Rubric — The Business Plan
The progressive project. Forty checkpoints, one document, graded four times.
Every chapter of this book ends with a 🍽️ The Business Plan callout that contributes one piece of a
real restaurant business plan. By Chapter 40 the student is holding a complete financing package for a
restaurant they invented in Chapter 2. This rubric grades that document.
It is the single most important assessment in the course, and it should carry the largest share of the grade — 50–60% is defensible, and 40% is the floor. Nothing else in the course tests whether the student can hold forty decisions in their head at once and make them agree.
1. What is actually being assessed
Not writing quality. Not concept originality. Not enthusiasm.
Whether the numbers hold together, and whether the student knows which of them they cannot defend.
A business plan is a set of interlocking claims. The menu price implies a food cost percentage; the food cost percentage implies a COGS line; the COGS line plus the labor line implies a prime cost; the prime cost plus occupancy, other operating, and G&A implies an operating profit; the operating profit implies a debt service coverage ratio. Change any one and four others move. A student who has understood this course produces a plan where they all move together. A student who has not produces forty individually reasonable pages that contradict each other, which is — and this is worth saying to the class out loud — exactly what most real business plans look like.
Grade for internal consistency first, realism second, presentation third. In that order, every time.
2. The five dimensions
| Dimension | Weight | What it asks |
|---|---|---|
| A. Arithmetic integrity | 30% | Does every number compute, and does every number agree with every other number? |
| B. Operational grounding | 25% | Do the financial claims correspond to a restaurant that could physically exist and be staffed? |
| C. Assumption disclosure | 20% | Is every load-bearing assumption stated, sourced, and labeled as to its confidence? |
| D. Risk and downside | 15% | Does the plan model what happens when it is wrong? |
| E. Document craft | 10% | Can a reader who is not the author navigate, verify, and act on it? |
3. Dimension A — Arithmetic integrity (30%)
This is the dimension where a plan fails outright. Check it with a calculator, not a reading.
The eleven ties that must hold. Give one point for each; this dimension is largely mechanical and should be graded mechanically.
| # | The tie | How to check in thirty seconds |
|---|---|---|
| 1 | Cost cards foot | Extended costs sum to subtotal; waste allowance computes; plate cost ÷ menu price = the stated food cost % |
| 2 | Menu mix sums to 100% | The mix percentages in the menu-engineering table add to 100.0 (±0.1 for rounding) |
| 3 | Weighted food cost ties to the P&L | Menu-mix-weighted food cost × food revenue ≈ the COGS food line |
| 4 | Beverage revenue ties to pour cost | Pour cost % × beverage revenue = the COGS beverage line |
| 5 | COGS food + COGS beverage = COGS | And the blended percentage is a result, not an input |
| 6 | The labor schedule produces the labor line | Positions × hours × rates × (1 + burden) = the dollar figure in the P&L, built from a schedule, not from a percentage |
| 7 | Prime cost = COGS + labor | And the prime cost percentage is computed from dollars, not asserted |
| 8 | The P&L foots | All expense lines + operating profit = revenue, to the dollar |
| 9 | Revenue ties to capacity | Covers × average check × operating days = revenue, and covers ≤ seats × turns |
| 10 | Break-even ties to the P&L | Fixed costs ÷ contribution margin ratio = break-even sales, using the plan's own fixed/variable split |
| 11 | The cash forecast ties to the P&L | Twelve months of the cash forecast reconcile to the annual profit, with the differences named (principal, depreciation, inventory build, draws, sales tax timing) |
Severity guide. Distinguish between a rounding difference and a reconciliation failure. A plan whose P&L is off by \$12 because percentages were rounded at each line has a presentation problem. A plan whose labor schedule produces \$570,000 while the P&L says \$500,000 has a credibility problem — and, critically, the fix is not to change one to match the other. The fix is to state which is which and why.
The single most common failure, and how to grade it. Students back-compute dollar figures from rounded percentages. They see "labor 32.3%" and write \$500,650 rather than the \$500,000 the schedule actually produces. Then everything downstream inherits the \$650. Dollars are canonical; percentages are rounded displays. Take points for this every time. It is the habit that turns a plan into a work of fiction one line at a time.
Scoring band.
| Score | Description |
|---|---|
| 27–30 | All eleven ties hold. Any residual differences are rounding, disclosed as such. |
| 21–26 | One or two ties fail, and the failures are disclosed or trivially mechanical. |
| 15–20 | Three or more ties fail, or one core tie (the P&L, the labor schedule, capacity) fails silently. |
| 0–14 | The plan does not foot and does not know it. The financial section is not evidence of anything. |
4. Dimension B — Operational grounding (25%)
The test: could a person run this restaurant on a Friday night?
Look for the specific places where a plan detaches from a physical kitchen.
- Capacity. Does the revenue projection exceed what the kitchen can produce? Check the equipment against the menu. A plan projecting 140 covers a night on a kitchen with one hearth that produces twenty-eight items an hour is not projecting revenue; it is projecting a ninety-minute ticket time and a dining room that empties itself.
- The schedule. Is there a written schedule — positions, days, shifts, hours, rates — or only a labor percentage? The schedule is the load-bearing document. Ask for it explicitly.
- Coverage at the edges. Who opens? Who closes? Who covers the salaried manager's two days off? A plan with 24 scheduled positions and no answer to those three questions has not been scheduled.
- The menu and the station map. Does every dish have a station, and does any single station carry more than it can execute at peak? This is where an otherwise excellent menu falls apart.
- Prep. Is there prep labor in the schedule, or does the plan assume à la minute execution of items that take four hours?
- The ramp. Does the plan open at its projected volume? It should not. A first quarter that runs at plan is the clearest single sign the student has not internalized the ramp.
Scoring band.
| Score | Description |
|---|---|
| 23–25 | Revenue, menu, equipment, and schedule are mutually consistent; the ramp is modeled; edge coverage is named. |
| 18–22 | Broadly coherent, with one operational gap (usually prep labor or edge coverage). |
| 13–17 | The financials describe a restaurant the operational sections could not run. |
| 0–12 | No schedule, or a revenue projection with no capacity check behind it. |
5. Dimension C — Assumption disclosure (20%)
This dimension is where the strongest students separate themselves, and it is the one instructors under-weight.
A plan cannot be right. It can be legible — meaning a reader can find every assumption, see where it came from, and change it. Grade for legibility.
Require an assumptions register: a single table listing every load-bearing assumption, its value, its source, and its confidence. Look for these being disclosed rather than buried:
| Assumption | What a strong plan says |
|---|---|
| Average check | Built from a priced menu and an estimated mix, not asserted |
| Covers per night | Tied to seats and turns, with the turn assumption defended |
| Food cost % | The weighted result of the cost cards, not a target imposed on them |
| Labor % | The result of the schedule |
| The ramp | An explicit month-by-month curve, not a flat year |
| Seasonality | Named, even if roughly |
| Occupancy | From the actual lease terms, including NNN and escalation |
| Sales mix | Food vs. beverage, dinner vs. brunch, on-premise vs. off |
Give substantial credit for a student who says "I do not know." A plan that marks its turn assumption as low confidence, most sensitive input and then shows what happens at 1.2 and 1.6 turns is a better plan than one that asserts 1.4 with fake certainty — and it is a much better predictor of the student's future as an operator.
Take points for invented precision. Any statistic without a source, any percentage carried to two decimals with no derivation, any claim about "the industry" that cannot be traced. If a student cites a figure, they must be able to say where it came from; "I read it somewhere" is a zero for that line.
Scoring band.
| Score | Description |
|---|---|
| 18–20 | Complete assumptions register with confidence levels; the weakest assumptions are named by the student before the instructor finds them. |
| 14–17 | Most assumptions disclosed; confidence is implicit rather than stated. |
| 10–13 | Assumptions are embedded in the text and must be excavated. |
| 0–9 | Key numbers appear with no derivation, or invented statistics are presented as fact. |
6. Dimension D — Risk and downside (15%)
Ask one question: what does this plan do when it is wrong?
Look for:
- A sensitivity analysis on at least three inputs — normally revenue, food cost, and labor — showing operating profit at each. One-variable-at-a-time is acceptable at this level; a combined downside case is better.
- A cash trough. Where is the low point of the first year, how deep is it, and what funds it? A plan whose worst cash week is not identified has not been cash-forecast. Students consistently discover that the trough is not in the slowest month; it is in an early payroll cycle, before revenue has ramped.
- Working capital sized against the trough, not against a round number. "\$45,000 of working capital" is an amount. "\$45,000 of working capital against a modeled trough of \$16,000 plus thirty days of operating cost" is a plan.
- A named list of what would have to be true for this to fail, and what the operator would do first.
- Honest treatment of the personal stake. Most first restaurants involve a personal guarantee. A plan that models the business's downside without acknowledging whose house is collateral is incomplete.
Do not reward pessimism for its own sake. A downside case that assumes revenue at 60% of plan and concludes "we would close" is not analysis. The useful question is where the break is: at what revenue does the business stop covering its obligations, and how far is that from plan?
Scoring band.
| Score | Description |
|---|---|
| 14–15 | Multi-variable sensitivity, an identified cash trough, working capital sized against it, and a stated break point. |
| 11–13 | Sensitivity on the main inputs; downside acknowledged; the trough is identified but not funded explicitly. |
| 7–10 | A single downside scenario, or a sensitivity table with no conclusion drawn from it. |
| 0–6 | No downside case. The plan assumes it is right. |
7. Dimension E — Document craft (10%)
The plan is read by someone who did not write it, is not obligated to like it, and has four others on the desk.
- Executive summary that states the ask, the use of funds, and the return in the first paragraph.
- Every table labeled and every figure numbered, so a reader can refer to something without quoting it.
- A single source of truth for each number. If revenue appears in six places, it is the same number in all six.
- Navigable structure — a reader looking for the labor model finds it in under thirty seconds.
- Appendices carry the detail; the body carries the argument. Cost cards belong in the back.
- No unexplained jargon, and no explained jargon that did not need to be there.
Scoring band. 9–10: a reader could act on it. 7–8: clear, with friction. 5–6: requires the author present to interpret. 0–4: a folder of documents, not a plan.
8. The four graded checkpoints
Do not grade this once at the end. A plan graded only in week fifteen cannot be fixed, and the entire pedagogical value of a progressive project is the revision.
| Checkpoint | After chapters | What is due | Weight of project grade |
|---|---|---|---|
| 1 — Concept and site | 1–9 | Concept statement, market and competitive set, site and lease summary, floor plan, licensing and pre-opening timeline | 15% |
| 2 — Product and people | 10–21 | Costed menu with cost cards, menu-engineering matrix, beverage program, organizational chart, labor model and schedule | 25% |
| 3 — Operations and channels | 22–30 | Service standards, revenue model, food safety plan, technology stack, marketing plan, off-premise strategy | 20% |
| 4 — The complete package | 31–40 | Three-year P&L, break-even, cash flow forecast, financial controls, growth and risk sections, and the assembled document | 40% |
Checkpoints 1–3 should be graded on Dimensions B, C, and E only. Dimension A cannot be assessed until there are numbers to tie together, and Dimension D cannot be assessed until there is a forecast to stress. Say this to students in advance, because it changes how they spend their effort.
Require revision. The most valuable single instruction in this course is: the checkpoint-2 menu you costed is now inconsistent with the checkpoint-4 P&L. Fix one of them and tell me which and why. That is the exercise. Everything else is preparation for it.
9. Grading efficiently
A complete plan takes forty minutes to grade properly. With thirty students that is twenty hours, which nobody has. Three compressions that preserve the assessment's integrity:
- Grade Dimension A first and mechanically, with the eleven-tie checklist and a calculator. It takes eight minutes and it tells you most of what you need to know. A plan that fails five ties does not need a close reading of its marketing section.
- Require a one-page reconciliation summary as the plan's first appendix: revenue, COGS, labor, prime, occupancy, other operating, G&A, operating profit, break-even, and the cash trough, each with the page where it is derived. This is a real professional artifact and it cuts grading time roughly in half. Students who cannot produce it have learned something by discovering that.
- Use peer review for Dimension E and reserve your own time for A through D. Document craft is the dimension students can assess in each other most reliably, and reading someone else's plan is the fastest way to see the holes in your own.
10. Two notes on academic honesty
Templates are fine; borrowed numbers are not. Business plan templates are a legitimate professional tool and students should use one. What must be original is every number and every operational decision. A student who downloads a template and fills it with a real restaurant's published figures has completed a formatting exercise.
Generated text is easy to spot in this assignment, and the reason is instructive. Generated business plans are fluent, well-organized, and do not foot. They produce a labor percentage and a labor dollar figure that disagree, cost cards whose lines do not sum, and a break-even computed from a contribution margin that appears nowhere else in the document. Dimension A is, incidentally, an excellent detector — which is worth telling students, because the point is not the detection. The point is that arithmetic consistency across forty decisions is the actual skill, it is the thing generation is worst at, and it is the reason the assignment is built the way it is.
11. A note on grading generously
Most students in this course will never open a restaurant. Of those who do, some will use this document.
Grade the plan the way a lender would read it — for whether it holds together — but write comments the way a mentor would: name the one structural problem, not the twelve symptoms. A student whose labor schedule and labor line disagree by \$70,000 has one problem. Telling them about it once, clearly, with the arithmetic shown, is worth more than eleven marginal annotations.
And when a plan is genuinely good — when the cost cards foot, the schedule produces the labor line, the break-even is computed from the plan's own split, and the student has flagged their own weakest assumption before you found it — say so plainly. That combination is rarer among working operators than it should be.