A case structure is a tree for this decision, not a museum of named models. Interviewers fail people who open with “I’ll use the 3Cs” and pass people who say “profit is down $14m; I will split price, volume, mix, and cost, and I will start with mix because outpatient share moved.” This page owns that choice. It is not a directory of tools — those live on the frameworks hub.
The method has four moves: name the decision, pick diagnose-versus-decide, split the metric with no overlap, then bet which branch to test first.
Diagnose versus decide
Most prompts are one of two jobs. Mixing them is how trees become laundry lists.
| Job | Question | What a good tree does |
|---|---|---|
| Diagnose | Why did the number move? | Algebraic or process split that can isolate one driver |
| Decide | What should we do? | Options that do not overlap, each with a kill test |
| Design | How should the offer work? | Levers (offer, price, channel, awareness) with unit economics |
| Allocate | Where does scarce capital go? | Portfolio units, not a P&L of one brand |
If the CEO asks “should we enter Ohio,” you are not diagnosing last year’s margin. If the CEO asks “why did EBITDA fall,” you are not drawing a product/price/place/promotion grid. Say which job you are in, out loud, in ten seconds.
Worked example: hospital system down $14m
Prompt. Meridian Health runs three hospitals. EBITDA fell from $62m to $48m. The board wants a recommendation in 20 minutes. A weak candidate names a famous framework. A strong candidate builds this tree.
Decision restated. Find the driver of the $14m drop, then recommend one fix that the CFO can run in 90 days.
Diagnose first. Profit = revenue − cost. Revenue = price × volume × mix. Do not tour “company, customer, competitor.”
Ask for the bridge:
| Line | Prior year | This year | $ impact vs prior |
|---|---|---|---|
| Inpatient discharges | 28,400 | 28,900 | Volume +$2.1m |
| Average collected price | — | +1.2% | Price +$1.4m |
| Mix: commercial share | 41% | 34% | Mix −$18.0m |
| Operating cost | $411m | $412.5m | Cost −$1.5m |
| EBITDA | $62m | $48m | −$14m |
The driver is isolated: payer mix, not volume, not “the hospital is inefficient.” Commercial lost 7 points to Medicaid and Medicare Advantage. Cost barely moved. Naming Porter or 7S here would have been a fail — this is not an industry-structure case and not an org case.
Then decide. Solution tree on the mix branch only: (1) win back two orthopedic groups that switched referrals to a rival with faster MRI slots; (2) renegotiate one commercial contract up for renewal (12% of commercial revenue); (3) cut Medicaid-heavy service lines. Option 3 is a political bomb and a small dollar. Option 1 is operational and hits mix. Option 2 is a pricing negotiation.
Recommendation. Recover mix by cutting MRI wait from 11 days to 4 (the rival’s number) before the orthopedic groups’ next panel meeting. That is a capacity-and-SLA move in service of mix, not a “marketing” idea. Risk: the rival matches wait times. Next step: slot-level MRI utilization by hour.
You never needed a brand name. You needed a diagnostic split, one isolated driver, and a solution tree on that driver.
How to mutate a canned tree
Memorized buckets are raw material. Mutate them in the first 30 seconds:
- Drop a branch that cannot move the metric. If cost is flat, do not spend three minutes on procurement.
- Add a branch the template forgot. Hospitals have payer mix. Airlines have load factor and yield. SaaS has churn and net expansion. Cement has utilization and energy.
- Rename buckets in the client’s words. “Collected price” not “price.” “Discharges” not “volume.”
- Sequence by the bet, not by the mnemonic. Start where the exhibit will kill the hypothesis fastest.
If you cannot say why branch two can wait, you do not have a hypothesis. You have a list.
When this method is the wrong move
Do not force a four-box onto a math question. If they handed you an exhibit of kiln energy $ / tonne, compute. Structure later if the so-what needs a decision tree.
Do not use this page as a tour of named tools. If you need profitability, market entry, M&A, or Porter, open that article. This page is the selector, not the encyclopedia.
Do not structure a PEI or fit story this way. Stories need an arc, not an issue tree.
The mistake that is unique to “knowing frameworks”
Announcing the name. “I’ll apply McKinsey’s 7S” on a pricing prompt is a meme and a score of 2. The unique fail of the complete-guide mindset is covering every tool you studied so the interviewer sees breadth. They see indecision. One mutated tree beats seven labels.
What to practice
Read one method article that matches this week’s case type. Then close it. Run a structure drill on a different industry so you cannot copy the example. If you hear yourself say a brand name, stop and replace it with the decision and the first test.
Pick a tree, then throw the name away
Timed structure reps grade whether the branches fit the prompt — not whether you remembered a label.