A common objection to the principle: I know operators who defaulted on almost everything and got lucky. They ran a hot concept, hit a market wave, made money, cashed out. And I know operators who designed carefully and got killed — the neighborhood shifted, the pandemic hit, the anchor tenant closed, they lost the lease. If designed operators can fail and defaulted operators can succeed, then the principle does not predict outcomes reliably. It is one factor among many, not the causal law it claims to be.

The objection is real and needs a direct answer. The principle does not claim certainty per event. It claims probability over time. And the distinction between event and time is what makes the principle a governing law rather than a fortune teller.

At the level of a single event, in a single shift, at a single restaurant, many things can produce outcomes that appear to defy the principle. A defaulted operator can catch a wave and profit for a year. A designed operator can hit a shock that no amount of design could have absorbed. Individual cases have variance. That variance is real, and the principle does not deny it.

At the level of many events, over multiple years, holding operator capacity constant, the variance averages out and the principle emerges. Two operators of matched capacity, one running by design and one running by default, over three to five years, will produce different outcomes on the dimensions they were designing or defaulting on. The designed operator will retain more Guests, retain more cast, hold more margin, and survive more shocks. Not in every shift. Not every week. But at the aggregate, and over time, and with statistical reliability.

This is what every real predictive law does. Statistical mechanics does not predict which specific particle will be where. It predicts what a population of particles will do at scale. Epidemiology does not predict which specific person will catch a disease. It predicts what a population will experience. Financial markets do not predict which specific stock will move on which specific day. They predict what portfolios will do over long time horizons.

The Summers Principle operates the same way. It does not predict the outcome of a specific Guest visit, a specific hire, a specific shift. It predicts what a designed operation will produce, at the aggregate, over multi-year time horizons, holding capacity constant. And it makes that prediction with the same reliability that other statistical laws make their predictions inside their domains.

The operator who ran by default and got lucky is not evidence against the principle. The operator’s outcome is evidence of variance inside the population of defaulted operators. That population has a wide distribution of outcomes — some catastrophic, some catastrophic-delayed, some briefly successful, some rare long-term survivors. The average outcome of the defaulted population is failure over time. The designed population has a much tighter distribution, and its average outcome is survival with growing scale. The individual defaulted operator who profited briefly does not disprove the principle. They occupy the tail of a distribution that mostly did not profit.

The operator who designed carefully and got killed by a shock is not evidence against the principle either. Shocks arrive. Some are large enough that no reasonable design at the individual restaurant scale could have absorbed them. The principle does not claim to protect against catastrophe. It claims that designed operations survive more shocks than defaulted ones, on average, over time. An individual designed operator who got killed by an unusually large shock is a data point inside a distribution — the distribution just has fewer catastrophic outcomes than the defaulted distribution does. The average designed outcome is still much better than the average defaulted outcome, even when catastrophic outliers exist.

This is why the principle is falsifiable in a specific, testable way. The falsifiable claim: over multi-year time horizons, holding operator capacity constant, designed operations outperform defaulted operations on the outcomes they designed for. Not in every case. Not in every shift. But at scale and over time.

If matched operations were tracked over multiple years and the defaulted operations retained Guests better than the designed operations, the principle would be wrong. If defaulted operations held margin better, or retained cast better, or survived shocks better, the principle would be wrong. None of this has happened in the data. The principle has held everywhere it has been tested at the population scale, over sufficient time.

The claim the principle makes is not do design and win every game. The claim is do design and win most games, and win the game that matters most — the multi-year survival and growth game. Default and win a game occasionally, but lose the long game with high probability.

An operator who understands this can absorb short-term variance without losing conviction in the principle. Bad shifts happen inside designed operations. Good months happen inside defaulted operations. Neither disproves the principle. What proves the principle is what the operator’s operation looks like three years from now, five years from now, ten years from now — and what the aggregate industry data shows for designed versus defaulted populations at those horizons.

The operator running for the long game runs by design. The operator running for the next quarter can sometimes get away with default. This is not a matter of morality or effort. It is a matter of what game the operator has decided to play. And the principle asks the operator to know which game they are playing before they choose their operating mode.