OPS.13:11 - SoTA-Echoing

The practice question is how to make service commitments under uncertain demand and constrained resources. The selected best-known line combines explicit work and time boundaries, a forecast appropriate to the claim, and a service decision by the responsible parties. Compared with a plan based on average utilization or an unqualified target, it exposes the answer-changing deficits shown by PumpWorks and the hospital case.

The Kanban Guide 2025.5 describes a service-level expectation through elapsed time and probability, informed by the relevant cycle-time history. Section 4.3 adopts that explicit forecast shape when a probabilistic claim is needed. It also asks whether the observed population and operating conditions fit the new promise. The guide’s workflow forecast does not itself provide resource authority or customer agreement.

The SRE Workbook’s service-objective guidance (2018) starts from user-relevant service behavior and distinguishes the intended indicator from how it is measured. Sections 4.1 and 4.2 adapt that distinction to the promised result, eligibility and acceptance evidence. This is a useful alternative to a resource-only account, while a software-service indicator remains insufficient for a clinical or engineering acceptance decision.

Steve Tendon, The Book of TameFlow (2022), printed pp. 215-220 and 259-269, connects reserve consumption, progress and management attention. Adopt its use of changing conditions to focus intervention, including information that arrives before a formal alarm. Section 4.5 adds the time and resource construction that tests whether a response can preserve the promise and exposes the need for another remedy or a revised agreement. The book’s heuristic conversion from a duration fraction to a delivery probability (p. 212), and its use of mean flow time as the 50th-percentile cutoff (p. 220), do not establish those probabilities. A probabilistic promise needs the applicable distribution; an alarm needs a workable response. Retain periodic meetings where they perform another needed function. Reconsider the observation policy when detection, authority, remedy duration or resource availability changes.

For a small known job set, a feasible sequence and explicit adverse case can answer the commitment question at lower effort than a stochastic model. PumpWorks uses that branch. A changed demand distribution, recurring service-level claim or consequential tail risk can justify the stronger forecasting work. Reopen the basis when work mix, windows, resources, acceptance or the promised result changes.