Your team has converged. You think the recommendation is wrong. Before you plan how to say so, do the part almost nobody does: find out whether you're right.
That's not a hedge. Being right feels identical to being wrong from the inside, and MBA cohorts are full of people who are confidently, articulately wrong—a room that rewards conviction produces more certainty than the evidence supports, and you're eligible for that. Clear the tests below and the rest is mechanics: raise it early, attack the assumption not the person, bring a replacement, commit fully if you lose. One situation exempts you from that last part, and it's at the end.
Start by trying to talk yourself out of it
Three tests. They take longer to read than to run.
Can you state the team's position better than they can? Not a version you can defeat—the version its strongest advocate would endorse word for word. Ask whoever proposed it to walk you through the reasoning, and take notes. Fairly often the disagreement dissolves right here: the leap you thought was unexamined had a reason that never reached the meeting. If it doesn't, you now know exactly which step you're objecting to, and that's most of the work.
What specific evidence would change your mind? Write it down before you argue. "If both comparables ran under 12% churn, I'd drop this" is falsifiable; a team can settle it in an afternoon. "I just don't think it'll work" is a preference, and a team can only route around it. Can't name the evidence? You aren't holding a position yet.
Are you arguing about facts or about risk tolerance? Teams misdiagnose this constantly and it costs weeks. Fact disagreements close with a source. Risk-tolerance disagreements never close by argument: both sides are right about the numbers and differ only on how bad the downside has to be before it disqualifies the plan. Say so out loud—"we don't disagree about the model, we disagree about how much downside we're willing to recommend to a real company"—and the conversation stops being a hunt for an error that isn't there.
Timing is most of the argument
The cost of changing a recommendation scales with everything built on top of it. On day three it's a conversation. At hour forty, with twenty-two slides built and an exhibit someone stayed up making, you're asking four people to throw away a weekend—and they'll weigh your reasoning against that cost whether they mean to or not. The identical argument wins on Tuesday and loses on Saturday.
So flag it the day you notice, before you've built the case. One sentence does it: "I'm not sold on the entry sequence. Give me until Thursday and I'll bring the specific reason." A placeholder costs nothing and prevents the worst version of this: an objection that arrives late and lands as an ambush on people who thought the question was closed.
Past the build point, raise your own bar: reopen only if the flaw is one a professor, judge, or sponsor would catch. Everything under that line goes in the retrospective, not into the last night. Locking the answer at a stated deadline—standard practice under case competition time pressure—makes that boundary explicit instead of leaving it to whoever has the most stamina at 1am.
Say the assumption, not the verdict
"I disagree" gives the team nothing to do. "Here's the assumption I think breaks" gives them something specific to check.
Name the load-bearing input, say what happens if it's wrong, and propose the smallest change that fixes it. Smallest is doing real work there. Most wrong recommendations aren't wrong end to end—usually one number, one segment, one sequencing choice. Framing a sequencing objection as "we should do something else entirely" makes the disagreement bigger than it is and much harder to win: you're asking for a reversal instead of a revision.
Two mechanics decide how it lands:
- Attach it to the work, never to a person. "The model runs 12% churn and both comparables came in above 20%" is a question the team can answer. "Dan's model is optimistic" is a status contest, and Dan will defend the number instead of examining it.
- Ask for a bounded test, not a reversal. "Give me two hours to rerun it at 20%—if the recommendation holds, I drop it." Cheap to grant, hard to refuse, and it commits you publicly to dropping it, which is what makes people willing to look.
And never recruit. "A few of us have concerns" turns an analytical disagreement into a faction fight, and that's what gets remembered.
When you lose, actually lose
Sometimes you do all of it and the team goes the other way anyway. Disagree-and-commit means three specific things.
You stop relitigating. You build your piece at full quality—the most common sabotage on student teams is unconscious, and it looks like slightly worse work on the recommendation you didn't want. And you don't leak the dissent sideways: no sighing in the work session, no "well, if we're doing it this way."
People break this in front of an audience. "For what it's worth, I argued against this" is the most damaging sentence available to you in a sponsor or judge Q&A. It doesn't read as intellectual honesty; it reads as someone distancing themselves from their team under pressure, which is exactly the trait that room screens for. Teammates remember it long after the grade posts—including on peer evaluations, where the mirror-image failure shows up as a concern nobody ever actually raised.
If you want a record, make it at the time and make it neutral: one line in the decision log. "Considered pricing 8% higher; rejected on competitive-response risk." Nobody's on trial and the team can find the reasoning in week nine. Keep a private note of what you predicted, too. If you were right you'll never need to invoke it—credibility accrues on its own next time. If you were wrong, which happens more often than the story you tell yourself allows, you'll notice.
The exception: when it's dishonest, not just weak
Everything above is for disagreements about judgment. There's a second category, and it doesn't get disagree-and-commit.
A number nobody can source. A market size someone estimated in their head that the slide presents as research. A quote attributed to a sponsor interview that didn't happen that way. A recommendation a real client would act on and be harmed by. One test: if the sponsor asked "where did this come from," would the honest answer embarrass the team? That's an accuracy problem, not a direction fight, and your name goes on the deliverable regardless.
Escalate in order, starting lower than instinct suggests.
- Ask a factual question, publicly, with no accusation in it. "What's the source on the 14%? I want it in the appendix." This ends most of these: what looks like fabrication is usually a placeholder someone forgot to replace.
- If there's no source, state the consequence and a date. "We can't present this unsourced. If nothing turns up by Thursday, I want to cut it or relabel it as our estimate."
- Only then go to the professor—and go to ask, not to report. "We have a figure I can't source and I'm not comfortable presenting it. How do you want us to handle estimates in the deck?" It's a real question with a real answer, and it protects the team instead of indicting anyone. Framing it as an honor-code accusation starts a process nobody can stop. The general rule on escalation holds: faculty are for misconduct, not for judgment calls.
With a real sponsor the stakes move again: a bad number costs a relationship the school keeps long after you graduate. Handle it like any sponsor-facing problem—early, in writing, before it reaches them.
What you're actually building
Nobody will remember, three years out, whether you were right about the churn number. They'll remember whether arguing with you was worth it.
Dissent is a budget. Two or three well-built objections a term land; nine get you tuned out by week four, at which point being right stops mattering because nobody's listening. Spend it where the reasoning genuinely breaks, raise it early, bring the fix, build at full effort when you lose. The payoff isn't winning arguments—it's getting asked "what do you think?" before the decision instead of after.