Rubric — Case Study Analysis
Every chapter in this book carries two case studies: one drawn from a real, public event, and one labeled composite built from patterns that recur across many operations. Eighty case studies in total. This rubric grades a written analysis of any one of them.
It is designed for a 900–1,500 word response, which is the length at which a student must choose what to leave out — and choosing what to leave out is most of the skill.
1. What is being assessed
Whether the student can separate what the case shows from what the student already believed.
Restaurant case studies are unusually vulnerable to a specific failure: the student arrives with a strong prior — they should have controlled food cost, they expanded too fast, the owner didn't listen to staff — and reads the case as confirmation. The analysis that results is fluent, confident, and contains no information that was not present before the case was read.
Grade against that. The strongest analyses change their own mind somewhere in the middle, or at minimum identify the point at which the evidence stopped supporting the obvious reading.
2. The five dimensions
| Dimension | Weight | The question |
|---|---|---|
| A. Evidence discipline | 25% | Are claims traced to what the case actually says? |
| B. Diagnosis | 25% | Is the identified cause the operative one, or the most visible one? |
| C. The counterfactual | 20% | What decision, made when, with what information, would have changed the outcome? |
| D. Transfer | 20% | What does this case tell an operator who is not in this situation? |
| E. Written craft | 10% | Is it organized, specific, and the right length? |
3. Dimension A — Evidence discipline (25%)
Every factual claim in the analysis should be traceable to the case, to the chapter, or to a labeled assumption. Students should be explicit about which.
What earns full marks:
- Direct reference to figures in the case, used correctly. If the case gives a P&L, the analysis works from its actual lines.
- Arithmetic done, not gestured at. "Labor was high" is worth nothing. "Labor at 38.5% against a 32% plan is \$100,750 on \$1.55M — roughly forty percent of the operating profit line" is the analysis.
- Clear marking of inference: the case does not say X, but Y and Z together suggest it.
- Explicit acknowledgment of what the case does not contain. Every case omits something material. A student who names the missing information — and says what they would have asked for — is doing the work.
What loses marks:
- Numbers introduced from outside the case with no source.
- Confident claims about the operator's motives, competence, or state of mind. The case shows decisions and outcomes; it rarely shows why.
- Treating a composite case as though it were a documented event, or a real case as though its published account were complete. Both distinctions are labeled in every case study in this book, and students should honor them.
On the real cases. These involve actual businesses and, in several instances, actual harm — a foodborne illness outbreak, a closure, a bankruptcy. Hold students to the same standard the book holds itself to: analyze the decisions, not the people. An analysis that reaches for contempt is not a better analysis, and it is usually a shallower one, because contempt closes the question of why a competent person did the thing.
4. Dimension B — Diagnosis (25%)
The most visible cause is rarely the operative one, and every case in this book is built so that the two differ.
The pattern to look for: a student who identifies a proximate cause and keeps going.
| Proximate cause the student names | The operative cause underneath it |
|---|---|
| "Food cost ran high" | No usage-based COGS — the operator was reading purchases ÷ sales and could not see the variance |
| "They ran out of cash" | Profitable on the P&L; the cash conversion cycle and a payroll calendar were never modeled |
| "The health inspection failed" | A schedule that left one cook alone during prep, which was a labor decision made two months earlier |
| "Turnover was crippling" | A hiring practice that filled seats fast, plus training that never happened, plus a schedule nobody could plan a life around |
| "The second location failed" | The first location's performance depended on the owner's presence, and nobody had measured that |
| "Reviews collapsed" | Service recovery was never systematized, so every failure became permanent |
Full marks require naming a mechanism, not a category. "Poor cost control" is a category. "Inventory was counted monthly, so a 4.2-point variance was invisible for five weeks and by the time it surfaced the period it belonged to could not be reconstructed" is a mechanism.
Credit multi-causal diagnoses that rank. Real failures have four causes. An analysis that names four and says which one, removed, would have changed the outcome is stronger than one that names the single "root cause" — a phrase worth discouraging, because it implies a singularity that restaurants rarely have.
5. Dimension C — The counterfactual (20%)
The hardest dimension, and the most valuable.
The question is not "what should they have done?" It is "what decision, made at what point, with what information that was actually available at that point, would have changed this?"
That constraint — information actually available at that point — is the whole exercise. It rules out the entire class of costless hindsight answers, which is the class most student analyses live in.
Look for:
- A specific decision point with a date or a triggering event. "In month four, when the second consecutive period showed a food cost variance above two points."
- The information available then, stated. Often less than the student assumes.
- The signal that would have prompted the decision, and whether the operator could have seen it. This is where students discover that the answer is usually a reporting change, not a judgment change — the operator was not wrong, they were blind, and blindness is a fixable system property.
- The cost of the alternative. Every counterfactual has a price. A weekly inventory count costs three hours a week of a salaried manager who is already fully committed. An analysis that proposes controls with no cost has proposed nothing.
Explicitly penalize the free-lunch counterfactual: "they should have negotiated better terms," "they should have hired better people," "they should have marketed more." These name an outcome, not a decision, and they are available in every case regardless of its facts — which is how you know they are not analysis.
6. Dimension D — Transfer (20%)
What does this case tell an operator whose situation is different?
This dimension guards against the opposite failure from Dimension A: an analysis so tightly bound to the case's specifics that it yields nothing.
Full marks:
- A generalized principle stated in one or two sentences, at the right altitude — neither "count inventory" nor "management matters."
- The boundary of the principle. Where does it stop applying? A case about a 68-seat independent may not transfer to a 300-seat operation or a two-person food truck, and saying where it breaks is more impressive than claiming it does not.
- Application to the student's own business plan, where the course is running the progressive project. This is the best version of this dimension: my plan has the same exposure, here, and here is the specific change I am making because of this case. Reward it heavily.
7. Dimension E — Written craft (10%)
- Answers the question asked, in the first paragraph.
- Specific over general — a number, a date, a decision, in place of an adjective.
- Within length. A 2,500-word response to a 1,200-word assignment is not generosity; it is a failure to choose.
- No restatement of the case. The instructor has read it. Every sentence spent summarizing is a sentence not spent analyzing, and this is the single most common structural flaw in student case work.
8. The scoring grid
| Exemplary (90–100) | Proficient (75–89) | Developing (60–74) | Insufficient (<60) | |
|---|---|---|---|---|
| A. Evidence | Every claim traced; arithmetic performed; missing information named | Claims mostly traced; some arithmetic | Assertions loosely tied to the case | Outside claims, invented figures, or motive attribution |
| B. Diagnosis | Mechanism named and ranked among causes | Operative cause identified, mechanism partial | Proximate cause only | Category label ("bad management") |
| C. Counterfactual | Specific decision, specific point, information-constrained, costed | Specific decision; cost or timing weak | Vague ("should have watched costs") | Free-lunch hindsight |
| D. Transfer | Principle plus its boundary plus application to own plan | Principle stated and applied | Principle stated at too high an altitude | None, or a platitude |
| E. Craft | Tight, specific, answers the question first | Clear, some summary | Substantially summary | Off-question or unreadable |
9. Using the case studies in class
- The pair is the point. Each chapter's real case and composite case are chosen to disagree productively. Assign both and ask which is more useful, and why. A frequent and correct student answer: the real case is more credible and the composite is more transferable, because the composite was built to isolate a mechanism that the real case has tangled with a dozen others.
- The discussion questions at the end of each case are not the assignment. They are conversation starters, and they are answerable in class in ten minutes. A graded written analysis should use one question, or a question of the instructor's own, and go deep.
- Have students grade each other on Dimension C once. Writing a counterfactual is hard; recognizing a free-lunch counterfactual in someone else's work is easy, and doing that once permanently changes how a student writes their own.
- The best single prompt in the whole set, usable with almost any case: identify the moment at which this outcome became likely, and say what could have been known at that moment. It forces every dimension of this rubric at once.