LEARN. PRACTICE. GROW.Start from all necessary work
Installing components is only part of an operational delivery. Break down deliverables and work needed to validate, train, transition, and close, including agreed recovery requirements. Each package needs an understood completion condition and responsibility. Also represent external dependencies, such as a firewall rule, with required timing and validation evidence. Writing network in parallel is insufficient. Where knowledge remains limited, detail near-term work and retain assumptions, ranges, and milestones for the remainder. Update the plan using prototype evidence and decisions without presenting nonexistent precision.
Calculate capacity before promising duration
Person-days measure effort; elapsed working days measure duration. Four people over twenty days, with twenty-five percent reserved for the project, provide twenty person-days before other losses. RUN capacity does not become available merely by appearing in the project organization chart. Two technically independent activities can also compete for one specialist. If each requires four full-time days from that person, they need eight days in total without a substitute. Show that constraint and compare authorizable alternatives: change sequence, reduce optional scope, obtain qualified capacity, or revisit the date.
Read the network and assess compression
In a simple network with two-day A, six-day B and C in parallel, and two-day D after both, duration is ten days. Shortening only B to four does not change the date because C still constrains D. To shorten the project, identify paths driving completion and recalculate after each change. Consumed float reduces future margin even when finish is not immediately delayed. Adding resources has cost and productivity limits; overlapping work can create rework. Neither technique removes the need to assess assumptions, dependencies, quality, and change authority.
Forecast cost through the end of change
A simple forecast can add actual cost to an updated estimate to complete. If forty-eight thousand euros have been spent and thirty-two thousand remain, forecast final cost is eighty thousand under the assumptions used. Include parallel operation, licenses, retirement, and closure when they belong to delivery. Do not confuse remaining cost with total cost or a forecast with authorized budget. If forecast cloud usage rises, communicate causes, uncertainty, and the gap against limits while options remain. Expected reduction becomes observed savings only when conditions and measurement support it.
Choose metrics that do not hide differences
A paid invoice may reflect an advance without establishing work acceptance. Keep funding, execution, and outcome dimensions separate. In earned value, CPI compares EV with AC and SPI compares EV with PV; monetary planned-value variance is not automatically a number of delay days. For adaptive teams, story points using local scales do not form a common unit for ranking productivity across teams. Use decision-appropriate metrics, explain denominators, and compare equivalent conditions. Update actual progress and remaining work without changing historical dates to retain a green status.
Connect quality, acceptance, and value without conflating them
Plan quality criteria and samples representing agreed conditions. Zero defects in normal operations does not establish behavior for excluded exceptions. In Scrum, merged code failing the Definition of Done does not become part of the completed Increment by moving a card. Compare prevention investment and rework over the same horizon: four hundred euros of preparation to avoid nine hundred of forecast corrections produces five hundred expected difference. After delivery, measure benefit using the agreed unit, such as manual minutes per thousand operations. Technical acceptance, usage, and realized benefit need related but distinct evidence.
Planning exercise: preparation 3 days, two configurations of 4 days each, and validation 2 days. With the same full-time specialist on both configurations, the minimum is 13 days; with two qualified available people, the logical model permits 9. Write the conditions making the second option a credible forecast and the decision needed to obtain that capacity.
Common pitfalls
Confusing effort and duration; overlapping one person’s work; compressing one branch without recalculating others; omitting retirement; treating payment as progress; comparing story points without a common unit; counting only convenient samples.
Related topics: Lead without becoming the bottleneck · Plan with real capacity · Risk, change, and informed decisions · Measure value and hand over autonomy
A credible plan connects scope, resources, dependencies, cost, and delivery evidence. Each forecast retains the assumptions making it executable.
Reference: Schedule Assessment Guide · PMP ECO July 2026; DR PMP 2026.5