A schedule is a model of how the work will unfold in time, built from activities, the order in which they can happen, and how long each takes. Its value is not the dates it prints but the structure underneath, which shows which delays matter and which do not.
Most of what a schedule tells you comes from the relationships between activities rather than from the durations. Get the network wrong and every date derived from it is wrong, and no estimating accuracy rescues it.
Sequencing and the four kinds of dependency
Activities are joined by logical relationships, each named for which end of the predecessor governs which end of the successor.
Finish to start is the common one, where the successor cannot begin until the predecessor finishes. Start to start lets the successor begin once the predecessor has begun, which is how paving follows grading down a road at a fixed distance behind. Finish to finish means the successor cannot complete until the predecessor does, which fits testing that has to close out alongside the writing it inspects. Start to finish is rare, since the successor cannot finish until the predecessor starts, and its standard illustration is a shift that must begin before the outgoing one may end.
Dependencies are also classified by where the constraint comes from. Mandatory logic is inherent in the work, since concrete has to be poured before it can cure. Discretionary logic is a preference drawn from practice, and it is the first place to look when compressing a schedule. External logic sits outside the project, such as a regulator's approval date, while internal logic is under the project's control.
Lead and lag
A lead lets a successor start before its predecessor finishes, so a finish to start relationship with a five day lead begins the successor five days early. A lag inserts waiting time, such as the fortnight concrete needs to cure.
Neither is padding. A lag is a property of the work and belongs in the network, whereas a margin added because somebody is nervous belongs in a reserve where it can be managed.
The critical path is the longest path
The critical path is the longest path through the network, and because every activity on it must happen in sequence, it is also the shortest possible duration for the project. Those two statements sound contradictory and describe the same fact.
The path through B and E runs to 22 days, which fixes the project duration and leaves that path with no room at all. The path through C and F takes 18 days, so the four days between it and the critical path are the float shared along it, and the path through D and G carries eight.
Position in the logic is the whole of it. An activity is critical because everything downstream waits on it, which has nothing to do with how difficult or how valuable the work is. There may be more than one critical path, and a project carrying several is more fragile, since a delay on any of them moves the finish date.
Float is the room the network contains
Total float is how long an activity can slip without delaying the project finish. Free float is how long it can slip without delaying any immediate successor, which is a tighter and more local number.
Activities on the critical path have zero total float by definition. Where a schedule shows negative float, the network is reporting that the imposed date cannot be met with the logic and durations as they stand, so something has to change rather than the team working harder against a number.
Float belongs to the path rather than to the activity looking at it. Two activities in sequence sharing ten days of float have ten days between them and not ten each, which is why consuming it early leaves nothing for later.
The network above shows the same arithmetic at a readable size. Both C and F report four days of total float, and the path they sit on carries four days in total, so C slipping by three leaves F with one. Spending float is a decision taken on behalf of everything downstream of you.
Crashing and fast tracking are different trades
Crashing adds resources to activities on the critical path, so it buys time with cost. It works only where an activity responds to more people, and it meets diminishing returns quickly.
Fast tracking runs activities in parallel that were planned in sequence, so it buys time with risk. Starting construction on a design that is eighty per cent settled means some of it may be undone when the last fifth arrives, and rework is the price of the overlap.
| Technique | What it does | What it costs | What it risks | When it fits |
|---|---|---|---|---|
| Crashing | Adds resources to activities on the critical path so each takes less time | Money, and usually more than proportionally, since the fourth engineer adds less than the second | Coordination overhead, quality slipping as people join late, and a new critical path appearing elsewhere | There is budget to spend, and the activity genuinely responds to more people |
| Fast tracking | Overlaps activities that were planned in sequence, so the successor starts on partial information | Nothing directly, which is what makes it tempting | Rework when the predecessor changes after the successor has started, and a much busier set of dependencies to manage | There is tolerance for rework, and the logic being overlapped is discretionary rather than mandatory |
Look at the network above before reaching for either. Adding people to C or D buys nothing at all, because those paths already carry float, and the only compression that moves the finish date is compression on A, B, E or H.
Neither produces time from nothing. Every compression technique trades one constraint for another, and a plan that shortens without conceding cost, risk or scope has usually moved the optimism somewhere less visible.
Rolling wave planning keeps the far future coarse
Detailed planning is done for imminent work and left deliberately coarse further out, with detail added as the horizon approaches. Planning packages hold the coarse portion until it is decomposed.
The coarseness is the rigour rather than the absence of it. Detail written about work eighteen months away will be rewritten before anybody uses it, so the effort is spent twice. The same reasoning underlies adaptive planning, which is why hybrid projects sit comfortably with a milestone schedule above short cycles.