Governments across the OECD spent an average of 42.6% of GDP in 2023, and none of it was meant to earn a return for shareholders. Every guide to public-sector cases opens with that fact, and it is not the one that decides your score. What the interviewer is listening for is whether you can define the denominator — cost per what, exactly — and then name which of four clocks (appropriation, statute, procurement, capacity) is stopping the client from buying more of it.
Most candidates do neither. They swap "profit" for "impact", draw a profitability tree with the labels changed, and spend the last ten minutes recommending something the client is legally unable to do this year.
By the end of this piece you will be able to open a government case by naming the unit of outcome the money buys, say out loud which clock binds, and work a backlog case with arithmetic you can reproduce on paper.
What the sheet rewards when there is no profit line
The scoring sheet does not change. Structure, quantitative work, judgment, synthesis, and communication are scored on the same scale whether the client is a retailer or a tax authority. What changes is what earns the point.
| What is scored | Retail case | Government case |
|---|---|---|
| Structure | Profit = revenue minus cost, broken into drivers | Outcome per dollar, with the outcome defined, plus the constraint (by when, under which rule) |
| Quantitative work | Margin, break-even, market size | Cost per outcome, capacity versus demand, time to target |
| Judgment | Which lever moves margin fastest | Which lever the client is allowed to pull this fiscal year |
| Synthesis | "Do X, worth $Y a year" | "Do X, hits the target by month Z, and here is who has to agree" |
In a typical first round of eight candidates, my estimate is that two or three open a public-sector case with an unaltered profitability tree. They do not fail for it. They lose four minutes, and they rarely recover the constraint they skipped.
Firms do not publish their case banks, so treat any claim about how often these prompts appear as folklore. What is documented is that some consultancies are almost entirely government businesses: Booz Allen reported that U.S. government agencies accounted for 98% of its revenue in fiscal 2025, on $12.0 billion of revenue. Interview there, at PA Consulting, or with a public-sector practice at a strategy firm, and the prompt will be public sector by default.
Define the denominator before you touch the numerator
"Cost per outcome" sounds obvious. The catch is that the client's target usually fixes the numerator (a budget) and leaves the denominator to be argued about, and the denominator changes the answer.
Take a work-placement program with two providers, same ministry, same year:
| Provider A | Provider B | |
|---|---|---|
| Annual cost | $10.0m | $12.0m |
| Placements into work | 5,000 | 4,000 |
| Still employed at 26 weeks | 40% (2,000) | 75% (3,000) |
| Cost per placement | $2,000 | $3,000 |
| Cost per sustained placement | $5,000 | $4,000 |
If the ministry counts placements, A wins by a third. If it counts people still in work six months later, B wins by a fifth. Neither number is wrong. The interviewer wants you to ask, in the first two minutes, "how does the department count success, and is that the number the minister reports?" The published target is very often the weaker denominator, and knowing that is a judgment point.
Name the denominator, then check whether it is the one the client is judged on. Traps: cost per FTE (an input), cost per application received (activity, not outcome), and "efficiency savings" with no unit at all.
Budgets are not P&Ls
Five rules a private-sector case never taught you. Each is written down by a treasury or an auditor; none is folklore.
| Private-sector habit | What is true for a government client | What to ask |
|---|---|---|
| Unspent money is saved | U.S. one-year appropriations left unobligated at year-end expire and cannot fund new obligations. A UK department can carry an underspend forward only if it forecast it before the Supplementary Estimate; unforecast underspends cannot be carried forward. | "Is this year's money committed, and does it lapse?" |
| Capex and opex are fungible | UK departments may not switch capital budgets to resource budgets; the reverse is allowed. A saving on buildings cannot pay for nurses. | "Which line does this hit, capital or current?" |
| Spend now, save later | Agencies may not obligate in advance of, or in excess of, an appropriation: the Antideficiency Act. A program that pays back in year three still needs year-one money voted. | "Who votes the up-front money, and when?" |
| Savings accrue to the saver | In my experience a department that underspends is treated as needing less next year, so the rational manager spends the money in March. | "Is the saving recycled inside the department or returned to the center?" |
| Raise prices | Fees are typically set in statute or regulation; changing them is a legal process with its own clock. | "Is the fee statutory? If so, park it." |
"We could reinvest the savings", said without asking whether the savings survive year-end, is a plan the client cannot execute. Say the rule, then design around it.
The wrong-pockets problem
Here is the thing other guides do not say. In a business, the entity that pays for an improvement is the entity that benefits. In government, it usually is not.
A city funds home insulation for low-income households. The city pays. The savings land in three other pockets: the national health system (fewer cold-related admissions), the energy regulator's bill-payers (lower peak demand), and the national welfare budget (lower fuel subsidies). The city's own budget sees almost nothing. That is why the "obvious" prevention program has not happened, and it is the answer to the interviewer's favorite follow-up: "if this is so good, why hasn't the client done it?"
Wrong pockets appears whenever a recommendation involves prevention, early intervention, shared data, or moving activity between levels of government. In federal systems (the U.S., Germany, Canada, and the UAE) it is the default state of the world.
Ned's rule. If your recommendation saves money for a budget the client does not hold, it is not a saving; it is a gift. Say who pockets it before you say how much.
The fix is not to abandon the idea. Name the pocket and propose the mechanism: a cost-sharing agreement, a pooled budget, a central grant that follows the outcome, or legislation. "The government as a whole saves money" earns nothing, because the government as a whole does not sign contracts.
The four clocks, and which one binds
Every public-sector recommendation is gated by one of four clocks. Usually only one binds, and the exhibits tell you which.
| Clock | What starts it | A rule or threshold you can quote | Where it hides in the exhibits |
|---|---|---|---|
| Appropriation | Asking for money not already voted | The U.S. fiscal year ends 30 September; UK carry-forwards are settled before the Supplementary Estimate | A budget table with this year's money already committed |
| Statute | Anything the law fixes: fees, the number of judges or inspectors, a wet signature, eligibility rules | The law itself; the interviewer will tell you if you ask | A process step nobody can explain except "it is required" |
| Procurement | Buying anything from outside | UK central government goods and services: £135,018 including VAT from 1 January 2026. U.S. simplified acquisition threshold: $350,000 from 1 October 2025, up from $250,000. Above these, a full competitive procedure applies | A vendor line, an IT build, an outsourced service |
| Capacity | Any plan needing more trained people or rooms | A judge, a nurse, or an inspector takes years to produce, not months | A demand-versus-capacity chart; staff at 90% are a bottleneck, staff at 55% are a scheduling problem |
Two things to notice. The thresholds are small: a $400,000 software contract in a U.S. agency heads into a full procedure. And durations are rarely printed anywhere reliable; a competitive tender from decision to signed contract runs, in my experience, four to nine months, and I would label that an estimate in the room rather than a fact.
The move that earns the point: after your structure, say which clock you think binds and why, then ask for the exhibit that would confirm it. "The target is 15 months away. If the fix needs new law we are already late, so I want to know first whether the three-signature step is statutory or habit."
Worked case: the Halvern courts backlog
An original case. All figures are invented; the shape is not.
Prompt. Halvern's civil courts receive 48,000 filings a year and dispose of 44,000. Pending cases stand at 33,000. The legislature has set a target: an average time to disposal of six months, within three years. The courts administrator has three options and wants to know which to fund.
Step 1: size the gap. Little's law says pending cases equal throughput multiplied by average time in the system. Today: 33,000 ÷ 44,000 per year = 0.75 years, about nine months. Six months at 48,000 filings a year means pending cases must fall to 48,000 × 0.5 = 24,000. So the courts must dispose of 9,000 more than they receive over three years and also close the 4,000-a-year gap that is growing the pile: (4,000 × 3) + 9,000 = 21,000 extra disposals, or 7,000 a year.
Step 2: cost per extra disposal, and which clock binds.
| Option | Extra disposals a year | Annual cost | Cost per extra disposal | Binding clock | Earliest start | Extra disposals inside the three-year window |
|---|---|---|---|---|---|---|
| A. Six new judgeships (1,000 disposals each) | 6,000 | $2.7m | $450 | Statute: the legislature creates judgeships | Month 18 | 1.5 × 6,000 = 9,000 |
| B. Mandatory mediation below a claim threshold (8,000 diverted, 55% settle) | 4,400 | $2.2m | $500 | Procurement: an above-threshold contract needs a full tender | Month 8 | 2.33 × 4,400 = 10,300 |
| C. Recalled retired judges plus evening sittings | 3,000 | $1.5m | $500 | Appropriation: in-year money, plus courtroom hours | Month 3 | 2.75 × 3,000 = 8,250 |
Cost per extra disposal sits within 10% across the three. The denominator does not decide this case; the clocks do. B plus C deliver about 18,500 of the 21,000 needed, roughly 88% of the gap. All three together cross 21,000 at about month 30.
Step 3: the denominator twist. Had the legislature written the target as "80% of cases disposed within six months of filing", clearing the backlog would make the metric worse before it makes it better, because the first cases disposed are the oldest. Mediation, which closes a case in weeks, would score fully; new judges working through two-year-old filings would not. Same courts, same money, opposite ranking. Ask which denominator is in the statute.
The recommendation as I would want to hear it. "Fund C this quarter from the current-year resource budget, tender B now so it is live by month eight, and introduce the judgeship bill this session, because A is the only option that survives once the temporary money ends. Expect the six-month average around month 30, and warn the legislature that the timeliness percentage will dip in year one while old cases clear. Two risks: the mediation savings land in the legal-aid budget, not the courts', so agree the cost-share first; and the tender slips if nobody has started the paperwork."
The real world is not kinder. In England and Wales the median time from a small claim being issued to trial was 36.1 weeks in October to December 2025, and fast, intermediate, and multi-track claims took a median 57.4 weeks. A backlog case is not an abstraction.
Practice this today
Take any case prompt, whether a hospital, a licensing office, a school district, or a transit agency, and before you draw a single branch write one sentence in this form: "The client wants more outcome per dollar, by date, and the thing most likely to stop it is clock." Then structure under that sentence, not over it.
Run three of those in a row on /start: three typed turns of a case with instant scores, no account needed. When the structure holds, do a full voice case with Ned on /interview and listen for whether you named the denominator before the interviewer had to ask. The case library has 64 original cases, several with a public body as the client, and the structure drill is the fastest way to break the profitability-tree reflex.
Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
CoachNed is independent and not affiliated with McKinsey, BCG, Bain, Deloitte, Booz Allen, PA Consulting, or any other firm named here.
Frequently asked questions
Do public sector case interviews use a profitability framework?
Rarely at the top of the tree. The objective is a defined outcome per unit of money under a constraint, so the first branch is the denominator (what counts as success) and the second is the binding constraint (budget, statute, procurement, or capacity). Cost trees still appear inside the structure; profit trees do not.
Which consulting firms do the most government work?
Booz Allen reported that 98% of its fiscal 2025 revenue came from U.S. government agencies. Deloitte, Accenture, PA Consulting, and the public-sector practices at McKinsey, BCG, and Bain are the other names candidates most often meet; the Booz Allen and PA Consulting guides cover how their interviews differ.
How do you size demand in a government case?
Population, times the share eligible, times the share who take up the service, times frequency. The take-up step is where candidates go wrong: entitlement is not demand. Then compare the result with capacity, because the answer is almost always a queue, not a market.
Is a public sector case easier than a private sector case?
The arithmetic is usually lighter and the judgment heavier. You lose points not on the math but on recommending something the client cannot legally, financially, or procedurally do inside the time the target allows.
Sources
- Government at a Glance 2025, full report (OECD) — general government expenditures averaged 42.6% of GDP in 2023, after a 48.3% peak in 2020. Checked 2026-09-24.
- Booz Allen Hamilton Form 10-K for fiscal 2025 (SEC) — U.S. government agencies accounted for 98% of revenue. Checked 2026-09-24.
- Booz Allen Hamilton fiscal 2025 results, Exhibit 99.1 (SEC) — revenue of $12.0 billion and about 35,800 employees as of 31 March 2025. Checked 2026-09-24.
- Principles of Federal Appropriations Law, Fourth Edition, Chapter 1 (GAO) — the bona fide needs statute; time-limited appropriations expire when unobligated. Checked 2026-09-24.
- Antideficiency Act resources (GAO) — the prohibition on obligating in advance of or in excess of an appropriation. Checked 2026-09-24.
- Consolidated budgeting guidance 2025-26 (HM Treasury) — no switches from capital to resource budgets; budget exchange only for underspends forecast before the Supplementary Estimate. Checked 2026-09-24.
- PPN 023: 2026 threshold amounts (Cabinet Office) — £135,018 central government and £207,720 sub-central thresholds from 1 January 2026. Checked 2026-09-24.
- Threshold changes effective 1 October 2025 (Acquisition.gov) — simplified acquisition threshold raised from $250,000 to $350,000. Checked 2026-09-24.
- Civil justice statistics quarterly, October to December 2025 (Ministry of Justice) — median 36.1 weeks from issue to trial for small claims and 57.4 weeks for fast, intermediate, and multi-track claims. Checked 2026-09-24.