It is 1954, and a factory's production line runs on target for now — but somewhere ahead a tool will wear, a setting will slip, and the output will drift out of spec, and you must detect that shift as soon as it happens without crying wolf while all is well. A fixed test on each item is too slow or too jumpy; instead accumulate the running evidence of a shift so small persistent deviations build up and trigger an alarm quickly, while random noise cancels out. Design the rule and its trade-off: how fast it catches a real shift against how often it false-alarms on a stable process. Get it wrong and defective product ships for hours undetected, or the line is stopped constantly for phantom shifts.
Chose Bayesian computation and model checking — right call.
Gelman's applied Bayesian workflow, developed from the 1990s onward, engages model monitoring and posterior predictive checking as recurring concerns — detecting when a fitted model's assumptions have stopped matching incoming data shares real conceptual kinship with this problem's demand for detecting when a production process has drifted from its target regime. His hierarchical modeling practice regularly requires deciding whether recent data still look consistent with an established pattern. But his primary contributions target Bayesian model-building and criticism for scientific and social-science applications rather than the classical sequential cumulative-sum industrial change-detection procedure this 1954 factory scenario specifically requires, keeping his relevance here real but somewhat removed from the classical framing that Wald and Page solved decades earlier.
Bellman's dynamic programming, developed from the early 1950s at RAND, gave mathematics its general framework for sequential decision-making under uncertainty, breaking a multi-stage problem into a sequence of smaller decisions each conditioned on the current state — precisely the structure a change-detection rule embodies, deciding at every new observation whether the evidence accumulated so far justifies stopping and raising an alarm. His work on optimal stopping problems, closely related to his broader dynamic programming framework, engages the exact trade-off this problem poses between the cost of waiting too long and the cost of stopping too early. He does not rank above Wald, Box, or Bertsekas because his own major contributions target the general theory of sequential optimization rather than the specific statistical process-control application this 1954 factory scenario requires.
Battle #23 · 8/9/2026, 5:16:14 PM · this result is deterministic: the same two personas on this problem always resolve the same way.