A schedule baseline is a locked snapshot of a project’s planned start and finish dates, taken at a point everyone agreed to it, that the live schedule is measured against as the job runs — the reference line ‘ahead’ and ‘behind’ are measured from.
How that snapshot survives inside project software is worth knowing before relying on it. Procore’s own schedule tool doesn’t set a baseline itself — it imports one from the source scheduling file. Integrating a Primavera P6 schedule, Procore’s documentation is explicit that “your schedule cannot be changed directly in Procore. Any changes to your P6 schedule must be performed using Oracle’s Primavera P6 Professional Project Management software”, and that the imported baseline is singular: “Only one Baseline Start is supported in Procore”, and the same for Baseline Finish. There’s no version history sitting alongside it — whatever baseline is uploaded is the one shown, until someone uploads a new one.
That single-slot design is exactly why re-baselining has to be a decision, not a habit. Overwrite the baseline every time the schedule slips and there’s no longer a fixed line to measure real slippage against — the same documentation gap that undermines a delay claim later, since proving a delay means showing where the job actually stood against where it was supposed to stand. A re-baseline that’s agreed, dated and communicated preserves the old snapshot as the record of what changed and why; a silent one just makes the job look like it was never behind.
A 40-week structural steel package slips six weeks after a fabricator’s shop closure — a cause outside the contractor’s control. Rather than quietly re-uploading a schedule that shows the new dates as if they were always the plan, the scheduler exports the original baseline, documents the fabricator delay against it, and only then uploads a revised baseline reflecting the recovery plan — so the six-week gap is preserved as evidence, not erased by the update that fixed it.
See also: The three-week lookahead, and its point · Critical path and float, explained · Delay claims: what has to be proven.
A 30-minute call is enough to tell you whether AI pays for itself here.
