Nobody announces that they've stopped contributing. What you get is accumulation: a section two days late and thin, a meeting missed with a good apology, a message read and never answered. By week six the team has quietly rerouted the work and nobody has said a word.
Two things, this week. Work out which of four things is actually happening, because the remedy differs completely for each and diagnosis is the step everyone skips. Then have a private conversation about one specific missed commitment and what happens next—not their character, not in front of the group, not the professor.
Before anything, check that the expectation was real
One test before you decide someone missed something: can you point to the moment they took ownership, with a date, in writing? If the answer is "we talked about it three weeks ago," that isn't a missed commitment. That's ambient hope—work mentioned, nodded at, never assigned—and teams read the resulting gap as somebody's personal failing. A startling share of these cases collapse right here.
The fix takes ten minutes: named owners with dates in one visible place, and a definition of done two people would agree on—a one-table responsibility matrix is the cheapest version. Then the next conversation points at a row, not a person.
The four reasons, and why they don't share a fix
Treating every case as the last one on this list is the standard mistake. Three of the four respond well to intervention, and the confrontation you'd run on the fourth burns the other three.
Overloaded. The tell: quality is fine when it lands, it just lands late, and the gaps track a calendar—superdays, work travel, a sick parent. Cutting their volume is half the fix. The other half is changing the shape of their work: give them the pieces with no downstream dependencies—the appendix, the data pull, the formatting sweep—so when they slip, nobody is stuck waiting. Recruiting crunch is forecastable in week one, if someone says so.
Unclear. The tell: they produced something, just not what you expected, or they've been "waiting on" an input nobody knew about. This is a team failure wearing a person's face, and running an accountability conversation on someone genuinely confused spends trust you won't get back. Test first: ask them to describe their piece back to you before you mention the miss. If their version and yours don't match, that's your answer—and it isn't about them.
Out of depth. They can't do the task and are too embarrassed to say so. Most common hidden cause; best response to intervention. The tell is specific: the delay attaches to one kind of work—the model, the regression, the financial section—while everything else they own arrives fine. They hedge with "almost done" for two weeks and go quiet exactly when the hard part starts. That's a silence that lengthens, not an availability that shrinks—which is why it gets misread as overload.
Disengaged. They've decided the grade is fine without them. The tell is the absence of a pattern—not lateness with a cause but a flat baseline of nothing, alongside perfectly normal social responsiveness. Confirm this one; don't suspect it. If they agreed to a named deliverable on a named date and that date passed in silence, you know after the first conversation what you were guessing at before it.
Which is why the first conversation is diagnostic: it's an experiment, and the result names which of the four you're in.
The one-to-one, and what to actually say
One person, privately, within a few days of the second miss. Anchor on a single specific commitment—the most recent—not a list of five. A list reads as a case being built, and people defend against cases.
Ask before you assert, then name the next thing:
"The competitor section was due Tuesday and came in Thursday night, so it missed the dry run. What's going on? … Okay. The sizing piece is due the 14th. Can you send me a rough version on the 11th, so there's room if it slips?"
That mid-point check-in does most of the work, and teams leave it out. A recommitment with only a final date tells you nothing until it's too late. A check-in three days early makes the promise checkable—and gives an out-of-depth teammate a natural moment to admit what they've been avoiding.
What not to say: "the team feels like you're not pulling your weight." It speaks for people who aren't in the room, converts a missed deadline into a verdict, and gives them nothing to do tomorrow. The broader conflict playbook covers the behavior-cost-question spine this sits on; non-delivery adds one rule—leave with a date, and a date before the date.
Write it down the same day
Send a two-line message afterward: what we agreed, by when. Not a threat—a record. Memories diverge fast, and the written version is what makes every later step fair rather than retaliatory. Keep a short log: date committed, date due, what arrived and when. Facts, not adjectives.
One test tells you whether you're documenting or building a case: would you be comfortable if they read it? If not, rewrite until you would. That's the line between a record and ammunition—and what lets a peer evaluation be honest rather than a grievance, because nothing in it is a surprise.
The trap: quietly doing it yourself
Absorbing their work feels generous. It protects the grade, it avoids an awkward hour, and it's almost always wrong.
It guarantees a repeat, because nothing was ever a problem in an observable way. It leaves you no evidence—you can't describe a miss that got covered, since from the outside the deliverable was fine. And it withholds the most useful thing business school can give your teammate: that this pattern has a cost, learned while the cost is a group project and not a job.
One real exception: eighteen hours before submission, you cover. Always. But cover the deliverable, not the fact—teams reflexively merge those. Do the work at midnight, then send the message next morning saying what happened. Absorbing the hours is sometimes necessary. Absorbing the information never is.
Involving the professor without it reading as tattling
Two thresholds: they've stopped responding entirely, or they missed a commitment they explicitly re-made in writing. And go before submission. Afterward you're requesting a grade adjustment—a much harder ask that looks like score-settling. Two weeks out, you're asking for a decision on a live project, which professors do all the time.
Arrive with three things: the written owner assignment, the dates, and what you already tried. Then ask for a ruling, not sympathy. "Do you want us to reallocate his section, or would you rather talk to him first?" is a question a professor answers in ninety seconds. "We're having trouble with a teammate" is a complaint they'd have to turn into a question themselves, and most will just tell you to work it out.
Tell your teammate first, without menace: "If the section isn't in by Wednesday I'm asking Professor Chen how to handle it." The tattling worry dissolves right here: the difference between reporting someone and asking for a decision is a record, a proposal, and no ambush.
If you're the one who isn't contributing
The tell is that you're avoiding the group chat. You know what's in there and you're not opening it.
The most effective move available to you is to flag it before anyone has to ask, with a specific recovery plan attached. Three sentences: what you'll miss, what you'll still deliver and by when, and what you'll take on afterward to square it. Not "sorry, crazy week"—that's a status update wearing a plan's clothes, and it buys four days.
This works better than it should, because a team's read of you is set by the surprise, not the shortfall. A known 40% contributor is a planning input; an unknown one is a threat. Teams forgive the first and remember the second for two years.
And if the real reason is that you can't do the task, say so specifically. It costs almost nothing in week three and the whole term's goodwill in week nine. No MBA team has ever preferred a preserved ego to a finished model. To script the conversation first—in either direction—MBA Lab's Compassionate Accountability Planner was built for this hour.