Capstones end the same way everywhere. There's a final presentation, the sponsor says something gracious, the team goes out, and within three weeks the shared drive is a graveyard nobody has access to anymore. Six months later you tell an interviewer "I did a strategy project for a manufacturer," and that sentence is the entire remaining artifact.
The fix is small, and it has to happen fast. Before the files scatter, produce one thing you can actually hand someone: a two-page case summary, a cleaned model, or a written case study your sponsor has agreed in writing that you can share. That's what a portfolio piece means for an MBA. Not a designer's portfolio, not a personal site with project tiles—one defensible artifact that outlives the grade.
Settle confidentiality before you write a word
Do this first, because it's what kills most attempts—usually after the work is done and the person has to quietly take it down.
Most capstone and sponsored experiential projects sit under an agreement: an NDA you signed, a course agreement between the school and the company, or both. Read yours. It's three pages and you skimmed it in week one. The common default is that the deliverable belongs to the sponsor and the school, and that you have no independent right to publish any of it.
What you can nearly always share:
- The method. How you framed the question, what analysis you chose, why that approach over the obvious alternative.
- The anonymized structure. "A mid-market industrial distributor, roughly $400M in revenue, eleven distribution centers" carries the shape of the problem without identifying anyone.
- Your specific contribution. What you personally owned, built, and decided.
- Work you sourced publicly. A market sizing you assembled from published filings is yours—it didn't come from them.
What you cannot share without explicit permission:
- The client's name, or details that identify them to anyone in the industry. A "regional grocery chain in the Pacific Northwest with 41 stores" is not anonymous.
- Real financials, customer lists, pricing, contract terms.
- Anything they handed you—their data, their internal decks, your interview transcripts with their staff.
- The final deliverable itself, in the form you gave it to them.
The co-signed version—a case study with the sponsor's name on it—is genuinely valuable, and it requires genuine permission. Ask by email, attach exactly what you want to publish, and get a written yes back. "My sponsor said it was fine" in the hallway after the final presentation is not permission. The person who said it may not have authority to grant it, will not remember saying it, and is not the legal reviewer who eventually sees your post. Copy your faculty advisor—it costs nothing and makes the ask read as institutional rather than a student going freelance with a client's material.
If the answer is no, build the anonymized version. It's worth most of what the co-signed one is, and carries none of the risk.
The 48 hours after the final presentation
This is the whole timing argument. In the two days after you present, everyone still has the files, the model still opens without a broken-link warning, and you can still explain why the churn assumption was 6% and not 8%.
By February you will remember none of it. The teammate who built the sensitivity tab has graduated and stopped answering email. The folder has been archived by someone in IT. Reconstructing a model whose logic you half-remember is worse than building a new one.
So put the block on the calendar before the presentation, not after—90 minutes, two days out, when you'll be tired and would otherwise skip it. Recruiting timelines will fight you for that slot, which is why it belongs on the calendar early rather than in the pile of things you meant to do between interviews. In the block:
- Export a personal copy of whatever you're permitted to keep—check the agreement first; this is the precise moment people breach it casually.
- Write down the three numbers that mattered and where each one came from.
- Write down two decisions you made under real uncertainty and what you knew at the time.
- Write the honest paragraph: what the team recommended, and what the sponsor said they'd do with it.
Then set a reminder for a month out and ask the sponsor what happened. "They implemented the pricing change in Q3" is the strongest closing line an interview answer can have, and almost nobody has it, because almost nobody asks.
Say exactly what you did
Attribution is where a portfolio piece flips from asset to liability.
A capstone deliverable is team output. Presenting it as yours is the fastest way to fail a reference check, and MBA reference checks are unusually easy to run—your teammates are on the same alumni list as the person checking, and your faculty advisor replies to email.
The fix isn't modesty. It's precision, in one sentence, up front:
Five-person capstone team for a regional health system. I owned the demand model and the pricing recommendation; a teammate owned the operations analysis.
Everything after that sentence is now unambiguously yours, and the specificity reads as more credible rather than smaller. "I led a strategic transformation" is a bigger claim that will be trusted less than "I built the model."
If you were the person who actually held it together, say that plainly—"I was the integrator; I owned the final narrative and the merge." That's a real and rare skill. Just don't quietly upgrade it into having done the analysis too.
Formats, and who actually reads them
- The two-page case summary. Problem, approach, your role, recommendation, outcome. One PDF. This is the one that gets read, because it's the only one that fits in the gap between a recruiter's calls. Write it first and write it even if you write nothing else.
- The cleaned model. Valuable for finance and analytics roles. An interviewer who opens a well-documented model learns more in ninety seconds than an hour of behavioral questions will tell them. "Cleaned" means assumptions isolated on one tab and sourced, no hardcodes buried inside formulas, and no broken links on open. An uncleaned model is worse than sending nothing.
- The public write-up. A post about the method rather than the client. Lower direct recruiting value than people expect; decent value as thinking practice.
- The sponsor-co-signed case study. Rare, high value, worth asking for. It's the only format where someone other than you vouches for the result.
- The deck. Nobody wants your 40 slides. Don't send it. Keep it as the quarry you pull one exhibit from—a single well-built chart in the two-pager does more than three paragraphs describing the analysis.
Some capstones produce nothing worth showing
A fair number don't. Pretending otherwise costs you a weekend and produces something that hurts you in the room.
Signs yours is one of them:
- The recommendation was generic. "Invest in digital capabilities" is not a portfolio piece no matter how many hours went into it.
- You can't name a decision you personally made. If your contribution was slide production and formatting, the honest artifact is a story, not a document.
- The sponsor never engaged. A project that was effectively a simulation with a real company's name attached has no outcome to point at.
- It wouldn't survive a smart question. If you can't defend the churn assumption to a curious interviewer, don't put it in front of one.
If that's your project, stop and take the other route: capture the moments instead, which are better interview material than most students realize. A thin artifact with a strong story behind it is fine. A polished artifact you can't defend is not—an interviewer who finds one soft number will spend the rest of the conversation looking for the next one.
Keep it somewhere you'll find it
Put the PDF, the model, and the notes in one folder outside school infrastructure, because your student account gets deactivated and nobody warns you twice. Name the file so a stranger understands it: capstone-summary-pricing-healthcare.pdf, not final_v7_REAL.pdf.
Then do the thing that makes it count: attach the two-pager to one application this month. An artifact you never send is the same as no artifact, and the first time you use it you'll see the three sentences that need rewriting.