. The real context includes behaviour — what people actually do, not what a process diagram says they do. It includes the workflow the solution must fit into, the other services and systems it depends on, the organisational constraints around it, and the question of trust: whether people will rely on it enough to change what they do.
Adoption is part of the product. An elegant solution that nobody adopts is an expensive ornament.
05 — What evidence would change the decision?
Research should not simply accumulate information. Before gathering evidence, it is worth agreeing what the evidence will decide. The findings should help determine whether the company should:
The evidence supports moving forward as proposed.
The direction is promising, but specific assumptions need to be addressed first.
A different proposition, experience or approach may create more value.
There is not yet enough evidence to justify this level of investment.
If no plausible finding would change the plan, the exercise is not research. It is decoration on a decision that has already been made.
A note on the strength of evidence
Not all evidence carries the same weight, and it is useful to notice the difference. Loosely:
Money, commitment, a contract or meaningful commercial action.
Real usage in the real context.
Behaviour in response to a prototype, proposition or experiment.
What people are actually seen to do.
What people say they do, need or want.
Concluded from adjacent evidence.
Believed, but not yet examined.
This is not a universal scientific scale, and not every decision needs every form of evidence. The appropriate evidence depends on the decision and the uncertainty around it. A small, reversible choice may need little more than reported and inferred evidence. A large, hard-to-reverse commitment deserves something observed, tested or behavioural before the money moves.