First, time the last few late launches stage by stage
Start from evidence rather than from the most recent complaint. Take the last few campaigns that went live after their planned date and, from the work system's history, write down when each one entered and left each stage of the route the campaign workflow guide draws: intake, brief signed, build, review and approval, launch check, live. Beside each stage, split the elapsed time into time someone was working on the item and time it sat waiting, as the tutorial on mapping a marketing process does for any recurring process. Mark every loop back, such as work returned from review to build, and the reason it returned.
Two readings usually narrow the diagnosis at once. First, where does the waiting pile up? The theory of constraints explains why one step sets the pace of the whole route and why work queues in front of it; if most of the late days sit before one gate, that gate is the place to look, whatever the people around it say. Second, do items loop? A campaign that passes review once and launches is slow somewhere else; a campaign that goes back to build twice has a cause upstream, usually at a handover where something arrived incomplete or changed after the fact.
If the history cannot be read because stages were never recorded, or the launch date lives in a spreadsheet rather than on the work item, that is already a finding, and one of the causes below. A campaign rushed through its final days also tends to have its name and tracking typed at launch, so the diagnosis for inconsistent campaign names often turns up the same campaigns from the other end. And if the dates slip because the plan itself is rewritten every few weeks, start with the diagnosis for a plan that keeps changing instead, because no workflow can hold a date the plan keeps moving.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Campaigns pass build on time, then go back for changes after review | Whether the brief was signed before build, and what changed between the first and last version | Build starts before the brief is signed, so the offer or message changes after the work is made | No build without a brief signed by the requester and the owner; changes go back through the signature |
| Most late days sit waiting for review, and every campaign queues for the same people | Who reviewed each late campaign, in what order, and how many rounds each review took | Reviewers chosen by seniority and working one round after another | Decide reviewers by risk with a matrix applied at intake, and review in parallel with one person consolidating |
| Many items are started, few finish, and each spends longer waiting than being worked on | Open items by stage against items finished per period | More work open than the team can finish | Set a limit at the stage where work queues and pull new work only when something finishes |
| Work sits paused for inputs, approvals or comments from the requester | Time each late item spent waiting on the requester, from the work system's history | No agreed turnaround on the requester's side | Write both sides of a service level per request type, with the clock pausing while work waits on the requester |
| The fix last time went to the step where the delay was noticed, and nothing improved | The traced map of a few recent items, with waits and loops marked | The process has never been mapped as it actually runs | Map recent items through the process and relieve the step where work queues longest |
| The slip is discovered on launch day, though a stage overran a week earlier | Whether launch dates live on work items or in a separate calendar | Launch dates kept by hand outside the work system | Turn the calendar into a view of the work system and read planned against actual dates by stage |
| Launches stall when one person is away | Which steps of the route only one person can do or has access to | A one-person step such as publishing, trafficking or a tag change | Add a second trained person, a procedure or a supplier, and test cover before an absence |
Fix, verify and prevent each cause
Build starts before the brief is signed
The campaign workflow guide is plain about where most rework at review comes from: a brief nobody agreed, or one that a single party changed afterwards without saying so. Building before the briefA short document telling the people making creative work what it must achieve, for whom, with which message and within which limits. is signed seems quicker, yet it drives most of the loops that send finished work back to build.
Fix: make the signed brief the entry condition for build, signed by both the requester and the campaign owner, and route every change after signing back through the same signature so it is visible as a change rather than discovered at review.
Verify: for the next few campaigns, compare the brief at signature with the brief at launch; changes should appear as dated re-signatures, and returns from review should no longer cite an offer or message that moved.
Prevent: add a return reason such as "brief changed after build" to the review gate's short list, so the next time it happens it is counted rather than absorbed.
Reviews queue for the same senior people, one round after another
The approvals and compliance guide describes how approval routes tend to pile up: after an error, a senior person asks to see every piece of that kind, and a few rounds later every piece passes several reviewers whatever it carries. The fix it gives is an approval matrixA table that says which reviews each type of marketing work needs, such as legal, brand or budget, and who gives each, matched to the risk of the work. that decides reviewers by the risk the work carries and is applied at intake, so reviewers are known before build starts and can hold time. Claims deserve their own place in that matrix. Advertising rules in most markets share the principle that claims must be capable of substantiationThe evidence a business holds, before a claim is published, that the claim is true; advertising rules in most markets expect claims to be capable of it. before they are made, as the guide notes the Advertising Standards Authority sets out in the UK, which is why new claims are flagged at intake and reach legal before the work is designed around them. A claims library lets routine work reuse approved wording, so the constrained legal reviewer's time goes only to what is new.
Fix: move reviewer selection to intake, run the required reviews in parallel against the signed brief, and give one person the job of consolidating conflicting comments.
Verify: the next campaigns show reviews starting on the same day and one consolidated round.
Prevent: review the matrix itself on a cadence; a row that never sends anything back may not be needed.
More work is open than the team can finish
The guide to limiting work in progress explains the mechanism with the theory of constraints: opening more items leaves the slowest step's output where it was and simply makes its queue longer, so each campaign spends longer waiting than being worked on and the time from request to launch grows while nobody works less.
Fix: count everything open, including requests held in inboxes, set a limit at the stage where work queues, count blocked and waiting items against it, and pull new work only when something finishes. Which requests come first stays a planning decision, covered by the guide to prioritising work and capacity.
Verify: over the following weeks, open items fall towards the limit while finished items hold or rise.
Prevent: when an item is stuck, chase whatever is blocking it rather than starting something new, and put stuck items at the top of every board review.
The requester's side has no agreed turnaround
The records often show the wait for feedback as the longest stage of all, as the tutorial on setting service levels for marketing requests notes, and that wait is invisible when only the team's turnaround is written down. A service levelA published promise between teams, or with a supplier, about what each type of request needs and how quickly it will be handled, agreed rather than imposed. is two-sided: the team commits to a typical turnaround for each request type, and the requester commits to a complete request, comments by an agreed date gathered into one set across all their reviewers, and a named person who can approve.
Fix: write both sides for the request types that run late, set them from the team's own records rather than a borrowed benchmark, and let the clock pause visibly while work waits on the requester.
Verify: the work system shows the paused stage, and the next late item can be read as late on one side or the other rather than argued about.
Prevent: publish the catalogue at the intakeThe single route through which every request reaches the marketing team, usually a short form, so work is triaged and checked before it starts. form and report misses on both sides as a measure of the system.
The process has never been mapped as it actually runs
The mapping tutorial points out that work is seldom held up at the step where someone first spots the lateness: a page that launches late may have sat waiting for copy, and the copy for a decision nobody knew they owned. Teams that repair the step where the pain showed often see nothing improve.
Fix: trace a few recent campaigns through the route together with those who did the work, lay the steps out in one lane per role, mark each handover, wait, loop and workaround, and choose the step where work queues longest as the one to relieve.
Verify: after the change, the waiting time in front of that step falls in the next few items, and the loops that kept returning to it stop.
Prevent: keep the map beside the procedure it describes and redraw it when the route changes.
Launch dates live in a calendar kept by hand
The guide to work management and marketing calendars treats the calendar as a view of the work system: every item carries its launch date, and the calendar is those items filtered by date and channel. A calendar kept by hand next to the work system falls behind, since keeping it current is an extra job with no owner, and that is how a stage that overran last week becomes a surprise on launch day.
Fix: move launch dates onto the work items, record stage changes, and run the guide's late launches report, planned against actual launch date by stage.
Verify: a stage overrun now shows on the calendar view the same day, and the report names the stage each slip came from.
Prevent: let stakeholders see the calendar view directly, so status requests by email dry up.
A step only one person can do
The guide to reducing key-person risk describes work that stops during a holiday or a busy week because only one person can publish to the site, change the tag container or traffic an ad, and nobody notices until something further down the line goes wrong. For a launch, that something is the date.
Fix: list the steps in the launch route that depend on one person or one login, then add a second administrator, a written procedure, cross-training or a supplier for each one that matters.
Verify: run the step once with the cover person before an absence forces it.
Prevent: keep the coverage matrix current and check it when someone joins, leaves or changes role.
Prevention: a planned-against-actual report, a work limit and a requester agreement
Every cause above leaves the same trace: a stage that took longer than it should and a date that moved without anyone deciding it would. The cheapest prevention is to make that trace visible while it can still be acted on. Record stage changes on every work item and run the late launches report monthly, planned against actual launch date by stage, as the work management guide describes. Read it as a measure of the system rather than a KPI for individuals, because a cycle time that becomes a personal target teaches people to skip gates and close items early. When the same stage appears behind most slips for two readings running, that stage is the constraint, and the effort for the next period belongs there.
Hold three agreements in place between those readings. The first is the entry rule: no build without a signed brief, and every change after signing goes back through the signature. The second is the work limit at the constraint, with blocked and waiting items counted and new work pulled only when something finishes. The third is the two-sided service level, published at the intake form, so requesters see the turnaround they can expect and the inputs, feedback and approver they owe in return. Each agreement removes a kind of waiting that no amount of effort on the team's side can recover.
Then close the loop on every campaign that still slips. The campaign post-mortem template notes which workflow stage added to a shortfall and what will change next time; for lateness, the useful entry is the stage, the wait and its cause, not the size of the delay. Keep the reviewer matrix and the coverage matrix on the same review cadence, and keep the launch check, done by someone other than the builder with the campaign launch QA checklist, as the last gate even when a date is tight, because a rushed launch that goes out wrong costs more time than the day it saved.
Finally, keep the report cheap to produce. If assembling it becomes a monthly chore of exports and spreadsheets, the diagnosis for a monthly report that takes days covers why reporting effort grows and how to bring it back to a fixed pack. When late launches arrive together with other operations symptoms, the marketing operations troubleshooting hub and the process hub help decide whether a broader review is the better start.
A garden centre chain finds where its seasonal launches lose their days (illustrative)
Hexlandry Garden Centres
Hexlandry Garden Centres, an invented chain of 7 garden centres with a marketing team of 5, launched its last 6 seasonal campaigns an average of 9 working days after their planned dates. The head of marketing pulled the stage history for those 6 from the work system. Build took 11 working days on average against a plan of 10. Waiting took the rest: an average of 5 working days queued for the commercial director, who reviewed every piece after the brand review had finished, and 3 days waiting for comments from the buying team, which arrived from 3 people on different days.
Four of the 6 campaigns had looped back to build after review, each time because the buying team changed the offer after the creative was made; none of the 4 briefs had been signed by the buying team. On one launch the only person who could publish to the website was on leave for a week. The team had 16 items open at once against 4 or 5 finished each fortnight.
The chain made the signed brief the entry rule for build, ran brand and commercial reviews in parallel with the director seeing only pieces carrying a price or offer, set a limit of 8 open items, agreed with the buying team a turnaround of 2 working days for consolidated comments from one named approver, and trained a second web publisher. Over the next season, 5 of 6 campaigns launched on their planned date, and the sixth slipped 2 days waiting for a supplier's product images, which the planned-against-actual report showed on the day the stage overran.
Common mistakes
- Padding every deadline. The extra days fill with more open work, and the same step still sets the pace.
- Fixing the step where the delay was noticed. Time the stages first; the wait usually started earlier.
- Blaming the reviewers. Review is where lateness becomes visible, not usually where it begins.
- Starting build while the offer is still moving. It seems quicker, and it is where most rework comes from.
- Writing only the team's turnaround. The requester's side is often where the longest wait sits.
- Picking up fresh work while the current item is stuck. Half-finished items multiply that way.
- Keeping the calendar in a separate spreadsheet. The slip is then found on launch day.
- Turning cycle time into a personal target. People skip gates to protect the number, and launches go out wrong instead of late.
