The Questions We Ask

We treat answers as the valuable part of thinking. The answer is what gets presented, implemented, approved and, occasionally, remembered.

But the quality of the answer is usually decided much earlier, by the quality of the question.

A precise answer to the wrong question is worse than useless. It is a beautifully engineered will-o’-the-wisp that leads a whole team confidently into the marsh.

For a long time I had this backwards. I assumed expertise was measured by the quality of someone’s answers — the people I admired always seemed to know what to do, diagnosing problems quickly and deciding with enviable confidence. Then I noticed that the people I respected most were rarely the quickest to answer. They were the quickest to ask the question nobody else had thought of.

Answers close things down. A good question opens them up.

This happens constantly in software. A stakeholder asks for a new report, so we dutifully discuss columns, filters and export formats. Weeks later it emerges that nobody wanted a report at all. Someone needed to know which customers required attention each morning.

The original request described a solution. The useful question concerned an outcome.

The distinction sounds obvious, but solution-shaped questions are seductive. They provide something concrete to design. They allow estimates, diagrams and tickets. They create momentum — which feels wonderful, even when it is aimed in entirely the wrong direction.

“Build a dashboard” feels actionable.

“What decision are we trying to improve?” is slower, vaguer and faintly irritating.

Yet the second question may save months of work.

The questions also change as you go on. Early in my career mine were almost entirely technical: which algorithm, which database, which design pattern. Useful questions, all of them — just rarely the important ones. These days they sound different: what outcome are we actually after? What are we quietly assuming? Who is this really for? What happens if we are wrong? Experience did not replace the technical knowledge. It simply wrapped it in slightly better judgement.

Good questions expose hidden assumptions. They also quietly decide the search space.

Ask “How do we automate this process?” and the answer will, astonishingly, involve automation.

Ask “Why does this process take so much effort?” and the answer might be automation, removal, simplification, better data or a quiet change in who is responsible.

This applies well beyond engineering.

In an argument, “How can anyone possibly believe that?” politely invites an answer about ignorance or bad faith. A more useful question is, “What experience or risk makes that position feel reasonable to them?” It does not require agreement; it simply produces a better model.

In management, “Why did they fail to follow the procedure?” points neatly toward correction or discipline. “Was the procedure visible, workable, understood and remotely compatible with the pressures of the actual job?” points at something systemic instead of merely someone to blame.

Two questions I now carry everywhere.

The first:

What would have to be true for this to make sense?

It is invaluable whenever a decision looks irrational. Perhaps the person has different information, or is optimising for something we have not considered, or the incentives are producing exactly the behaviour any sane person would have predicted.

The second:

What evidence would tell these explanations apart?

Without it, discussion becomes a contest between stories, each fitting the visible facts. The next step is not a louder opinion, but a test.

Questions also shape culture. A leader who asks “Who caused this?” builds a very different organisation from one who asks “What allowed this to happen?” Both may matter, but sequence does too: if blame arrives first, information develops a remarkable talent for disappearing.

All of this matters more now, not less. AI has made answers astonishingly cheap — ask, and you get code, documentation, a design or an explanation in seconds. That does not shrink the value of a good engineer; it moves it. When anyone can generate a plausible answer, the scarce resource becomes the quality of the question, the challenge and the follow-up. The fastest answer stops being the differentiator. The most interesting question becomes it.

It also takes a certain nerve to ask a real question — to admit, out loud, that you do not yet know. I have made my peace with that. Asking before you have formed a conclusion is not weakness; it is how understanding grows. Learning out loud, if you like.

The aim is not endless enquiry. Questions can become their own species of procrastination dressed as rigour, and eventually you have to choose, act and accept the uncertainty.

But a good question does more than request information. It changes what everyone in the room can see — and, often enough, the best ones do not merely uncover good answers.

They create them.

• #thinking
• #engineering
• #problem-solving

@online{byatt2026-thequestionsweask,
  author  = {Stephen Byatt},
  title   = {The Questions We Ask},
  year    = {2026},
  date    = {2026-01-29},
  url     = {https://learningoutloud.stephenbyatt.com/blog/the-questions-we-ask/},
  urldate = {2026-01-29},
  note    = {Learning Out Loud}
}