I was on holiday in a remote bit of Portugal earlier this month, mostly unplugged and doing the thing I always do on holiday – reading random stuff that I’ve been saving up because I never have time to get to it normally.

Deep in that list was a psychology paper about how researchers measure problem solving in children. It’s not obvious holiday reading, I’ll admit, but bear with me.

The paper picked apart the famous marshmallow test: the one where a child is left alone in a room with a marshmallow and told they’ll get a second one if they don’t eat the first before the researcher comes back.

For decades this has been treated as a clean measure of self-control, until a group of developmental psychologists ran it with children from Yucatec Maya communities in Mexico and most of them “failed” it.

That happened not because they couldn’t resist a marshmallow, but because they saw no good reason to sit alone in a room doing nothing. In their world, waiting quietly by yourself isn’t a virtue – noticing what needs doing and getting on with it is.

Put another way, the test wasn’t measuring what the researchers thought it was measuring. It had quietly stripped out all the context that made the “problem” mean anything in the first place.

That’s been rattling around in my head ever since, because it’s exactly the mistake I watch a lot of founders I work with make, not with marshmallows but with everything else on their plate.

Some problems really are what they look like. You’ve got the information you need right there in front of you, and a decision will settle it – the invoice didn’t go out, the pricing page is wrong, someone needs to own the Tuesday stand-up. Let’s call these narrow context problems – the kind you can solve from where you’re sitting.

Others aren’t, because they can’t be understood from one seat. The “problem” changes shape completely depending who’s describing it to you.

Ask a founder why a manager won’t delegate and you’ll get one answer but ask the manager and you’ll get a different one entirely. Ask the person he’s supposedly not delegating to and you’ll get a third. None of them will quite agree. Let’s call this a wide context problem, the kind where you need the surrounding history, the relationships and the unwritten rules before a fix means anything.

To be honest, I’ve done this myself plenty of times over the years. People who start businesses are wired for action. That means reaching too readily for the fast, narrow fix – a new process, a clearer job description, a firm chat in a one-to-one – on a problem that was actually wide. And making that mistake means the fix lasts not very long at all. Then the same problem comes back, usually somewhere I wasn’t looking.

Here’s where it gets properly difficult: wide problems are very good at looking like narrow ones. They arrive in your to-do list as a single, nameable issue – things like a manager who won’t delegate, a leadership team that can’t agree on priorities or a process everyone nods along to in the room and has abandoned by Friday.

They feel comfortably solvable with a decision yet they rarely are, because the decision only ever addresses the version of the problem you happened to see.

So how do you tell which one you’re actually looking at before you spend a week’s worth of mental energy on the wrong fix?

A few things I check for now:

• Would five people in the business describe it the same way? If everyone would write roughly the same problem statement, it’s probably narrow. The moment the descriptions start to diverge – “he’s not delegating” against “she never gives me anything that matters” against “I stopped asking because it never landed well” – you’re looking at something wide.
• Has it happened before under a different name? Narrow problems tend to be new. Wide problems have a history and you’ll hear “we tried that already” more often than you’d like.
• Does the fix survive being moved to a different team? A solution that works brilliantly in one part of the business and does nothing in an apparently identical part usually means you solved the narrow version of a wide problem.
• Is there a bunch of things that nobody’s saying out loud? I’m talking about old resentments. Who’s actually trusted, not who has the title? What happened the last time someone tried to change this? Wide problems nearly always have one, sitting just under the version people are willing to vocalise to you as the founder.

None of this necessarily means wide problems need more meetings to get to the bottom of them. If anything they need fewer, better ones – the time spent genuinely trying to understand the field the problem is sitting in, rather than collecting opinions before you announce the fix you’d already decided on. Remember: narrow problems reward speed and wide ones punish it.

So a few things worth sitting with this week:

• Which problem on your plate right now would get you a different answer from every person you asked about it?
• When did you last solve something quickly only to watch it quietly reappear somewhere else a few months later?
• And are you giving your wide problems the narrow problem treatment, mostly because narrow is faster and easier to feel good about by Friday?

This is one of my favourite things to dig into with a leadership team – if it’d help to talk through where you might be misjudging the size of a problem, I’m always up for it.