Public Sector Cases 2026: Cost Per Unit Of Output, Worked

The term a public sector case loses from the equation

Start with what a commercial case assumes. Profit equals price times volume minus cost, and the client controls price. Remove that control and the structure changes shape rather than losing a branch.

What replaces it is a ratio: cost per unit of output delivered. Per filing processed, per patient treated, per pupil taught, per permit issued. The unit is the first thing to pin down, and it is usually the first clarifying question worth asking.

Then a second ratio, which is the one that makes a public case feel unfamiliar: cost per unit of outcome, which is not the same as output. Processing a filing is output. Collecting the tax owed is outcome. An agency can raise output per dollar while outcome per dollar falls, and a recommendation that improves only the first is the standard weak answer.

There is a third feature with no commercial analogue. The people served cannot leave. A company with a service problem sees volume fall next year, which is a signal with a number attached. An agency sees the same volume, because filing is compulsory, so the feedback that a commercial case relies on is missing from the data entirely.

Write those three facts down before branching, and the structure follows from them rather than from a shape. Our page on structured thinking covers why branches should fall out of the equation you wrote.

An agency's unit cost, from its own data book

Federal agencies publish enough to compute their own unit costs, which means a public sector case can be worked with real numbers instead of described in the abstract.

The Internal Revenue Service publishes its budget section at irs.gov, checked 30 September 2026: "IRS's actual expenditures were $19.0 billion for overall operations in Fiscal Year (FY) 2025" and "In FY 2025, the IRS used 95,226 full-time equivalent (FTE) positions in conducting its work".

The returns section at irs.gov, checked the same day, gives the volume: "More than 224.2 million returns and other forms were filed electronically. These represented 82.6% of all filings."

Those two figures give a total the agency does not print. Divide 224.2 million by 0.826 and you get 271.4 million filings, rounded to the nearest tenth of a million. That derivation is ours, and it is the denominator everything below uses.

Every number in the right column is rounded, and each is announced as rounded rather than presented as exact. The cost per filing lands on 70.0 dollars, which is close enough to a round number to be worth a sanity check: 271.4 million times 70 dollars is 19.0 billion, so the division holds.

Cost per hundred dollars collected, as an upper bound

The ratio a finance ministry actually watches is cost per unit of revenue collected, and here the published figures need care, because the Data Book puts its all-taxes total in a spreadsheet we did not open.

What is printed in the HTML is two lines. Individual income tax withheld and payments combined totaled 2.9 trillion dollars before refunds, and business income taxes added 486.4 billion dollars before refunds.

Against individual receipts alone: 19.0 divided by 2,900, times 100, gives 65.5 cents, so about 66 cents per 100 dollars collected.

Against individual and business receipts together, 3,386.4 billion dollars: 19.0 divided by 3,386.4, times 100, gives 56.1 cents, so about 56 cents.

Now the interpretive step, and it is the one worth borrowing. Both figures are upper bounds on cost per 100 dollars of total collections, because employment, excise and estate taxes would enlarge the denominator and every addition to a denominator lowers the ratio. So the true all-taxes figure is at or below 56 cents, and we can say that without having the total.

That sentence is what a bounded answer looks like when a figure is missing. The alternative is to invent a total, which is the failure CONTENT.md section 6.3 exists to prevent. Say the bound, say its direction, and name the table you have not opened.

The flagship efficiency lever, priced

Now run the recommendation an efficiency prompt invites, and watch it underdeliver.

Electronic filing stands at 82.6 percent of all filings. Suppose the agency targets 90 percent. The shift covers 271.4 million times the 7.4 point gap, which is 20.1 million filings moving from paper to electronic.

Assume paper processing costs 8 dollars more per filing than electronic. That figure is ours, declared, because the Data Book publishes no unit processing cost at all. The saving is 20.1 million times 8 dollars, so 160.8 million dollars a year.

Set that against the budget: 160.8 million over 19.0 billion is 0.85 percent. Cost per filing falls from 70.00 dollars to 18,839.2 million over 271.4 million, which is 69.41 dollars, a move of 59 cents.

So the headline efficiency lever, pushed seven and a half points in one year, moves the unit cost of the agency by under a percent. That is the finding, and it generalizes: in a public case the large numbers live in the mandate and the tax base, and operations move the third decimal.

A candidate who stops at the saving has answered a smaller question than the one asked. The useful next sentence is about where the leverage actually is: 47.2 million paper filings remain today, since 271.4 minus 224.2 is 47.2, and even eliminating every one of them saves 377.6 million dollars on our assumed differential, which is 2.0 percent of the budget.

Decision rights belong in a public sector case structure

In a commercial case, feasibility is a risk you mention at the end. In a public case it is a term, because the client does not own most of the levers.

1. The rate. Set by legislation. The agency cannot move it, and a recommendation that relies on it is a recommendation to a different institution.
2. The base. Set by legislation and by the economy. Same constraint.
3. Administration. Owned by the agency. This is where the 0.85 percent above lives.

One lever of three, and it is the smallest. Writing that split into the structure early does two things: it prevents an unusable recommendation, and it tells the interviewer you understood who the client is.

It also reframes the objective honestly. If administration is the only owned lever, then the question is rarely how to collect more; it is how to deliver the same mandate for less, or deliver more service for the same money. Those are different prompts and they need different structures.

The Data Book itself hints at where the remaining leverage sits. It names four budget activities, Taxpayer Services, Enforcement, Operations Support and Business Systems Modernization, and describes what each funds. The amounts per activity are in spreadsheet tables we did not open, so we make no claim about the mix, and a candidate in a real interview would ask for exactly that table first.

What one firm publishes about a public sector case

Bain lists a public sector case among its practice material at bain.com, checked 30 September 2026. Its own description reads: "This case involves a public sector organization aiming to identify strategic growth opportunities while addressing inherent bureaucratic challenges and shifting market dynamics. It requires candidates to evaluate various initiatives to drive sustainable improvement and enhanced operational efficiency in a complex, regulated environment."

Two things are worth extracting from three sentences. The firm reaches for commercial vocabulary, "strategic growth opportunities", and it names the constraint in the same breath, "a complex, regulated environment". Both registers are present, which matches the structure above: an owned lever inside a boundary set by someone else.

One disclosure is owed here. The case itself is hosted on a third-party preparation platform, which is a direct competitor of ours, so we cite the description written on the firm's page and do not link or read the exercise. That means the Bain public sector case is a case described by Bain and published by someone else, and anyone who tells you the firm publishes it has not checked.

The wider record is thin in a way worth stating. Across our research files, the published case material with a non-commercial objective amounts to that description, McKinsey's Diconsa exercise, and one Oliver Wyman case with a social product. No firm in our files publishes a worked public sector case with figures on its own domain, and the five hiring steps Booz Allen publishes contain no case interview at all, with security clearance as the filter that firm documents most carefully.

What the scale rewards on a public prompt

The weights below are ours, read from the grading code in src/lib/grading. No consulting firm publishes a marking scheme, which holds across every research file we keep, and the absence is the reason we publish ours.

Drilling the public sector ratios by hand

These ratios drill well because the published inputs are stable and the inverses are quick.

1. Solve for the differential. What paper-to-electronic cost gap would make the 90 percent target worth 500 million dollars a year? Divide 500 million by 20.1 million filings and the gap is 24.88 dollars, rounded. Then ask whether that is plausible, which is the real question.
2. Solve for the headcount. At 2,850 filings per full-time equivalent, handling 300 million filings needs 300 million over 2,850, which is 105,263 positions, about 10,000 above today.
3. Bound a missing figure on purpose. Recompute cost per 100 dollars collected assuming all-taxes receipts of 5 trillion dollars: 19.0 over 5,000 times 100 is 38 cents. Notice the direction matches the bound, which is the check that your reasoning about the denominator was right.

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, it knows nothing about federal budgets, and it will not tell you which unit your prompt means.

For a budget raised from donors rather than appropriated, see non-profit cases. For an objective measured in metric tons, see ESG cases.