Stop 2 of 9 · Weekly
Reading the P6 export
The columns that make an export readable, the longest path and the near-critical band, and the questions only P6 can answer.
The export, not the file
P6 hands you an XER, which is a text dump of database tables written for P6 to read back. Claude can open one, but the useful move takes ten seconds longer and pays for itself: export the activity table and the relationship table to Excel with the columns that carry meaning.
| Column | What it answers |
|---|---|
| Activity ID, name, WBS | Which scope, which area, which contract line |
| Original duration, remaining duration, percent complete type | Whether reported progress is measured or asserted |
| Early and late dates, actual start and finish | What is planned and what has happened |
| Total float and free float | How much room there is, and whose room it is |
| Constraint type and date | Whether a date was calculated or forced |
| Calendar | Whether fourteen days means fourteen days |
| Predecessors and successors with relationship type and lag | The logic that makes the path a path |
Without the last two rows Karim is reading a to-do list. With them he is reading a schedule.
The longest path is not the float zero list
On a job with constraints and multiple calendars, the set of activities at zero total float and the longest path are different sets, and the difference is where arguments start. Ask for both, and ask for the near-critical band underneath: on a 940 day contract, anything under ten working days of total float is one bad week from being the path.
[upload the activity export and the relationship export for the current update] Read these as a schedule, not a list. Give me, in this order: the longest path as flagged in the export, then activities at zero or negative total float that are not on it; the near-critical band under ten working days of total float grouped by WBS; every activity carrying a hard constraint with its type and date; and the ten activities whose total float dropped most against the column I have marked prior float. State the number of activities the export contains and how many are complete, so I know what you read.
[upload the activity and relationship exports] Run a data quality pass on this update and give me a count and a list for each: activities with no predecessor, activities with no successor, relationships with a lag over five days, relationships with a negative lag, start to finish relationships, hard constraints, activities with an actual start after the data date, activities showing progress whose predecessor has not started, and remaining durations equal to original duration on activities reported as started. Rank the findings by how much they would change a critical path reading.
From the two lists above, write the questions I send to the contractor's scheduler. One line each, activity ID first, what I observed, and what I am asking for. No accusations and no conclusions about the effect. Order them so the ones that block my review come first.