Finding the critical path in a marketing programme

The longest task is almost never the one that decides your launch date. The longest chain is.

Critical path method finds the longest chain of dependent work through a project. That chain, not the biggest task, determines the earliest possible finish. In marketing and martech programmes the critical path usually runs through approvals, third-party access and data availability rather than through production work, which is why teams that only manage production keep missing dates.

What the critical path is

The critical path is the longest chain of dependent activities from start to finish. Its length is the shortest possible duration of the project. Any task on that chain that slips by a day pushes the end date by a day. Any task off it has float, meaning it can slip by some amount without affecting the finish.

The technique comes from construction and heavy engineering, and it transfers to digital work better than most people assume. What does not transfer is the assumption that durations are known and dependencies are physical. In marketing programmes the dependencies are informational and organisational, which makes them harder to see and easier to ignore.

The practical payoff is not the diagram. It is knowing which three or four items to chase every week and which fifteen to leave alone. Most programme managers chase everything equally, which means they chase the wrong things with the same energy as the right ones.

Where the critical path actually runs in marketing work

Almost never through production. Building landing pages, writing ads, cutting creative: these are parallelisable and the team controls the resourcing. If you need them faster you add a person, and the chain shortens.

It runs instead through four things you do not control. Client approval cycles, which are serial by nature and gated by one person's calendar. Third-party access, meaning ad account permissions, analytics property access, CMS credentials, and DNS control. Legal and brand review, which is serial and has its own queue. And data availability, because measurement work cannot start until the data exists and has accumulated enough history to be meaningful.

Conversion tracking is the clearest example. You cannot validate an attribution change until enough conversions have flowed through the new setup, and that duration is set by traffic volume, not by effort. No amount of staffing shortens it. If a reporting deliverable depends on that validation, the wait is on your critical path and it was probably never in your plan.

The same applies to any account with a learning period. Bidding changes, audience changes and structural changes all carry a settling window before the numbers mean anything, which is why account structure decisions should be made before a campaign launches rather than after.

Building the network in one sitting

You do not need software. Take the work packages from your breakdown, write each on a sticky note or a row, and for each one answer a single question: what must be finished before this can start? Not what should be finished, what must be. The difference between those two words is where most fake dependencies come from.

Then walk forward through the chains and add durations. The longest total is your critical path. Everything else has float equal to the difference between its chain and the longest one.

Two rules make the exercise honest. First, use elapsed duration, not effort. An approval that takes two minutes of someone's attention and sits in a queue for six business days is a six-day activity. Recording it as two minutes is how programmes miss dates while everyone reports being on track.

Second, do not add buffer inside the estimates. Buffer hidden in individual tasks disappears, because work expands to fill it. Keep the durations honest and hold one explicit buffer at the end of the chain where you can see it being consumed.

Duration is not effort: the distinction that decides your date

Duration is not effort: the distinction that decides your date
ActivityEffortElapsed durationOn the critical path?
Build 12 landing pages60 hours8 days with two peopleRarely, it parallelises
Client approves creative1 hour5 to 10 business daysAlmost always
Obtain ad account admin access15 minutes3 to 15 days across an agency handoverAlmost always
Legal review of claims3 hours10 business days in a queueOften
Accumulate conversion data to validate tracking0 hours14 to 28 days, set by trafficYes, and usually unplanned
Write ad copy variants20 hours3 daysNo, it has float

Read the effort column and the duration column together. The three activities with the least effort carry the most elapsed time, and those three are what decide whether you launch on the date you promised.

This is why staffing a programme harder frequently changes nothing. Adding a second developer compresses row one, which had float. It does not touch rows two, three and five, which do not consume your team's hours at all.

Compressing the path when the date will not move

Only two levers exist and both cost something. Fast-tracking runs activities in parallel that were planned in sequence, which buys time and adds rework risk, because the second activity starts on an assumption about the first. Crashing adds resource to a critical activity, which buys time and adds cost, and only works where the constraint is effort rather than elapsed time.

Neither lever works on an approval queue, which is where the marketing critical path usually sits. The lever that does work there is procedural: pre-book the review slot rather than requesting it when the work is ready, and reduce the number of serial approvers. Two serial reviewers at five days each is ten days. The same two reviewing in parallel is five, and the only cost is that they see a draft simultaneously rather than sequentially.

Access requests deserve the same treatment. Start them in week one regardless of when the dependent work is scheduled, because the request is free to make early and expensive to make late. On any agency transition this is the single highest-value thing to do on day one.

A Gantt view is the standard way to see the effect of these moves, because it renders sequence and dependency together. Atlassian's overview describes the chart as a way to visualise tasks against a timeline with their dependencies made explicit, which is exactly the picture you need when arguing that a date is infeasible.

What breaks the technique

Three things, and they are all common in agency work. The first is fake dependencies. Teams mark work as sequential because that is the order they intend to do it in, not because one genuinely requires the other. Every fake dependency lengthens the apparent critical path and hides the real one.

The second is a resource-blind network. Critical path assumes unlimited resource. If two parallel branches both need the same specialist, the branches are not really parallel, and your chain is longer than the diagram says. In a small team this happens constantly, because there is one person who can do the data work.

The third is treating the path as fixed. It moves. When a critical activity finishes early or a floating branch slips, the critical path relocates, and the item you have been chasing for three weeks stops mattering while something you have been ignoring becomes urgent. Recalculate weekly or the analysis actively misleads you.

Microsoft's Azure Boards documentation on planning and tracking work across a team is a reasonable reference point for keeping the network live in a tool rather than in a document that ages the moment it is exported.

Critical path versus theory of constraints

For recurring delivery rather than one-off projects, critical path is the wrong tool. A retainer that ships the same kind of work every month does not have a start and a finish, so it has no critical path. What it has is a bottleneck: the one stage where work queues.

Find it by looking for the stage with the largest backlog in front of it, then improve only that stage. Improving any other stage adds nothing, and typically makes the queue in front of the bottleneck longer. This is the discipline that turns a chaotic monthly reporting cycle into a predictable one, and it is why reporting cycles are worth engineering as a process rather than a scramble.

The two techniques answer different questions. Critical path answers when will this finish. Constraint analysis answers where should I spend my improvement effort. Programmes running several projects at once usually need both, which is the coordination job program management describes.

What this looks like on a real programme

On a transit BI migration, the visible work was moving 154 data sources without an unplanned outage. The critical path did not run through the migration scripts. It ran through source system owners confirming what could be cut over and when, which is an approval chain, and through validation windows that could not be compressed because they required a full reporting cycle to complete.

Recognising that early changes what you manage. You stop pushing the engineering team, who have float, and start managing a calendar of confirmations from people who do not report to you. That is uncomfortable work and it is the work that protects the date.

The failure mode when nobody does this is familiar: a programme where everyone is busy, every individual workstream reports green, and the date slips anyway. That gap between local status and global truth is one of the clearest signals that nobody owns the sequence, and it is the specific problem delivery ownership as a defined engagement is meant to close.

Start smaller than a full network diagram if you need to. Pick your committed date, work backwards through the chain of things that must happen before it, and stop at the first item where the elapsed duration is outside your control. That item is your programme, and everything else is production.

Start a conversationMore insights