Chapter 34 — Instructor Notes

What this chapter is actually for

Almost every student arrives at this chapter with a settled frame, and the frame is wrong. They believe financial controls are about catching dishonest employees. They have seen the trade articles, they have possibly worked somewhere that installed cameras after a bad month, and a meaningful fraction of them have been on the receiving end of a badly run suspicion. The teaching problem is not the arithmetic. It is that the chapter's central claim — that a control's product is trustworthy information, and that theft is the seventh explanation rather than the first — sounds to a student like a softness they are being asked to adopt for reasons of manners.

It is not, and the chapter proves it with numbers rather than exhortation. Use the numbers.

The strongest single teaching asset is Bellwether's \$53,122, because it is decomposed. Six lines: \$22,282 of food variance, \$16,169 of bar leaks, \$8,680 of comps over target, \$3,348 of receiving error, \$1,395 of cash loss, \$1,248 of time-clock exposure. Not one of the six requires a dishonest employee to exist. Put that list on the board before you teach anything else and ask the room to guess which line is theft. They will guess wrong, and they will guess wrong in a specific direction: they will assume the biggest line is people. That five-minute exercise buys you the rest of the chapter.

Timing. Three 75-minute sessions:

If you have only one session, teach §34.4 and §34.6 — the daily sales report and the ladder. Do not teach §34.6 without §34.5's thresholds. Without the thresholds, the ladder is an ethics lecture; with them, it is a method, and the difference is entirely whether students leave with something they can do on Monday.

Common misconceptions, in the order students produce them

1. "A control catches a thief." The first one, and it is in the room before you open your mouth. The correction is definitional and should be made immediately: a control produces a record. Give them the five-way ambiguity from the Overview — a compressor that failed on Sunday, a cook who started plating seven ounces in March, a special nobody costed, an invoice that crept eleven cents a pound, and a person taking something — and point out that without records all five look identical on a bank statement. The control does not distinguish them because it is suspicious; it distinguishes them because it wrote things down.

2. "Cash is where the money is." Universally believed and demonstrably false. \$1,395 against \$53,122 — 2.6% of the list — while cash absorbs, by Figure 34.8's estimate, ninety of the roughly one hundred and fifty minutes of weekly control attention. Do not let students conclude that cash controls should therefore be cut; the chapter is explicit that they should not, for reasons that are not financial. That tension is Discussion Guide prompt 7 and it is one of the best conversations in the chapter.

3. Voids and comps get swapped, constantly. Students latch onto voids because there are more of them — 41 against 34 comped items in Bellwether's week — and because "void" sounds serious. Drill the distinction until it is automatic: a void touches no financial statement at all; a comp reduces sales while the product cost stays in cost of goods sold, moving both halves of the food cost ratio the wrong way at once. The diagnostic question that catches it: "Which one is invisible on the profit and loss statement, and why?"

4. "The exception report proves something." Expect an operator-in-training to look at Figure 34.3's eighteen voids and reach a conclusion. Slow it down. Eighteen against a house average of five is a sample of one against a baseline of seven, and the chapter supplies four ordinary answers — new hire, menu design, printer routing, a manager's undocumented instruction — which are right better than four times in five. The sentence to make them repeat: the report generates questions; only people answer them, and most of the answers are boring.

5. Confusing "percentage of ideal usage" with "points of food cost." A genuine arithmetic trap, and it appears every year. Period 8's variance is +6.7% of ideal usage and +2.0 points of food cost. They are different denominators — \$25,488 of ideal usage against \$86,400 of food sales — and a student who mixes them will report a variance three times its true size in the units that matter. Make them state the denominator out loud, every time, the same way Chapter 32 makes them name the revenue base.

6. Treating one period as a pattern. The failure mode of the diligent student. Figure 34.6's rule is crisp and worth memorizing: one period is a data point; two of three in the same direction is a pattern; three periods crossing zero is your counting error plus the volatility of the product. Related and worth five explicit minutes: a negative variance is not good news, it is a recount.

7. "Just install cameras." It will be proposed in every section. The answer is not "cameras are bad"; it is arithmetic. Cameras see the bar. They do not see a cost card that was last updated in April, a special that ran three weeks uncosted, an invoice above the quote sheet, a ribeye at 12.9 ounces, or a keg that foamed out on a Sunday — which is \$997 of Bellwether's \$1,062 and \$1,606 of Case Study 2's \$1,680.

8. Believing the industry theft percentage. Half the room will have encountered the three-to-four percent claim. Case Study 1 dismantles it in three structural steps — wrong denominator, a median loss per case is not a percentage of sales, and above all it cannot be allocated, so it tells you nothing about where to stand in your own walk-in. Teach the last point hardest. An unallocatable number sends an operator looking for a person; a decomposition sends them to four pieces of paper.

The hardest point to teach

That "theft last" is an empirical claim, not a moral one.

Students will assent to it in the seminar room and abandon it on the first case they feel strongly about. The tell is a sentence that begins "I understand the principle, but in the real world…" — and when you hear it, the point has not landed, because the principle is the real-world claim. It is not a concession to fairness. It is a statement about base rates.

The reason it is hard is that the fairness argument is so much easier to state that it crowds the accuracy argument out. Students remember "don't accuse people" and forget "you will be wrong about the cause, and the leak will keep running."

The demonstration that fixes it — the four-document afternoon. Twenty-five minutes, and it is the best use of class time in the chapter.

Hand each pair four documents and one number:

• the number: meat and poultry, +\$1,062 on \$9,850 of ideal usage, Period 8
• Document A — two cost cards, with ground beef and pork shoulder priced at last quarter's numbers against invoices showing about a 6% move
• Document B — a specials log showing a Wednesday pork special that ran three weeks, with no cost card attached
• Document C — an invoice for one case of chicken above the quoted case price, with "credit at door" written in the margin and nothing posted
• Document D — a portion log showing ten weighed ribeyes averaging 12.9 oz against a 12 oz spec, and a receiving note showing birds at 3.8 lb average against a 3.5 lb spec

Give them twenty minutes and ask for a decomposition. Most pairs will find \$997 of the \$1,062 — \$186 + \$214 + \$92 + \$298 + \$207 — and land on a **\$65 residual, 0.66% of ideal usage**, inside a hand count's honest error.

Then ask the two questions that do the work:

1. "How long would the rung-seven version of this investigation have taken?" (Weeks. With a worse answer.)
2. "If you had started there, would you have found any of what is on your page?" (No. And the stale cost card would still be there after the firing — and so would the variance.)

Do not moralize afterward. The exercise has already made the argument, and a moral coda will convert an empirical demonstration back into an ethics lecture, which is precisely the failure you are trying to prevent.

A second demonstration: the threshold-design trap

Ten minutes, and it inoculates against the most common real-world failure.

Ask the room to propose a whole-book variance threshold for Bellwether. Somebody will say a round dollar figure — "anything over a hundred dollars." Take it seriously and run it: on \$25,488 of ideal usage where a hand count carries about \$127** of honest error, a \$100 threshold fires every period, forever, on nothing.** Then run the opposite proposal ("5% of the whole book"), which is \$1,274 and would have let Period 8's \$1,714 through until it was nearly \$2,000.

Then reveal the chapter's two-level structure and the asymmetry, which is the part worth an exam question:

• Category level, greater-of (5% of that category's ideal usage or \$250; liquor and wine 3% or \$150)
• Whole-book level, smaller-of (1.0 point of food cost or \$400)

Ask them why the tie-break points in opposite directions. The answer is cost of a look: the whole-book test is a tripwire whose false alarm costs one afternoon of reading paper you already own, so you want it sensitive; the category test decides which walk-in you physically stand in, so you want it specific. Sensitivity where a look is cheap, specificity where a look is expensive. Every student who can say that sentence has learned §34.5.

Extension if you have time: at what period volume do the two whole-book tests cross? (\$40,000 of period food sales, where 1.0 point equals \$400. Bellwether at \$86,400 is well above it, so the dollar test binds and is equivalent to 0.46 points.) And: does Period 8's 2.0 points trigger escalation? (No — the clause requires 2.0 points sustained across two periods, and Period 8 is one. This one catches almost everybody, including instructors reading quickly.)

The arithmetic behind the un-answered exercises

Exercise 30 has no worked solution in the answers appendix, and it is a good in-class problem, so here are the figures.

Weekly counting costs 2 people × 1.5 hours × \$22 × 52 = **\$3,432. Period-end-only counting — thirteen times a year — costs 2 × 1.5 × \$22 × 13 = **\$858. The saving is \$2,574.

Now put it against detection lag. The measured food variance is \$22,282 a year, which is **\$428.50 a week. Weekly counting costs \$66 a week. So the weekly count pays for itself if it shortens detection by \$66 ÷ \$428.50 = 15.4% of a week — about one day.** It routinely shortens it by two to three weeks. The saving is real and the trade is terrible.

The two sentences to an owner who wants the \$2,574: "Period-end counting means the earliest you can find a problem is four weeks after it starts, and the earliest you can act is six. The weekly count only has to buy you one day of that to pay for itself."

Exercise 31's ranking is also worth having on hand: portion scale (\$3,874 addressed, \$0 spent) · jiggers (\$7,631 for \$233 = 32.7×) · weekly count (\$22,282 for \$3,432) · second pair of eyes (\$1,395 for \$1,184 = 1.2×). The pair at the top costs **\$233 between them and addresses \$11,505 — about 49 times** — which is the number that tends to end the "controls are expensive" objection permanently.

Assessment guidance

The single best exam question is a variant of exercise 22: hand them a daily sales report that fails one of the three ties, contains an over/short near the tolerance boundary, and carries an exception block with a bad un-coded comp rate. It tests the three reconciliations, the tolerance rule, the difference between an arithmetic problem and a control problem, and the posture toward an exception — in one artifact, and it is nearly impossible to answer from memory.

Require the tie-out, always. Sold must equal tendered; expected cash must be derived, not asserted; the processing calculation must show both components. Students who do not tie out produce answers that are wrong in ways neither they nor you can locate.

Require the covers conversion. A leak stated in dollars is half an answer. \$53,122 is 10.5 dinner covers a night and the \$4,849 program is about one — and that pair is the deliverable the chapter actually promises to a floor manager. Make it a scored line item.

Never award full credit to an answer that reaches rung seven without documenting the first six. State this in the rubric before the assignment, not after. It is the chapter's whole argument, and a rubric that does not enforce it is teaching the opposite of the text.

Grade the "what I will not do" list as heavily as the plan. In exercise 35 the three prohibitions — never confront on the floor, never interview alone, never quietly cut hours — carry more of the chapter's value than the ordering of the steps, and they are the part students omit under time pressure.

One thing to watch for and mark down hard: any answer that constructs a scenario in which a named role is the culprit. Case Study 2 is explicit that it establishes nothing about the person who was terminated, and the discipline of not filling in that blank is itself an assessable skill.