Ch34 Discussion

Discussion Guide

Prompt 1 — "Controls impose a cost on trust. Is a clipboard an accusation?"

What to listen for. The chapter refuses to resolve this cheaply and neither should the room. The weak positions are at both ends: "controls are just good business, feelings are irrelevant" and "if you trusted your people you wouldn't need to count." Push past both to §34.8's actual resolution, which is a claim about the counterfactual: in a restaurant with no controls, an unexplained loss lands on whoever the owner happens to suspect — so the choice is never "controls or trust," it is "a clipboard or an unfalsifiable suspicion." Listen for students who reach the three concrete formulations on their own: a manager who counts with a witness cannot be accused of a shortage; a bartender counted on the same rotation has evidence their variance is normal; a cook whose portioning is measured can demonstrate the problem is purchasing. The strongest contributions notice that this only works if the controls are universal, routine, and explained — and that an owner who counts only the bar, only after a bad week, and never mentions the result has built a signal rather than a control.

Prompt 2 — "The industry says restaurants lose three to four percent of sales to theft. This book says that number cannot be sourced. Who is right, and what should you do while you are unsure?"

What to listen for. Start with epistemics, end with practice. The productive path runs through Case Study 1's three structural objections — the sample is detected-and-investigated fraud in organizations that employed a fraud examiner; a median loss per case cannot be converted into a percentage of your sales; and above all the number cannot be allocated. Push the room on that third point, because it is the one with operational consequences: an unallocatable number sends you looking for a person, while a decomposition sends you to four documents. Watch for the false symmetry — "so we just don't know anything" — which is wrong; we know the six lines Bellwether measured, and every one of them is measurable in any restaurant in an afternoon. The best answers land on: refuse the aggregate, build the decomposition, and treat anyone who leads with the aggregate as having told you something about their evidence standards.

Prompt 3 — "Case Study 2's operator counted the bar to the tenth every period, noticed the trend by Period 4, and acted. They did more than most operators do — and got it badly wrong. What separates diligence from method?"

What to listen for. This is the chapter's deepest question and the room usually starts in the wrong place, either faulting the operator's character or concluding that measurement is futile. Redirect to sequence. Everything the operator did was correct except the order: they collected excellent data and read it in the one sequence guaranteed to contaminate it. Listen for the observation that measurement and control are different things, and that this restaurant had the first without the second. The strongest contributions notice the falsifiability failure — the variance did not move after the termination and that was not treated as information for two more periods — and connect it to the general principle that an investigation without a stopping rule does not end, it just runs out of people. If the room is strong, ask what the operator's routines would have had to look like for the right thing to happen automatically, rather than depending on somebody having better judgment on a bad Tuesday.

Prompt 4 — "Bellwether has four managers and two of them own the place. Who audits the owners?"

What to listen for. The chapter puts this in the open questions carried forward and does not answer it, so the room is genuinely on its own. Good answers converge on the outside bookkeeper as the only structurally independent party and then notice the limit immediately: the bookkeeper works for the owners. Listen for the practical substitutes — cross-review between the two partners, which is the reason partnerships in this business survive; mandatory absence, which is the only test of whether a process is a process or a person; and the fact that a control everyone knows is read has most of its value before anyone reads it. Push on the scaling question, since Chapter 35 is next: what changes at two units, when the owner is no longer in the building most nights? The best contributions notice that "who audits the owners" is a rhetorical question at one unit and a real one at two — which is exactly why the chapter hands it forward rather than closing it.

Prompt 5 — "\$76 of Friday's \$187 in comps bought back tickets that ran long. Is that a controls failure?"

What to listen for. The single most common misdiagnosis in this part of the business, and the answer is no — it is a kitchen throughput problem being paid for out of the comp line, and the fix is in Chapter 14's ticket-time work and Chapter 19's staffing guide. Listen for the crucial second step: a comp policy tightened in response to that number would make the guest experience worse without recovering a dollar, because the underlying tickets would still run long; you would simply have stopped paying to fix them. Then push the room onto the line that is a control problem — the single \$16 comp with no reason code, which is 8.6% of the night's comp dollars made permanently un-reviewable. The strongest answers state the general rule: measure the reason-code failure rate as a percentage rather than in dollars, because the dollars are always trivial at the moment you could still fix it and never trivial by the time you notice. If the room is quick, connect it to Chapter 23: the comp is the cheapest retention instrument in the building, and a control that makes hospitality slower has to justify itself against the revenue it costs.

Prompt 6 — "You have worked all six rungs and something still does not explain itself. Walk the first forty-eight hours."

What to listen for. Run this as a live sequencing exercise, not an essay. The three first moves are document before you speak (memory reorganizes itself around a suspicion with remarkable speed), preserve the records (copies, not originals; noted, dated; the system's retention limit is now your deadline), and call an employment attorney before the first interview rather than after the termination. Listen for who gets told — the smallest possible number, and nobody on the floor — and for the three prohibitions. Expect at least one student to propose confronting someone directly "to get it over with"; that is the moment to make the room articulate what an accusation costs when it has to be withdrawn. Reserve the last five minutes for the stopping rule, which is the part almost every group omits: write down the innocent explanations in advance, rule each out on evidence, and commit to a falsifiable prediction — if this cause is real, the variance falls next period. Note that legal specifics vary enormously by state, county, and city, and that this is a place for counsel rather than instinct.

Prompt 7 — "Cash is 2.6% of the exposure and gets more than half the attention. Defend keeping the nightly two-person count."

What to listen for. A deliberately awkward closer, because on dollars alone the second pair of eyes is the worst control on the program — \$1,184 spent against \$1,395 of measured loss, a ratio of about 1.2×, and that generously assumes a witness prevents every dollar of it. Watch for students who follow the arithmetic straight off a cliff and cut it. Then surface the three costs that never appear in a dollars-per-dollar ranking: personal safety (the chapter has watched an operator lose a manager to a parking-lot mugging, not to theft); staff protection (a server who carried a drawer and left without a witnessed count has no defense if the number is wrong tomorrow — say that out loud when you introduce the rule and it stops feeling like suspicion within a week); and the credibility of every other control (a business that visibly stops counting \$612 has announced how seriously it intends to count \$25,488 of walk-in). The best answers arrive at the general principle: judge a control on everything it does at once, not on its stated purpose — the same lesson exercise 29 teaches about the receiving window, where the twenty minutes is simultaneously an invoice control, a food-safety check, and a spec enforcement mechanism. Close by asking what the ranking was good for. The honest answer is that it correctly identifies the scarce resource — and the scarce resource in Figure 34.8 is not money, it is the forty-five minutes on Monday.