Most completion surprises do not start with a dramatic delay event. They start with a weak update. Progress is collected against one week, cost against another, remaining durations are “tidied” to protect a milestone, and the data date is treated as a label rather than a hard cut-off. The file still calculates. The finish date still looks precise. It is no longer comparable to last period.
Data date discipline is the boring control that keeps forecasts honest. Without it, every other controls product—SPI, CPI, critical path, board pack—drifts.
What the data date actually means
The data date is the line between known and forecast. Everything before it should be progressed. Everything after it should be remaining work with current durations, logic and calendars. If that line moves around the room from week to week, the model of the future moves with it.
In Primavera P6 and MS Project, the software will calculate from whatever cut-off you give it. It will not refuse a soft process. Discipline is a team rule, not a button.
The update errors that quietly move completion
- Progress collected on Thursday, but the data date set to Friday or Sunday to “look current.”
- Remaining duration shortened to hold a milestone without a method change the site owns.
- Out-of-sequence work ignored, so successors stay driven by incomplete predecessors.
- Actuals and narrative closed on a different week from the schedule update.
- Activities marked complete while punch lists, tests or documentation still sit on the path.
Each error can look small. Together they move float, rewrite the critical path and shift completion without a clear event in the narrative.
Why the finish date moves without a story
When remaining durations are optimistic, the network says the work finishes sooner. When incomplete work is marked done, successors unlock early in the software and late on site. When the cut-off is fuzzy, one package reports a newer world than another. The calculated completion then changes for process reasons, not site reasons.
Directors see a date move and ask what happened. The honest answer—“the update rules moved”—rarely appears in the pack. So the room invents a delivery story for a controls failure.
| Weak practice | What the file does | What leadership thinks |
|---|---|---|
| Soft data date | Mixes two weeks of progress | The programme “caught up” |
| Optimistic remaining duration | Float appears; finish improves | Productivity recovered |
| False completes | Successors unlock early | Path is clear |
| Late progress dump | Big percent jump in one update | A surge week happened |
| Narrative after the file | Slide contradicts the network | Controls are confused, not the site |
A disciplined update sequence
Hold the cut-off first. Agree the data date before progress is collected. Collect facts against that date only. Apply the progress rules already written for the activity type. Review remaining duration with the person who will finish the work. Recalculate. Then draft the narrative and the indices.
Do not reverse the order. A pack written before the file is closed will force the schedule to match the story.
Rules that keep completion comparable
- One data date for schedule, earned value and the leadership pack.
- No remaining-duration cut whose only purpose is to protect a date.
- No complete status without the done criteria the baseline assumed.
- Out-of-sequence work recorded and explained, not silently overridden.
- A saved period file so next week can be compared, not overwritten into oblivion.
What directors should ask every period
Ask three questions. What was the data date? Was every package measured against it? Did any remaining duration change without a method or scope reason? If those answers are vague, do not argue about the finish date yet. Fix the update.
Also refuse a recovery story that starts with “we re-forecasted” and never states whether the cut-off and progress rules stayed the same.
How this links to earned value
EV is only as honest as the progress measure and the cut-off. A soft data date makes SPI and CPI jump for reporting reasons. That is why data date discipline belongs with consistent progress rules, not as a separate admin habit.
A practical standard
Treat the data date as a contract with yourself. Everything before it is evidence. Everything after it is forecast. If the evidence is mixed, the forecast is theatre.
EVM Solutions runs schedule updates and project controls with a fixed cut-off, named progress rules and a narrative written after the file calculates. If your completion date moves every period without a clear site event, send the last two update files and the stated data dates. The quiet error is usually in the process, not in the critical path story the room prefers.
