Insights
Critical path analysis shows which activities decide your finish date and how much slack the rest have. This guide walks through the forward and backward pass, a worked example and the mistakes that make plans look safer than they are.
By Christopher Fagan, Managing Partner, PEF Consulting Partners
Delivered for and alongside
















Critical path analysis is a scheduling technique that finds the longest chain of dependent activities in a project. That chain, the critical path, sets the earliest date the project can finish. Any delay to an activity on it delays the whole project, while activities off it have float, meaning they can slip a little without moving the end date.
It is one of the oldest tools in project management and still one of the most useful, because it answers the question every sponsor eventually asks: what actually decides when we finish? This guide covers the method, a worked example you can follow with a pencil, and the judgement calls that matter once the plan meets reality.
In this article
A project plan is a network of activities, and most of them depend on something else finishing first. Some chains through that network are long and some are short. The longest one fixes your end date, and the analysis tells you which chain that is and how much float every other activity has.
That gives a sponsor three practical things. You know where management attention should go, because a day lost on the critical path is a day lost at the end. You know where there is room to move people or absorb a late supplier. And you can test a proposed change, such as a new requirement or a shorter test window, against the end date before anyone agrees to it.
The critical path method, usually shortened to CPM, was developed in the late 1950s by Morgan Walker of DuPont and James Kelley of Remington Rand, who were looking for better ways to schedule DuPont’s engineering projects. Kelley and Walker published the method in 1959.
PERT, the Program Evaluation and Review Technique, appeared at almost the same time. It was built for the US Navy’s Special Projects Office to manage the Polaris missile programme, where much of the work had never been done before and durations were genuinely uncertain. That difference still shapes how the two are used. CPM assumes you can estimate each duration with reasonable confidence, while PERT works with a range.
You need four things before you can calculate anything.
There are two kinds of float worth separating. Total float is how far an activity can slip without moving the project end date. Free float is how far it can slip without delaying the earliest start of the next activity. The difference matters when two teams both think they own the same spare days.
Scheduling software does this arithmetic for you, but working one small example by hand is the quickest way to understand what the software is telling you.
The forward pass and backward pass are the whole technique. Everything else is presentation and judgement.
The example below is invented for illustration. Imagine a finance team rolling out a new expenses system, with seven activities, A to G, and durations in working days.
| Activity | Description | Duration (days) | Depends on |
|---|---|---|---|
| A | Agree requirements | 4 | None |
| B | Configure the system | 10 | A |
| C | Prepare the data extract | 6 | A |
| D | Build the payroll interface | 8 | A |
| E | Load and reconcile data | 5 | B, C |
| F | System and integration testing | 7 | B, D |
| G | Train users and go live | 3 | E, F |
We count from day 0, so an activity that starts at day 4 and lasts 10 days finishes at day 14. Some tools count from day 1, which shifts every number by one but changes nothing about the result.
The forward pass runs left to right. A starts at day 0 and finishes at day 4, and B, C and D, which depend only on A, all start at day 4.
E needs both B and C, so it waits for the later of the two, B, and starts at day 14. F also waits for B and starts at day 14. G needs E, finishing at day 19, and F, finishing at day 21, so it starts at day 21 and the project finishes at day 24.
The backward pass runs right to left from day 24. G must start by day 21, so E and F must both finish by then, which means E must start by day 16 and F by day 14. B feeds both, so it must finish by the earlier of those two dates, day 14, and start by day 4. C must finish by day 16, D by day 14 and A by day 4.
| Activity | Duration | ES | EF | LS | LF | Total float | Critical |
|---|---|---|---|---|---|---|---|
| A | 4 | 0 | 4 | 0 | 4 | 0 | Yes |
| B | 10 | 4 | 14 | 4 | 14 | 0 | Yes |
| C | 6 | 4 | 10 | 10 | 16 | 6 | No |
| D | 8 | 4 | 12 | 6 | 14 | 2 | No |
| E | 5 | 14 | 19 | 16 | 21 | 2 | No |
| F | 7 | 14 | 21 | 14 | 21 | 0 | Yes |
| G | 3 | 21 | 24 | 21 | 24 | 0 | Yes |
The critical path is A, B, F, G, and it runs to 24 working days. Those four activities have no float at all, so a day lost on any of them is a day lost at the end.
The float column is where the management value sits. C shows six days of total float but only four days of free float. If C slips by more than four days, it starts using E’s float, and E only has two. Float is shared along a chain, so it can only be spent once, by whoever gets there first.
Look at whole paths as well as single activities. A, B, E, G and A, D, F, G each run to 22 days, just two days short of the critical path, so this plan really has three paths that matter. If you shortened F by three days, for example by adding testers, the project would finish only two days earlier, because A, B, E, G would become the new critical path at 22 days.
This is why one red line on a Gantt chart can mislead a steering group. Near critical paths deserve the same attention, and they belong in the programme status report, not buried in the scheduling tool.
The critical path method uses one duration per activity, which works when the work is familiar. When it is not, PERT asks for three estimates: optimistic, most likely and pessimistic. The expected duration is the optimistic figure, plus four times the most likely, plus the pessimistic, all divided by six.
Take activity F, the testing window. If the team says it could be done in 5 days, will most likely take 7 and could take 15 if defects pile up, the expected duration is (5 + 28 + 15) divided by 6, which is 8 days. That extra day lands straight on the end date because F is critical, and the pessimistic case would move it by eight days. The real value of PERT is less the arithmetic than the conversation it forces about where the risk sits.
On most business programmes, full PERT statistics are overkill. Three point estimates for the handful of activities with real uncertainty, usually testing, data migration and anything that depends on a third party, combined with a visible schedule contingency, give most of the benefit for a fraction of the effort.
It earns its place on any project with a fixed end date and real dependencies: system implementations, office moves, plant shutdowns, regulatory deadlines and multi site rollouts. It is most valuable at two moments, when the plan is first baselined and when something has gone wrong and you need to know quickly whether the date still holds. The cost of delay calculator helps put a value on each day of the critical path, which makes trade offs easier to argue.
It is less useful where the work is not a network of dependent tasks. Agile teams working from a prioritised backlog get more from cadence and flow measures, although the release they are building usually sits inside a wider plan that does have a critical path. It also struggles when people, not logic, are the real constraint, which is why scheduling tools add resource levelling and some teams use critical chain methods on top.
The analysis is only as good as the plan underneath it. That plan starts with a clear scope, broken down into deliverables, then sequenced and estimated in a project plan. On ERP and billing programmes the critical path usually runs through data migration, testing and cutover, which is why our ERP implementation timeline guide spends so much time on them.
Keeping the path current is a PMO job. A good PMO refreshes the schedule each cycle, challenges optimistic durations and makes sure the steering group sees the near critical paths as well as the main one. If yours does not, that is the gap our PMO setup and uplift work is designed to close.
The same logic scales up. When a strategy depends on several programmes landing in the right order, the critical path runs across all of them, and managing it becomes part of strategy execution rather than project scheduling.
Christopher Fagan, PEF’s Managing Partner, has spent 19 years delivering systems and transformation programmes. On BP Pulse, where he was Global Transformation Director, the Oracle to Microsoft Dynamics migration completed 39 days ahead of forecast. Dates like that come from knowing which work sets the end date and protecting it every week.
Critical path analysis is a way of scheduling a project by finding the longest sequence of dependent activities from start to finish. That sequence is the critical path, and its length is the shortest time in which the project can be completed. Activities on it have zero float, so any delay to them delays the end date, while activities off it can slip by their float without moving the finish.
List every activity with its duration and dependencies. Run a forward pass to find each activity’s earliest start and finish, then a backward pass from the end date to find its latest start and finish. Total float is the latest start minus the earliest start. The activities with zero float, linked from start to finish, form the critical path, and the earliest finish of the last activity is the project duration.
Float, also called slack, is the amount of time an activity can be delayed without causing a problem. Total float is how long it can slip before the project end date moves. Free float is how long it can slip before it delays the earliest start of the next activity. Activities on the critical path have zero total float, which is why they need the closest management.
The critical path method uses a single duration estimate for each activity and suits work that is well understood, such as construction or a repeat system rollout. PERT uses three estimates, optimistic, most likely and pessimistic, and combines them into an expected duration, which suits uncertain work better. Both build the same network of dependencies, and many plans use CPM overall with three point estimates for the riskiest activities.
Yes, and it usually does. As activities finish early or late, another chain of work can become longer than the original critical path, and a path with only a few days of float can take over after one late delivery. That is why the analysis should be refreshed every reporting cycle, and why near critical paths should be reported to the steering group alongside the critical one.
Related reading: Project dependencies explained, Work breakdown structure, Project plan template and How to read a programme status report.
If your end date keeps moving and nobody can say why, we can review the schedule with your team, find the paths that really matter and help you keep them under control.
Book a confidential conversation
Sources: Kelley, J. E. and Walker, M. R., Critical Path Planning and Scheduling, Proceedings of the Eastern Joint Computer Conference, 1959; Malcolm, D. G., Roseboom, J. H., Clark, C. E. and Fazar, W., Application of a Technique for Research and Development Program Evaluation, Operations Research, 1959
Written by Christopher Fagan
Managing Partner, PEF Consulting Partners
Christopher has led systems and transformation delivery for 19 years, including the National Grid separation, the £9bn Energy Bills Support Scheme and the BP Pulse EV charging rollout.
Where next