LEARN. PRACTICE. GROW.Funding a decision in stages
An initiative estimated at €120k depends on customers accepting a format. A €12k trial can observe acceptance before building every integration. Funding it limits initial exposure and enables a better-informed later decision. It does not turn the €108k not yet authorized into realized savings or guarantee a return. Define what the trial can clarify and which decisions depend on it. Include options to continue, change, or stop. If the result cannot change any possible choice, reconsider the trial’s usefulness. Budget and its authority are explicit contextual conditions; Scrum does not automatically assign every financial power to the PO.
Recovered capacity and achieved outcomes
Consolidating variants can reduce maintenance effort and create room to respond to needs. If maintenance falls from 60 to 40 days within a constant monthly total of 100, capacity outside that activity rises from 40 to 60 days. That is 20 additional days, not doubling and not proof of customer value. Record how that capacity is used and its effect on the ability to learn or deliver. Recovery may need improvement, interruptions may need reduction, or an uncertain need may warrant exploration. Avoid automatically demanding more features in return. Measurement should support sustainable choices rather than hide the operational work that keeps the product usable.
Unit costs and interpretation limits
Lower total spending can accompany worse performance per useful outcome. In the example, €480 for 80 correctly completed tasks equals €6 per task. Later, €360 for 45 tasks equals €8. Total spending fell, but this ratio worsened. Confirm that costs, period, and the definition of a correct task are comparable. If case mix shifted toward harder cases, investigate before attributing the change to the product. With zero correct tasks, the ratio cannot be calculated; do not present it as zero cost. The final choice may consider other dimensions, such as coverage or resilience, but should keep this consequence explicit.
Retirement with verifiable continuity
An application with no August logins may remain essential to quarterly close. Look for automated processes, external consumers, and seasonal usage before inferring no need. In an exercise with a fictional continuity and historical-lookup policy, acceptance requires demonstrating the task on the successor and retrieving records with approved permissions. A data copy or a new server does not demonstrate these capabilities. Define who accepts, which cases were rehearsed, and what remains missing. Retain coexistence costs while dependencies persist. The team can recommend staged exit or defer retirement, with criteria for revisiting the decision. These examples do not establish any bank’s internal rules.
Stopping conditions and decision rights
Before a pilot, make the desired benefit and conditions preventing expansion explicit. If the agreement requires no incorrect reconciliations and two occur, a better mean processing time does not permit ignoring that boundary. Suspend expansion, investigate, and take any proposed change to those with appropriate authority. Do not exclude inconvenient cases to confirm the hypothesis. The PO can recommend reallocating funding based on learning while a committee retains the financial decision. Bring options and consequences to that committee. Also challenge incentives rewarding only delivered scope and penalizing justified abandonment: these make it difficult to act on evidence that contradicts the initial plan.
Transition and portfolio decision workshop
Prepare a record with three options: maintain, migrate in stages, or retire after acceptance. For each option, identify required capacity, future costs, consumers, missing evidence, and decision authority. One participant represents the quarterly process; another challenges the savings calculation; another asks for pilot stopping criteria. If training and interface changed together, distinguish observed improvement from isolated cause. End the workshop with a conditional recommendation and an action reducing the greatest uncertainty, such as rehearsing historical lookup on the successor. The reviewer should be able to reconstruct the reasoning and identify new evidence that would justify a different decision without relying on an opaque score.
# Fictional retirement decision record
dependent_process: quarterly close
continuity_evidence: accepted successor rehearsal
history_evidence: authorized lookup tested
funding_authority: portfolio committee
saving_starts: after agreed retirement
missing_evidence: prevents unconditional recommendationAn application with no logins in one month supports a quarterly close. The team identifies the consumer, rehearses the successor, and confirms historical lookup before proposing infrastructure retirement.
Common pitfalls
Counting released effort as delivered value; forgetting seasonal use; excluding pilot failures; measuring only total spending; treating product accountability as unlimited financial authority.
Related topics: Discovery, hypotheses, and experiments · EBM and measurement interpretation · Portfolio, investment, and the product operating model
A well-prepared exit retains required capabilities and makes it verifiable when costs and dependencies disappear. Learning includes being able to stop.
Reference: The Evidence-Based Management Guide · PSPO II; Scrum Guide November 2020 and EBM Guide May 2024; no public numbered exam revision