Help With Stuck — Mentoring Engineers, Part 4
Mentoring Engineers · Part 4
Two junior engineers on the same team get stuck in the same week. The first messages their lead within ten minutes of every obstacle. The lead, who is kind and busy, answers each one with the fix. After a month the first junior is fast on easy work and helpless on anything new. The second junior doesn't ask. They spend three days on a failing integration test that turns out to depend on a timezone setting in the CI runner, and only mention it at stand-up when asked why the ticket hasn't moved. Their lead finds out about the three days afterwards.
Both leads believe in "letting people struggle a bit". Neither had an agreed way to tell useful struggle from waste, so each junior settled into the habit the environment rewarded: ask instantly, or never ask.
The working claim for this part: stuck is valuable while the junior is still learning something about the problem, and wasteful once they're repeating the same moves. Your job isn't to decide in the moment whether to help. It's to set up the time boxes, the signals and the hints so that the line is visible to both of you.
Before you start
Ask your junior engineers, privately: "When you get stuck, how long do you usually wait before asking, and why?" The answers will vary more than you expect.
Think of the last time you gave someone the answer to a problem. Could you have given a hint instead? Keep it in mind for Step 3.
Check whether your team has a written norm for when to ask for help. Most teams have an unwritten one, and it's usually "don't".
Step 1: Separate productive stuck from unproductive stuck (10 min)
Research on learning supports some struggle. Kapur's 2008 study of "productive failure" found that students who wrestled with complex problems before being taught the method did better on later tests than students taught first, even though they performed worse while learning. The struggle built a better understanding of the problem's structure.
Two conditions mattered, and they translate to engineering. The problem has to be within reach, and the struggle has to be followed by instruction that makes sense of it. Struggle with no debrief is just lost time.
Productive stuck | Unproductive stuck |
Forming and testing new hypotheses | Repeating the same attempt with small variations |
Learning about the system with each try | Blocked on missing access, knowledge or context |
Can say what they've ruled out | Can't say what they've tried, or keeps trying the same thing |
Frustrated but curious | Anxious, or hiding it |
Test you can run: next time a junior says they're stuck, ask "what have you ruled out?" A list means productive stuck. "I'm not sure" or the same item three times means unproductive, and it's time to help.
Step 2: Agree a time box, and make asking the expected move (10 min)
Set a default together: for example, 45 minutes of genuine trying, then ask, with "I've been stuck for 45 minutes" treated as a normal report, not a confession. Shorten it for blocked-by-access problems (there's nothing to learn from waiting for a permission). Lengthen it for problems that are the point of the task.
The time box has to be modelled, not just announced. When you're stuck, say so in the team channel, with how long you've been at it.
Test you can run: at your next stand-up, ask who has been stuck on something for longer than the agreed time box this week. If the answer is "nobody" every week, either your juniors are unusually lucky or the norm isn't working.
Step 3: Use a hint ladder before giving the answer (15 min)
Wood, Bruner and Ross coined "scaffolding" in 1976, studying how a tutor helped young children solve a problem "beyond [their] unassisted efforts". The tutor gave far more direct, hands-on help to the youngest, least able children and progressively less as competence grew, leaving more of the task to the learner. The engineering version is a ladder you climb one rung at a time, going only as far down as this person needs today:
Question: "What does the error actually say?" or "What changed since it last worked?"
Direction: "I'd look at how the test environment differs from your machine."
Specific pointer: "Check the timezone in the CI config."
The answer, with the reasoning.
Start at the top. Move down only when the current rung doesn't unblock them in a reasonable time.
Test you can run: for the next five times you help a junior, note which rung you started on. If it's usually rung 4, you're solving their problems, not teaching them to solve problems.
Step 4: Know when to give the answer straight away (10 min)
The hint ladder isn't always right. Go straight to the answer when the problem is outside what the junior is meant to be learning (an obscure build-tool flag, an undocumented internal API), when production is affected, or when they've clearly reached the end of what they can work out alone. A hint about something they have no way to reason about is just a riddle.
Test you can run: before giving a hint, ask yourself: "Could they work this out from what they currently know or can find?" If no, give the answer and the reason, and move on.
Step 5: Teach the debugging process, not just the bug (15 min)
Juniors are often stuck because they don't have a method, not because the bug is hard. Agans's Debugging: The 9 Indispensable Rules (2002) is still a good checklist to borrow from: understand the system, make it fail reliably, quit thinking and look, divide and conquer, change one thing at a time, keep an audit trail, check the plug, get a fresh view, and if you didn't fix it, it isn't fixed. When you help, name which rule you're applying: "Let's make it fail on purpose first."
Test you can run: ask a junior to talk you through how they'd approach a bug they haven't seen before. If the answer is "I'd look at the code and see what's wrong", they need a method before they need more hints.
Step 6: Recap after every unstick (5 min)
When the problem's solved, spend two minutes on what unblocked it and what they'd try first next time. This is the "instruction that makes sense of it" from Kapur's work. Without it, the struggle teaches the fix but not the pattern.
Test you can run: a week later, ask the junior what fixed the last thing they were stuck on and how they'd spot it sooner. If they only remember the fix, the recap didn't happen or didn't stick.
Step 7: Make stuck visible without making it a performance issue (10 min)
In an office, a lead can often see that someone is stuck: the same screen for an hour, the sigh, the third coffee. On a distributed team, stuck is invisible until stand-up, and stand-up is a poor place to admit it, because it happens in front of the whole team. Give stuck a lower-cost channel. Some teams use a "stuck" label on the ticket that anyone can set and that a senior checks twice a day. Others keep a dedicated help channel where the norm is to post the problem, what's been ruled out and how long it's been.
The important part is what happens next. If a stuck flag leads to a conversation about the junior's pace, nobody will set it twice. If it leads to help within an hour and a short recap, it becomes a normal part of the work.
Test you can run: check how many times the stuck signal was used last month, and by whom. If only seniors use it, juniors don't yet believe it's safe.
Worked example: the timezone test (composite scenario)
At a retail software company, a junior engineer spent three days on an integration test that passed locally and failed in CI. Nobody knew until stand-up. Their lead, rather than fixing it, sat down with them and worked through the ladder.
Rung 1: "What's different between your machine and CI?" The junior listed the operating system and the database version, which they'd already ruled out. Rung 2: "What about environment settings, not software versions?" Silence, then: "Locale? Timezone?" The test compared a date generated "today" against a fixture written in UTC; the CI runner was in UTC and the junior's laptop wasn't. It took eight minutes.
The recap took five. The lead asked what they'd check first next time, and the junior wrote "env differences, not just versions" in their notes. Then the lead asked the harder question: why three days without asking? The junior said they thought asking would look bad in their first quarter. The lead raised it with the team, and the team agreed a 45-minute norm, with the lead posting their own "I'm stuck" messages to make it real. Over the next month, the junior's longest silent stretch dropped from days to under an hour, and their questions got sharper, because they arrived with a list of what they'd ruled out.
Trade-offs to make deliberately
Choice | You gain | You risk |
Short time box | Little wasted time | Less struggle, so less built understanding |
Long time box | Deeper learning on the right problems | Days lost on problems with nothing to learn |
Hint ladder | Builds problem-solving skill | Slower; frustrating when used on the wrong kind of problem |
Direct answers | Fast; good for problems outside their remit | Dependency, and the same question next week |
Common failure modes
Instant rescue. The junior learns to wait for you.
Refusing to help "on principle". Struggle with nothing to learn from is just waste.
No agreed time box. Each junior invents their own, usually the wrong one.
Hints on unknowable problems. A riddle about an undocumented flag teaches nothing.
No recap. They learn the fix and miss the pattern.
Final takeaway
Stuck is valuable while the junior is still ruling things out, and wasteful once they're repeating themselves or blocked on something they can't know. An agreed time box, a hint ladder, a debugging method and a short recap make that line visible to both of you. Your next action: agree a time box with your team this week, and the next time someone asks for help, start on the top rung of the ladder.
Sources
Kapur, M. (2008). Productive Failure. Cognition and Instruction, 26(3), 379–424.
Wood, D., Bruner, J. S. and Ross, G. (1976). The Role of Tutoring in Problem Solving. Journal of Child Psychology and Psychiatry, 17(2), 89–100.
Agans, D. J. (2002). Debugging: The 9 Indispensable Rules for Finding Even the Most Elusive Software and Hardware Problems. AMACOM.
Continue the Mentoring Engineers path
Previous — Part 3: Give Feedback Juniors Can Use
Part of the Mentoring Engineers learning path. Related: Debug a Failing Test · Debugging in Production Without Losing Your Mind.


