PERT common mistakes
PERT · Understand the traps and how to avoid them.
Treats every dependency in the network as finish-to-start with zero lag, even when the problem states start-to-start, finish-to-finish, start-to-finish, or a lag.
Most candidates learn the basic FS relationship first and default to it without re-reading the dependency labels in the question.
What to do instead
Before calculating dates, rewrite each dependency as 'predecessor relationship lag successor' and apply the matching forward/backward-pass formula for that relationship type.
Omits dummy activities when converting an activity-on-arrow network with overlapping predecessor sets, causing two activities to appear to share only part of a predecessor set.
Dummy activities are abstract and easy to skip when the diagram 'looks connected' without them.
What to do instead
Whenever a unique combination of predecessors is needed in AOA, insert a zero-duration dummy arrow to preserve the exact logic, then verify each activity has the intended predecessor set.
Computes the PERT expected duration as the simple average (O + M + P) / 3.
The arithmetic mean is more familiar and the weighted formula is easy to misremember under time pressure.
What to do instead
Use TE = (O + 4M + P) / 6 and sanity-check that the result lies closer to M than to the midpoint of O and P.
Adds activity standard deviations along a path to obtain the path standard deviation.
Durations are additive, so candidates incorrectly assume standard deviations are additive too.
What to do instead
Convert each activity SD to variance by squaring it, sum the variances along the path, then take the square root to get the path SD.
Uses (P - O) / 6 as the variance, or uses (P - O) as the standard deviation.
Formula recall blurs the range, standard deviation, and variance steps.
What to do instead
Write the chain explicitly: SD = (P - O) / 6; Variance = SD squared = ((P - O) / 6) squared.
Identifies the critical path as the shortest path, the path with the most activities, or the path that 'looks busiest' instead of calculating path durations.
Critical-path terminology suggests importance, and visual complexity is mistaken for duration.
What to do instead
Sum expected durations for every complete path from start to finish; the critical path is the longest total duration and has zero total float.
Uses total float and free float interchangeably when deciding whether delaying an activity delays a successor.
Both are called 'float' and the formulas are similar.
What to do instead
Compute TF = LS - ES = LF - EF and FF = earliest successor ES minus EF (for FS with zero lag); use FF to judge immediate successor impact and TF to judge project finish impact.
In the backward pass, sets an activity's late finish to the maximum successor late finish instead of using the minimum successor late start adjusted for the relationship and lag.
Forward-pass logic of taking maximums is carried into the backward pass.
What to do instead
For each successor relationship, calculate the latest allowable finish from that successor's late start and lag, then take the minimum across all successors.
Computes a completion-date probability using the critical path mean but an individual activity's standard deviation, or a sum of standard deviations, rather than the path standard deviation.
The z-score formula is remembered, but the correct mean and SD inputs for the whole path are not.
What to do instead
For the path, sum expected durations for the mean, sum variances and take the square root for the SD, then use z = (Target - Path Mean) / Path SD.
Reports a single PERT probability from the deterministic critical path and ignores near-critical parallel paths or merge bias.
Candidates equate 'critical path' with 'only path that matters' and forget that a slightly shorter path with high variance can be riskier.
What to do instead
Calculate completion probability for every path with duration close to the critical path, compare them, and use the path that gives the lowest probability of meeting the target as the governing risk.