Glossary of scheduling and delay-analysis terms

Short definitions, in alphabetical order, of the terms used across this site and in the reports the tool produces. Each entry says what the word means and, where the idea is genuinely contested, points at the page that treats it properly.

Several terms below have no single agreed definition, and the entry says so where that is true. That is not evasion: a glossary handing you one definition of "critical path" has hidden the disagreement most delay arguments turn on. Concepts is the long form of every contested entry here.

A

Activity. A discrete element of work carrying a duration, a calendar assignment, and a position in the network of relationships. In a Primavera export it is a row in the TASK table. It is the unit almost every specification clause quantifies over, so what counts as one — and what is a milestone, a summary or a span instead — decides whether a clause fires. See the XER format.

Actual start, actual finish. Recorded dates on which work began and ended, as distinct from computed early and late dates. An actual date is a statement about history, and whether a date recorded on the data date is history depends on the data date convention — see Data date below and Concepts §3.

ALAP (as late as possible). An activity placement that schedules work as late as the network allows rather than as early. It is a scheduling mode rather than a date constraint, and it changes which float figure a report should quote, because an ALAP activity's published early dates are derived from its late dates.

As-built schedule. The record of what actually happened. Most as-built records are collections of dates rather than networks: they carry no logic, no durations that recompute, and nothing to reschedule. That distinction decides whether a subtractive method can be run at all — see the method comparison.

As-planned schedule. The programme as it stood before the events under analysis, usually the baseline. Additive methods impact it; observational methods compare against it.

B

Baseline. The schedule adopted as the plan of record, against which progress and variance are measured. Which submission is the baseline is normally a contract question rather than a technical one, and several specifications regulate its content tightly — see UFGS scheduling clauses and state DOT scheduling specifications.

C

Calendar. The definition of which days and hours are worked, including holiday and shutdown exceptions. Durations, lags and float are all counted on a calendar, so "five days" is not a fixed quantity, and on a network using several calendars day counts taken on different calendars must not be added together. See calendars in a P6 export.

Collapsed as-built. A modelled, subtractive method: remove the delay activities from an as-built network and reschedule to ask when the work would have finished but for them. Also called but-for analysis. The method comparison sets out what it needs and where it fails; Methods treats it method by method.

Concurrency. Two or more delays running such that both bear on the completion date over the same span. There is no single test for it: the authorities differ both on what "at the same time" means and on which of two overlapping delays is the effective cause. Concepts §5 sets out the two axes and the four findings they produce. This tool reports the arithmetic and, under US jurisdictions, returns no governing answer.

Constraint. A date imposed on an activity independently of its logic. Soft constraints act as a floor on the forward pass or a cap on the backward pass and leave the rest of the arithmetic intact — start no earlier than, finish no later than and their counterparts. Hard constraints fix a date whether or not the logic supports it, so they can override the network and distort float. "Hard constraint" as reported by schedule-quality checkers is a reporting class wider than the set that actually overrides the calculation. See the DCMA 14-point checks.

Critical activity. An activity identified as critical by whichever operational test the schedule or the specification uses. There is no single test: AACE RP 49R-06 documents four operational methods for identifying the critical path and declines to endorse one of them. The two in common use are a total float threshold and membership of the longest path, and they disagree on real files. See Concepts §2.

Critical path. The work controlling the completion date — a definition that sounds settled and is not. AACE RP 10S-90 separated "critical path" and "longest path" into distinct entries in its December 2023 revision, and specifications pick sides: a federal section can require criticality to be defined as longest path rather than by float. A report quoting a critical path without naming the sense it used has not answered the question.

D

Data date. The boundary between what is recorded and what is planned. Float and criticality are defined forward of it. Its convention is genuinely unresolved: a data date of 1 February can mean work is reported through 1 February, or to 1 February with the period ending on 31 January. Both readings produce a schedule that looks right, and they place a day's delay in different windows — often windows owned by different parties. See Concepts §3.

Driving relationship. A relationship that actually sets its successor's early dates, as opposed to one that merely permits them. Where two predecessors tie at the same bound, both drive, which is why the set of driving relationships is generally wider than any single driving chain through the network.

Duration. The working time an activity takes, counted on its own calendar. Three figures are commonly distinguished: original duration, remaining duration and at-completion duration. Most duration-cap clauses in specifications are written against original duration, and they do not all count the same population of activities.

F

Float. Spare time. The word on its own is not an answer: a float figure means nothing until you know which float, against what anchor it was measured, and on which calendar it was counted. An activity can drive the project finish and still carry float, when the backward pass ran from a stated deadline that sits later than the computed finish.

Fragnet. A small network of activities modelling a delay event, built so it can be inserted into a base schedule. Additive methods work by inserting fragnets; their durations, their attachment points and the date each event is taken to have begun are all the analyst's, and all contestable.

Free float. The amount by which an activity can slip without moving any successor's early dates. It diverges from total float in ways that depend on what the successor is doing — its own constraints and placement — and, unlike total float, it carries no cap from the project completion date. Published P6 data contains rows where free float exceeds total float, so a reader who treats that as impossible will misread real files.

L

Lag. A delay applied to a relationship rather than to an activity: the successor's bound is advanced by the stated amount. Two things about it change dates and are easy to miss. It is counted on a calendar, and which calendar is a setting rather than a fact about the link. And an FS or SS lag lands in start space while an FF or SF lag lands in finish space. Several specifications cap, prohibit or require justification for lags, and no two of those caps count the same population.

Lead. A negative lag, permitting a successor to begin before its predecessor's driving event. Many specifications prohibit leads outright, on the ground that they model an overlap the logic should state directly.

Level of effort. An activity type whose duration is derived from the activities it spans rather than stated on its own — supervision, dewatering, site facilities. Authorities disagree sharply: one federal section prohibits the type entirely, a state specification presupposes it and regulates its naming, and others require it to be identified and kept off the critical path.

Longest path. The chain walked back from the activity that sets the project finish, following one driving predecessor at each step. Distinct both from the total-float definition of criticality and from the wider driving set, which branches wherever two predecessors tie.

M

Method Implementation Protocol (MIP). A coordinate in AACE RP 29R-03's taxonomy of forensic schedule analysis methods, numbered 3.1 to 3.9. The numbers matter because common names do not map onto them one-to-one: "time impact analysis" names two of them and "windows analysis" three. See Concepts §1 and Methods.

Milestone. A zero-duration activity marking an event rather than work — award, notice to proceed, substantial completion, an interim completion date. Start milestones and finish milestones behave differently in the arithmetic, and several clauses exempt milestones from rules about duration.

N

Negative float. The condition where an activity's late dates fall before its early dates, which happens when a required date is earlier than the logic-driven one. It is a symptom rather than a defect in itself, and the authorities are talking about different documents rather than disagreeing: several prohibit submitting a baseline carrying negative float, while a federal section expects the negative sign to appear on an update genuinely late against an interim completion date. See the standards crosswalk.

O

Open end. An activity with no predecessor or no successor, so that nothing in the network drives it or is driven by it. A dangling start or finish is the narrower case where an activity is tied only by relationships that leave one of its two ends free. Almost every authority checks for open ends; they differ mainly in which activities they exempt.

Out-of-sequence progress. Work recorded as started or finished before the logic said it could be — the field has already worked around a tie the schedule still contains. It forces a choice about what to do with that tie: see Progress override and Retained logic below. Authorities differ, one federal section forbidding it outright while another permits it case by case.

P

Predecessor. An activity that another activity's start or finish is tied to.

Progress override. A scheduling mode that drops the tie from an incomplete predecessor to a started successor, so the successor runs from the data date on its own remaining duration. It produces the earliest of the three modes' finish dates, and several specifications prohibit it for that reason. See Concepts §4.

R

Relationship. A logic tie between two activities. Four types:

• Finish-to-start (FS) — the successor may start once the predecessor finishes. The default, and the overwhelming majority of links in most networks.
• Start-to-start (SS) — the successor may start once the predecessor starts.
• Finish-to-finish (FF) — the successor may finish once the predecessor finishes.
• Start-to-finish (SF) — the successor may finish once the predecessor starts. The rarest, hardest to justify, and prohibited outright by several specifications.

Remaining duration. The working time an in-progress activity still needs. In Primavera it is stored independently of percent complete, so the two can disagree and a progressed schedule can state a remaining duration that its own percent complete does not support.

Retained logic. The scheduling mode that keeps every relationship, so an out-of-sequence successor still cannot finish before its predecessor does. It produces the latest of the three modes' finish dates, and is the mode several federal and state sections require. See Concepts §4.

S

Schedule update. A snapshot of the programme at a data date with progress recorded against it. The ordered chain of updates is the evidence that contemporaneous methods read, and the reason a delay analysis asks for every update rather than a baseline and a current schedule.

Successor. An activity whose start or finish is tied to another's.

T

Time impact analysis (TIA). A name, not a method. It describes inserting a delay fragnet into a base schedule and measuring the movement, but does not say which base schedule — one fixed plan, or the update current when each event occurred — and those are different analyses producing different numbers on the same facts. It is also used both prospectively and retrospectively, which are different evidentiary postures under different documents. See the method comparison.

Total float. The time an activity can slip before it delays the project completion, computed as the difference between its late and early dates. Three qualifications matter. It has two readings — late start minus early start, and late finish minus early finish — which can differ on the same activity. Some documents treat it as a property of a network path rather than of an individual activity, and those part company where an activity sits on more than one path. And it is measured against whatever anchor the backward pass ran from, which may be the computed finish or a stated completion date.

W

Windows analysis. A family of methods that cuts the project into periods and measures within each, rather than treating it as one span. The name is ambiguous between at least three distinct methods — static logic across every window, the project's own contemporaneous updates, or a modelled delay event per window — and they are not interchangeable. See the method comparison.

Work breakdown structure (WBS). The hierarchical decomposition of the project into which activities are grouped. It is an organising structure and not logic: grouping does not drive dates, and summary or rollup elements should not appear on a driving path.

Terms this glossary does not define are usually defined in Concepts, Methods or the standards crosswalk. If a report uses a word none of those explain, that is worth telling us — see talk to us.

Source: web/pages/glossary.md. Source commit date: 2026-09-11.