Most candidates open a nonprofit case with a profitability tree and then apologize that "the client is mission-driven." That is the wrong swap. Profit is not soft here; it is replaced. The unit of progress is cost per outcome under funding that is often not fungible. If you treat restricted cash like a surplus, I stop listening within the first two minutes.
After this page you should be able to open a nonprofit case the way a partner would: split dollars by donor restriction, define the outcome before the output, compare programs on cost per incremental outcome, and treat the overhead ratio as a political constraint rather than the objective.
What the interviewer is actually scoring
Nonprofit and social-impact rounds still follow the same choreography as any other case: clarify the question, structure, run the math, recommend. The scoring sheet changes on three lines.
| Scoring line | What a good answer does | What a weak answer does |
|---|---|---|
| Objective | Names a measurable outcome and a time horizon | Says "maximize impact" without a unit |
| Money | Splits restricted vs unrestricted before any allocation | Adds all cash into one pool |
| Decision metric | Cost per incremental outcome, then risks | Overhead ratio, meals served, or "brand" |
I am not looking for charity-sector jargon. I am looking for whether you notice that a dollar with a purpose restriction is not the same asset as a dollar the board can move. That distinction is not folklore. U.S. not-for-profits that follow FASB ASC 958 report net assets with donor restrictions and net assets without donor restrictions on the face of the financial statements, and the IRS Form 990 mirrors that split on Part X, lines 27 and 28.
First split: dollars that move vs dollars that do not
Open every nonprofit case by asking two money questions before you touch a program tree.
- Which dollars are with donor restrictions (purpose, time, or both)?
- Which dollars are without donor restrictions — including board-designated funds that look "reserved" but are still legally movable?
Board designations are internal. Donor restrictions are external. Confusing the two is how candidates invent a fake surplus. A $3m meals grant sitting in cash is not a $3m discretionary pool if the grant agreement forbids rent, measurement, or another program. On Form 990 that cash still appears as an asset; the constraint lives in the net-asset classification and the grant terms.
Say this out loud early: "I will not allocate restricted dollars to a use the donor did not fund. Unrestricted dollars are the only true discretionary capital in this case."
If the interviewer hands you a single revenue line, ask for the restricted share before you recommend anything. That question alone separates candidates who have seen a 990 from candidates who are improvising.
Cost per output is not cost per outcome
The second trap is counting the wrong unit. Meals served, nights of shelter, and nets distributed are outputs. Grade-level reading, relapse-free months, and deaths averted are outcomes. Serious evaluators price the second.
GiveWell publishes cost-per-outcome estimates for its top interventions. As of their 2022–2024 grant average (rounded), insecticide-treated malaria nets cost about $5,500 per life saved and about $6 per net delivered. Those two numbers are not interchangeable: the first is the decision metric; the second is an operational input. GiveWell also publishes ranges by geography — nets have been estimated in a roughly $3,000–$8,000 band per death averted depending on location — which is a useful reminder that cost per outcome moves with baseline risk, not with accounting elegance.
J-PAL makes the same point from the evaluation side: comparative cost-effectiveness analysis needs a shared outcome (for example, additional years of schooling) and an explicit stance on which costs count. Their education work shows that expensive programs that reliably raise attendance — including many cash-transfer designs — can still lose on cost-effectiveness to cheaper approaches such as addressing health barriers or reducing distance to school. In interview language: a program can "work" and still be the wrong place for the next unrestricted dollar.
When an exhibit shows "cost per beneficiary," ask whether the beneficiary is an output or an outcome, and whether the count is incremental (would not have happened otherwise). Attribution belongs here the same way it belongs in a marketing case.
The overhead ratio is usually a red herring
Boards and donors love administrative-cost ratios. Interviewers love watching you genuflect to them and then refuse them as the decision rule.
A low overhead ratio can mean starved measurement, unpaid finance staff, and outcome numbers nobody should trust. A higher ratio can be the price of M&E, grant compliance, and the finance system that keeps restricted funds from being co-mingled illegally. In a case, use the ratio as a constraint ("the board will not approve anything that pushes admin above 20%") and then optimize cost per outcome inside that constraint. Do not maximize the ratio itself.
If a restricted grant forbids admin recovery, that is a funding-design problem, not proof that admin is waste. A clean recommendation is often: renegotiate an allowable indirect rate, or stop accepting grants that strand cash while the lights go unpaid.
Worked example: HarborLine cannot spend the "leftover"
Prompt. HarborLine is a U.S. youth nonprofit with two active programs: NightStay (emergency shelter beds) and SkillTrack (job-readiness coaching that places 18–24-year-olds into jobs lasting 90+ days). A foundation just wired a $2.4m restricted grant for NightStay bed-nights only. HarborLine's fully loaded cost is $48 per bed-night. Operations can deliver at most 40,000 bed-nights this year at current staffing. Separately, the board has $900k unrestricted. Two proposals compete for it:
- Proposal A — new county expansion of SkillTrack: claims 1,200 additional 90-day job placements.
- Proposal B — deepen SkillTrack in the current city: claims 2,000 additional 90-day placements with the same curriculum.
The board chair wants any plan that "keeps overhead under 15%." NightStay's grant agreement disallows rent, shared finance staff, and SkillTrack costs. What do you recommend?
Step 1 — Do not invent a surplus. Max NightStay spend at capacity: 40,000 × $48 = $1.92m. Cash remaining in the restricted grant: 2.4m − 1.92m = $480k. That $480k is stranded for HarborLine's other needs. It is not profit, not a float for SkillTrack, and not a reason to expand geography. Either raise NightStay capacity toward the grant's implied ceiling (2.4m / 48 = 50,000 bed-nights) if operations and facilities allow, or go back to the donor for an indirect-cost allowance and a measurement line. Presenting $480k as "available capital" fails the case.
Step 2 — Rank unrestricted dollars on cost per outcome.
| Use of $900k unrestricted | Claimed incremental outcomes | Cost per placement |
|---|---|---|
| A — new county | 1,200 placements | $900,000 / 1,200 = $750 |
| B — deepen current city | 2,000 placements | $900,000 / 2,000 = $450 |
Even before incrementality, B is cheaper per placement. Then discount A's claim: a new county often counts youth who would have been served by another provider. If even 25% of A's placements are non-incremental, true cost per incremental placement rises to 900k / 900 = $1,000. B still wins on the math the interviewer can grade in sixty seconds.
Step 3 — Handle overhead as a constraint, not a goal. If deepening SkillTrack requires $120k of shared M&E and finance (pushing reported admin from 14% to 17%), say so. Recommend approving B with a board note that a 17% ratio buys trustworthy placement verification; cutting M&E to defend 15% would make the 2,000 figure theater. Offer a 12-month unrestricted fundraising target so NightStay grant concentration is not a going-concern risk.
Recommendation. Put the $900k unrestricted into B. Treat the $480k NightStay remainder as non-deployable until the donor amends the agreement or capacity rises to ~50k bed-nights. Next exhibit I would ask for: true incremental placement rates in the current city, and the share of NightStay fixed costs the restricted grant is actually allowed to cover.
Ned's rule for nonprofit cases
Ned's rule. Restricted cash that you cannot legally spend on the decision in front of you is not a resource. Count it as zero for allocation purposes, then either renegotiate the restriction or leave it stranded on purpose. Candidates who "optimize the whole balance sheet" are optimizing a fiction.
What a profitability tree misses (and what to build instead)
A standard revenue–cost tree assumes fungible dollars and a profit objective. Rebuild the opening structure like this:
- Outcome definition — unit, population, time horizon, incremental vs gross.
- Funding stack — with donor restrictions vs without; any matching or spend-down rules.
- Program options — cost per incremental outcome; capacity and utilization.
- Constraints — overhead ceiling, reserve policy, donor concentration, mission boundaries you cannot exit.
- Recommendation — where unrestricted dollars go, what to do with stranded restricted cash, what to measure next.
Mission is a boundary condition ("we do not abandon the last-mile population the charter names"). It is not a license to skip arithmetic. Inside the mission, programs still compete on cost per outcome the way product lines compete on contribution — see operations and cost cases for the fixed/variable discipline, then change the objective.
Hospitals with a charity-care mission are usually still healthcare economics plus payer mix. Pure public-agency KPI cases are a different species. This article is NGOs, foundations, and mission-driven operators where Form 990 logic and grant terms dominate.
Practice this today
Take any case you already know and force a nonprofit rewrite for fifteen minutes. Invent a 60/40 restricted–unrestricted split, pick one outcome unit, and refuse to spend restricted dollars off-purpose. Then run the cost-per-outcome comparison on unrestricted dollars only.
On CoachNed, start with a timed math rep on /drills/math or the free forever /drills/quick-math, then put the same discipline into a full case at /start (three typed turns, instant scores, no account). Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month. CoachNed is independent and unaffiliated with McKinsey, BCG, Bain, Deloitte, or any firm named in interview-style practice.
Frequently asked questions
Are nonprofit case interviews different from regular case interviews?
The process is the same: structure, math, recommendation. The objective and the funding constraints change. Replace profit with cost per incremental outcome, and split restricted from unrestricted dollars before you allocate.
How do I handle restricted grants in a case?
Treat purpose-restricted funds as spendable only inside the donor's rules. Leftover restricted cash is not a surplus. Either scale the funded activity, renegotiate allowable costs, or leave the cash stranded and say why.
What is a good cost-per-outcome benchmark?
There is no universal number. Use the case exhibits first. Outside the room, published evaluators such as GiveWell quote intervention-specific figures (for example, roughly $5,500 per life saved for malaria nets on their recent top-charity average). In a domestic youth or education case, build the comparison between the options on the page rather than importing a global health benchmark.
Should I optimize the overhead ratio?
No. Treat it as a board or donor constraint if the prompt says so. Optimize cost per outcome inside that constraint. Cutting measurement to decorate the ratio is usually the wrong move.
Do I need nonprofit experience to pass these cases?
No. You need clean logic on fungibility and outcomes. Knowing that Form 990 and ASC 958 separate net assets with and without donor restrictions is enough institutional literacy to sound adult in the room.
Sources
- Instructions for Form 990 (IRS) — Part X lines 27–28: net assets without / with donor restrictions for ASC 958 filers. Checked 2026-09-24.
- GiveWell: How We Produce Impact Estimates — cost per life saved and cost per output for top charities (2022–2024 averages). Checked 2026-09-24.
- GiveWell: Mass Distribution of Insecticide-Treated Nets — geographic range roughly $3,000–$8,000 per death averted. Checked 2026-09-24.
- J-PAL: Conducting Cost-Effectiveness Analysis — CEA method, shared outcomes, and which costs to include. Checked 2026-09-24.
- J-PAL: Reducing Costs to Increase School Participation — cash transfers raise attendance but often lose on cost-effectiveness to cheaper interventions. Checked 2026-09-24.