We'd refined the ticket properly. Product wanted the feature, we'd talked it through as a team, everyone had nodded in the right places, and the whole thing seemed entirely plausible for about as long as it takes to start building it.
Then it didn't work. Not in a "this is fiddly" way. In a "the thing we all agreed to is not going to do what product thinks it does" way.
So I went to my team lead, who was approximately one hour from a fortnight of annual leave and had the specific energy of a man whose laptop lid was already at forty-five degrees. I explained what I'd found. He listened, and then he said three things, and I've only really appreciated the shape of them years later.
The first was: good, I'm glad you came to me now rather than next week.
The second was the one that stuck, and it's the reason this post exists: go to the TPM with a solution, not a problem.
And the third, which I nearly didn't register at the time, was: spend an hour on it, no more, and then take it to them either way.
The Poster Version Is Missing Two Thirds of It
You'll have seen the middle sentence on its own. It's a management poster. Don't bring me problems, bring me solutions. Divorced from the other two, it's one of the more quietly destructive things you can tell someone.
Think about what the isolated version incentivises. You find something alarming. You don't have a solution yet. The rule says you're not allowed to escalate without one, so you go away and work on it. A day passes. Then two. On day four you either fix it, in which case nobody ever knows, or you come back and admit you can't, which is now much worse news than it was on Monday, because the intervening days were the cheap ones. Projects have gone over cliffs at a very stately pace with everyone involved politely holding their bad news until it was sufficiently well-dressed to present.
What my team lead actually said had guard rails on both ends. Coming early is correct. And the hour is the budget, not the goal. Take it to them either way, solved or not.
Strip those out and you get the poster, which teaches people to sit on things. Keep them, and you get a genuinely useful instruction that I've been following ever since.
What the Hour Actually Buys You
Here's what I had before the hour, and what I'd have sent if he hadn't stopped me:
hey, think there's a problem with the new feature
Look at what that gives the person receiving it. Not information. An alarm with no dial on it. They don't know how bad, what it touches, or what their options are, and the only thing they can do with it is ask me a series of questions to extract things I already knew. It hasn't moved forward. It's moved sideways, onto someone with less context than me, with a bit of dread attached for the journey.
Here's what an hour turns it into:
The way we refined this doesn't hold up, and it's the [X] behaviour that breaks it rather than the feature as a whole. Two options I can see: we ship the simpler version without [X], which works and is maybe two days, or we do it properly and it's a couple of sprints because of [Y]. I'd ship the simpler one and revisit. The thing I haven't checked is whether product cares more about [X] than I think they do.
Same problem. Same hour I'd have spent anyway answering their first three follow-up questions. But their job has completely changed: they now have to decide something, instead of find out what's going on. Deciding is their job. Finding out was mine.
And that last sentence, the one naming what I hadn't checked, is the bit people leave off and shouldn't. Being precise about where your knowledge runs out isn't a weakness in the take. It's what makes it usable, because it tells them exactly which part to go and verify, and it means when you're wrong about it nobody feels misled.
You don't have to be right. A take that gets demolished in ninety seconds is still worth far more than a bare problem, because reacting to something specific is enormously faster than generating something from nothing. Being wrong in a correctable direction is a service.
The Hour Isn't Always an Hour
The timebox is the part that makes this safe, and it's worth saying plainly that for some things the budget is zero.
Anything where the delay compounds. Data being corrupted right now. A credential in a public repo. Every minute you spend forming a view is another minute of damage, and your view is worth less than the minute.
Anything with a legal, security or privacy dimension. Not yours to triage, and not yours to decide how serious it is. Tell the person whose job that is, immediately, and let them be the one who concludes it was nothing.
Anything affecting a promise somebody is making today. If your TPM is in a room this afternoon committing to a date you now believe is wrong, they need to know before the room, not after you've built an elegant alternative. Their ability to soften a commitment in the moment is worth more than your solution.
The actual skill, the one that took me a lot longer than the original conversation, is telling which bucket you're in. The test is roughly: does the delay cost more than the analysis? If it does, send it now, half-formed, with "I've only just found this and haven't thought it through yet" on the front. That sentence is free, and it sets expectations perfectly.
A Take Isn't a Solo Effort
One more misreading worth heading off, because "bring a solution" can sound like "handle it alone".
"Here's what I think is happening, here's what I'd do, can you sanity-check the second half because I'm out of my depth on the infrastructure" is a take. A completely legitimate one. You've done your share of the thinking and you're being precise about where it ran out, which is the same move as above, just pointed at a colleague rather than a manager.
What isn't a take is the raw problem with an implied question mark after it. That's the only thing worth training yourself out of.
The Taste Test
Before you send the bad news:
If they read this now, what do they do next? If the honest answer is "work out what you meant", it needs another twenty minutes.
Does delaying cost more than the analysis? If yes, send it immediately and say it's half-formed. This overrides everything else here.
Have I given a recommendation, or just options? Two choices with no lean is still homework. Say which you'd pick, even weakly.
Have I said where my knowledge stops? The gap you name is the gap someone else can cover.
Am I still working on this because it needs work, or because I don't want to send it? These feel identical from the inside and are not remotely the same thing.
Wrapping Up
The thing that changes when you start doing this isn't gratitude. It's that people begin handing you vaguer work, because someone who reliably comes back with a view can be trusted with a mess. Being handed messes is more or less the definition of the next level up.
But I'd keep all three sentences, not just the memorable one. Come early. Spend an hour, not a week. Then go anyway, solved or not. The poster only ever prints the middle one, which is how a decent piece of advice ends up teaching people to go quiet at exactly the wrong moment.
My team lead went on holiday. I took it to the TPM that afternoon with two options and a recommendation, and the conversation took about four minutes.