Most unreliable forecasts do not come from bad judgment. They come from small update errors that compound. A stale data date. An activity that started before its predecessor finished. A constraint added two years ago that nobody removed. These ten checks catch most of them before the update goes out.
The ten checks
- Data date is current. It matches the reporting period, and the schedule was recalculated on it.
- No actuals in the future. Actual starts and finishes never fall after the data date.
- No forecast in the past. Unfinished work is never scheduled to happen before the data date.
- Out-of-sequence progress is resolved. Activities that started before their predecessors finished have logic corrected or documented.
- Remaining durations are reviewed. They reflect the work left, not the original duration minus elapsed time.
- Logic changes are logged. Every added, deleted or changed relationship is listed with a reason.
- Constraints are justified. Every constraint maps to a contract date or a real outside commitment.
- Calendars are correct. Holidays, shift patterns and weather days match how the site is working.
- The baseline is untouched. Comparisons run against the approved baseline, not a quietly refreshed one.
- The narrative matches the file. Every statement in the report can be traced to the XER.
Why the log matters most
An update with a clean change log can be audited in an hour. One without it can take weeks to rebuild, and in a dispute that gap usually favors the other side. Keep the log every month, even when nothing seems to change.
Practice on a sample schedule in Schedule Studio, or ask about Primavera P6 scheduling and progress reporting that follow these checks every period.
