Non-Profit Cases 2026: Cost Per Beneficiary, Worked

A non-profit's two economies, and only one has a paying customer

Write the two ledgers separately before you branch anything.

The funding economy has a cost per dollar raised. Grants, major gifts, events and direct mail each have their own yield, and the organization can choose among them.

The delivery economy has a cost per beneficiary served. That cost varies by who the beneficiary is, by geography, and by how deep the service goes.

The link between the two is a single ratio: benefit delivered per dollar of funder money. That is the objective function of a non-profit case, and it is the sentence to say out loud in the first thirty seconds.

One structural fact makes these cases unlike commercial ones, and it is checkable. In the United States, the financial statements of an exempt organization are reconstructed from a public return. The Internal Revenue Service states it plainly at irs.gov, checked 30 September 2026: "Data are compiled from Form 990, Return of Organization Exempt from Income Tax, and Form 990-EZ."

So the funder can read your cost structure before the first meeting. A commercial client's competitor cannot. That asymmetry changes what a recommendation has to survive, and it is the reason the overhead argument later on this page has real teeth.

A published non-profit case where the client is a foundation

McKinsey publishes the Diconsa case at mckinsey.com, checked 30 September 2026 and read through a text proxy because the site does not answer us directly. The client is the Gates Foundation, working with the Mexican government and a state-owned bank.

The opening question is published verbatim: "What should the team investigate to determine whether the Diconsa network could and should be leveraged to provide a range of basic financial services to Mexico's rural population?"

Notice the two verbs. Could and should. The first is a feasibility question and the second is a mission question, and the firm asks both in one sentence. A profit equation answers neither.

The assumptions are printed. Collecting benefit payments costs a family 50 pesos a month in travel and food today. If payments were collectible at the local Diconsa store, that cost would fall 30 percent for the 60 percent of the rural population living furthest from a bank branch, and 20 percent for the 40 percent living relatively close. Twenty percent of Mexico's population is rural, half of those receive state benefits, the population is 140 million, and families average four members.

The firm's own hint under that question is worth reading as an instruction about pace: "Take time to organize your thoughts before answering. This will help show your interviewer that you have a logical approach and can think in an organized way, regardless of the accuracy of the outcome."

Working the benefit, peso by peso

The firm asks: "If every family could collect state benefits at their local Diconsa stores, how much in total per year would be saved across all Mexican rural families receiving state benefits?" Here is the chain, and every line below is our arithmetic on the printed assumptions rather than a quoted result, because the case exhibits are images.

1. Rural population. 140 million times 20 percent is 28 million people.
2. Beneficiaries. Half of those receive benefits, so 14 million people.
3. Families. 14 million divided by four members is 3.5 million families.
4. Current spend. 50 pesos a month is 600 pesos a year, so 3.5 million times 600 is 2.1 billion pesos a year across all of them.
5. Far group. 60 percent of 3.5 million is 2.1 million families, saving 30 percent of 50 pesos, so 15 pesos a month and 180 a year. That is 2.1 million times 180, which is 378 million pesos.
6. Near group. 40 percent of 3.5 million is 1.4 million families, saving 20 percent of 50 pesos, so 10 pesos a month and 120 a year. That is 1.4 million times 120, which is 168 million pesos.
7. Total. 378 million plus 168 million is 546 million pesos a year.

Check it twice, because a single chain is a single point of failure. By month: 2.1 million times 15 plus 1.4 million times 10 is 31.5 million plus 14 million, so 45.5 million pesos a month, and times 12 that is 546 million. By share: the weighted saving rate is 0.6 times 30 plus 0.4 times 20, which is 26 percent, and 26 percent of 2.1 billion pesos is 546 million. Three routes, one number.

The third check is the one worth keeping, because it converts the answer into a sentence a client can act on. The program returns a quarter of what these families currently spend just reaching their own money.

One total, two right answers, and one wrong one

Now divide that total, and watch the denominator.

Per family: 546 million divided by 3.5 million families is 156 pesos a year. Per beneficiary: 546 million divided by 14 million people is 39 pesos a year.

Both are correct. They answer different questions, and the case hands you the material for both without saying which it wants. That is the unit trap, and it is set by the firm rather than by carelessness, since the family size is given without being told what it is for.

The wrong answer is available too, and it is a factor of four out. Apply the 50 peso monthly cost to 14 million people instead of 3.5 million families and the total becomes 2.184 billion pesos of annual saving, since 14 million times 156 is 2.184 billion. Every multiplication in that version is correct. The error sits above the arithmetic, in what the multiplicand counts, which is exactly where a re-check never looks.

The defense is to speak the unit beside the number every time: "families, at 600 pesos a year each". Our page on clarifying questions gives the test for when to ask about a unit rather than assume it, with a second worked example on a public denominator.

The cost the non-profit case never gives, and its threshold

The Diconsa page quantifies the benefit and never mentions what the program costs to build. There is no implementation figure, no capital number, no operating budget. A candidate who tries to compute a return on investment is computing with an invented numerator.

What you can do instead is invert the question, which is the move the missing number invites. Ask what the program may cost before it stops being worth doing, and the answer is on the page.

Read the third row. A program that costs more than 39 pesos per beneficiary per year destroys value on this benefit alone, and 39 pesos is a small number, which is the useful finding. Read the last row: at an annual program cost of 100 million pesos the ratio is 546 divided by 100, so 5.46 to one.

Say that out loud as the recommendation and you have answered a question the page did not ask, using only figures the page supplied. This is also the honest way to handle a missing cost: name the threshold rather than filling the gap.

Why the overhead ratio ranks two non-profits backwards

The metric non-profits are judged on publicly is the share of spending that goes to programs. It is available to any funder, because the Form 990 is public. It also ranks organizations wrongly, and two lines of arithmetic show why.

Two charities each take in 1,000,000 dollars of gross receipts. Charity A spends 20 percent of receipts on fundraising and administration and delivers its service at 100 dollars per beneficiary. Charity B spends 40 percent and delivers at 60 dollars per beneficiary. All four figures are ours.

Charity B has twice the overhead ratio and serves 25 percent more people for the same money. On the metric that matters, total cost per beneficiary served, it wins by 25 dollars a head.

Find where they tie, because that is the defensible version of the argument. Charity B matches A's 8,000 beneficiaries when 600,000 divided by its delivery cost equals 8,000, so at 75 dollars per beneficiary. Below 75 dollars, the higher overhead is worth paying. Above it, it is not.

That threshold is the recommendation. A funder who caps overhead at 20 percent has quietly capped delivery efficiency too, and now there is a number attached to the cap. Our list of nonprofit consulting firms covers who does this work for a living, and Bridgespan covers the one employer of the nine that publishes its own pay and process.

The non-profit metric that improves by abandoning the mission

Here is the arithmetic that makes the mission a constraint rather than a preamble, and it is the reason a non-profit answer cannot be a pure optimization.

An organization serves 10,000 people on a 1,000,000 dollar program budget, so 100 dollars per beneficiary. The population splits in half: 5,000 people who are cheap to reach at 60 dollars each, and 5,000 who are hard to reach at 140 dollars each. Check: 5,000 times 60 is 300,000, plus 5,000 times 140 is 700,000, so 1,000,000 in total, at an average of 100 dollars.

Now optimize the metric. Drop the hard half and spend the whole budget on easy cases. Reaching 10,000 easy beneficiaries costs 600,000 dollars, and the remaining 400,000 dollars is free to expand further.

Average cost per beneficiary falls from 100 dollars to 60, an improvement of 40 percent. Beneficiaries served stay at 10,000. Every reported figure gets better, and the 5,000 hardest people, who are usually the reason the organization exists, receive nothing.

So the constraint has to be written into the structure as a term, in the form of a floor on coverage of the hardest segment. Without it, the optimization is correct and the recommendation is unusable. This is the single most common way a strong commercial candidate fails a non-profit prompt, and the failure looks like rigor on the way past.

What the scale rewards here, and who wrote it

The weights below are ours, read from the grading code in src/lib/grading. No consulting firm publishes a marking scheme for any case type, and none of the pages cited here attaches a weight to anything.

Drilling non-profit cost per beneficiary

Non-profit arithmetic drills well because every ratio has an inverse worth computing.

1. Invert the benefit. Recompute Diconsa with families of five instead of four. Beneficiary families become 14 million over five, so 2.8 million, and the total saving becomes 2.8 million times 156, which is 436.8 million pesos. A 25 percent change in family size moved the answer by 20 percent.
2. Find the overhead cap that binds. In the two charity table, hold delivery costs and ask what overhead ratio Charity B can afford before it drops below 8,000 beneficiaries. It needs 8,000 times 60, so 480,000 dollars of program budget, which allows 520,000 dollars of overhead, a ratio of 52 percent.
3. Price the mission constraint. Put a floor of 3,000 hard beneficiaries on the last example. That costs 3,000 times 140, so 420,000 dollars, leaving 580,000 for easy cases, which buys 9,667 of them. Total served is 12,667 at an average of 78.9 dollars, rounded.

Run these against a clock in the math drill, which returns Answer, Working, Units and Interpretation separately with the first correction named. It is free and needs no account, and it will not decide which denominator your prompt wants.

For a budget that is appropriated rather than raised, see public sector cases. For an objective measured in tons rather than people, see ESG cases.