Needs and measurement boundaries | PMI-PBA: needs, requirements, and benefits

Separate the need from the suggested solution

In a fictional funds application, the sponsor requests more servers because requests take too long to finish. APS shows a chart with six minutes of average processing. Before accepting the solution, ask what the user expects, when waiting begins and which outcome marks completion. The problem may lie in a queue, manual authorization, handover between teams or execution itself. You need not reject servers as an option; you need evidence connecting that option to the need. Define the outcome in observable terms, such as eligible requests completed within an agreed window, identifying who experiences delay and its consequence for close.

Use boundaries that explain the journey

The lab contains five completed requests. Processing times are 4, 5, 6, 7 and 8 minutes, but preceding waits are 10, 15, 20, 25 and 30. Mean execution is six minutes; mean wait is twenty and mean end-to-end duration is twenty-six. The means can be added because they use the same five requests and non-overlapping phases. If each team uses a different population, that sum no longer has the same justification. In actual work, confirm identifiers, events, units and clocks before subtracting timestamps. A negative duration should trigger investigation of data or semantics rather than silent conversion to zero.

Retain requests that did not finish

Besides the five completed requests, one failed, one is pending and one has unknown status. The agreed indicator uses all eight eligible requests as denominator: five out of eight is 62.5%. If the report includes only six terminal requests, it presents five out of six, about 83.3%, for a different question. If it shows completed requests only, it can appear 100% successful by construction. None of those slices should use the original indicator’s name without explaining the change. Unknown status is neither zero minutes nor confirmed completion. Known-status coverage is seven out of eight, or 87.5%, and should remain visible as an evidence limitation.

Choose the next observation

Timings show where waiting accumulates but do not prove its cause. To distinguish capacity shortage, an external dependency and manual control, observe arrival, start, rerouting and completion using authorized examples. Include normal periods and close, different teams and relevant exceptions. If the team collects only successful low-volume requests, the dataset may not address the need. Combine data with interviews and observation of work, retaining disagreements still needing confirmation. A recommendation may start by measuring an unobserved phase before investing in a change. Explain which decision that collection will support and when enough information will be available to proceed.

Write a definition that can be reviewed

For each metric, record the question, eligible population, numerator, denominator, start and end, exclusions, unit, source, frequency and owner. Distinguish an attempt from a business operation: a retry can generate two technical rows for one movement. Identity policy should match the indicator’s purpose. In the exercise, the nearest-rank 95th percentile of five completed requests is 38 minutes, the largest observed value. Such a small sample does not stably estimate the production tail. Retain the calculation definition and limitations with its result. A committee should understand what the number answers and what still needs investigation before approving the suggested solution.

python3 content/labs/pmi-pba-benefits-evidence/run.py --output /tmp/pba-benefits.json
# Compare processing, waiting, elapsed time and eligible completion.
# Completed-only timings do not describe unresolved requests.

IN PRACTICE

Five requests have six-minute mean execution but twenty-six-minute mean end-to-end duration. Eight eligible requests produce 62.5% success.

Common pitfalls

Completed-only mean as everyone’s experience; failures excluded without notice; unknown as zero; queue time as an already demonstrated cause.

Related topics: Needs and business case · Metrics and acceptance criteria · Operations and benefit realization

Take this idea with you

Define the question and population before comparing numbers or recommending a solution.

Create account

Reference: PMI-PBA Examination Content Outline · Five-domain ECO / verified 2026-10-01