Why more work in progress means less finished
Most overloaded marketing teams are not short of effort. They are short of finished work. A typical week has a dozen campaigns, emails, pages and content pieces in flight, each partly done, each waiting for something: a review, an answer from a product manager, an asset from an agency, a slot in a developer's queue. Everyone is busy and stakeholders still ask why nothing seems to launch.
The reason is the one the theory of constraints describes. Any system that produces an output is limited by one step, and work piles up in front of that step. Starting more work does not raise the output of the limiting step; it lengthens the queue in front of it. Each item then spends longer waiting than being worked on, so the time from request to launch grows even though nobody is working less. The framework's answer is to pace everything else to the constraint, and a work in progress limit is the practical way to do that: the team stops starting new items while the queue in front of the limiting step is full.
There is a second, simpler effect that any marketer recognises. Every open item asks for a little attention: status questions, reminders, a reread of the brief before picking it back up. Spread across too many items, that overhead crowds out the work itself. This guide does not lean on research about multitasking; the team's own records will show the pattern more convincingly than any study. When open items rise and finished items stay flat, the extra work in progress is producing waiting, not output. Where that waiting surfaces as campaigns missing their launch dates, the diagnosis for campaigns that keep launching late separates it from the other reasons a date slips.
Limiting work in progress is not the same as prioritising. Which requests the team accepts, and in what order, is decided at intake against the plan and the team's capacity, as the guide to prioritising marketing work and capacity sets out. A limit decides when accepted work starts. The two support each other: prioritisation keeps the queue honest, and the limit keeps the queue from turning into a pile of half-finished work.
Seeing everything that is open
A limit is meaningless until the team can count what it is limiting. The first job is to make every open item visible in one place, which is what the work management guide means by work that exists only if it is in the work system. Its report table already lists work in progress and ageing among the measures a work system provides; this guide turns that report into a rule.
Counting usually surprises people, because the work system shows only part of what is open. The rest is hidden:
| Where open work hides | What it looks like | How to bring it into view |
|---|---|---|
| Inboxes and chat | Requests agreed in a message and started without an item | Create an item for each, then send future requests to the intake form |
| Personal lists | Tasks a specialist is doing for a colleague as a favour | Ask each person to list what they are working on that is not on the board |
| Agency and freelancer queues | Briefs sent out and not yet returned | Give each external brief an item with the agency as the current holder |
| Recurring work | The newsletter, the monthly report, the social calendar | Show each instance as an item, so it uses capacity like any other work |
| Stalled projects | Work nobody has touched in weeks but nobody has closed | Flag items that have not moved, and decide to restart, park or close each |
Recurring work deserves particular care. The editorial calendarA shared schedule of what content will be published, on which channel, when and by whom., the monthly report and the regular email are rarely on the board as items, yet they take a steady share of the team's time. Leave them off and every limit looks generous, because the team is quietly working on more than the board shows.
Do the first count without judgement. The point is an honest picture of how much is open at each stage, how long each item has been open and what the team finished in the same period. The creative production playbook already reviews work in progress at its production stand-up; the count makes that review cover everything, not only creative.
Setting limits by stage or by person
There are two common ways to set limits, and teams can combine them.
Limits by stage cap how many items may sit in a stage at once, such as build or review. They suit most teams, because the stages of the campaign workflow are where work waits: a full review column shows reviewers are the constraint, and a full build column shows the team has started more than it can finish. Stage limits make the queue visible to everyone, which is half their value.
Limits by person cap how many items one person holds at once. They suit teams where each person owns a whole type of work end to end, or where one specialist, such as the only designer or developer, is the step everything waits on. A person limit protects that specialist from being handed work faster than they can finish it.
Set the first limits from the team's own records, never from a benchmark. Look at what is open in each stage now and what the team finishes in a typical period. Set the first limit somewhat below what is open today, so it bites, and then adjust it by watching. If work still piles up in front of a stage, the limit upstream is too high. If people sit idle waiting for work to arrive, the limit is too low. Read movements in small counts with the caution the guide to reading small numbers recommends; one unusual week is not a trend.
Whatever the method, the limit is a team agreement about flow. It is not a target for individuals, and it should not become one. As Goodhart's law warns, a measure that becomes a target stops measuring well: if people are judged on staying under a limit, items get closed early, split into pieces or kept off the board. If the team wants a KPIA metric tied directly to an objective. for this work, the time from request to launch says more than any count of open items.
Blocked, waiting and parked work
Every board fills with items that are not moving, and how they are treated decides whether the limit holds.
Blocked items cannot move because something outside the team's control is missing: an answer from legal, a product decision, a missing asset. They still count against the limit. That is the point: when blocked items fill a stage, nothing new can start there, and the team's attention turns to clearing the block. Mark the blocker on the item, name who is chasing it and look at blocked items first at every board review.
Waiting items are ready but sit in a queue, such as finished work waiting for review. They also count, because they show where the constraint is. A queue that keeps growing in front of one gate is the clearest sign of the limiting step, and that is where improvement effort belongs.
Parked items have been deliberately stopped by a decision, not by a blocker: a campaign postponed to next quarter, a page on hold until a product launches. Move them off the board into a parked list with a reason and a date to look at them again, so they stop counting against the limit. The test is whether anyone is expected to work on the item soon. If not, it is parked, and pretending otherwise hides the true load.
The one thing to avoid is quietly starting a new item because the current one is blocked. That is how teams end up with many items each a little further along and none finished.
Pulling new work instead of pushing it
In most teams work is pushed: a request arrives, someone accepts it, and it lands on a person's list whether or not they have room. A work in progress limit changes the direction. New work is pulled from a ready queue when a person or stage finishes something and has space under the limit.
Three changes make pulling work in practice:
- A ready queue in priority order. Accepted requests wait in one visible list, ranked by whoever owns prioritisation, with a complete creative briefA short document telling the people making creative work what it must achieve, for whom, with which message and within which limits. or campaign brief where the work needs one. Nothing enters build until it is in this list.
- Finishing before starting. When someone has space, they first help move an item already in progress, such as reviewing a colleague's draft or chasing a blocker, before pulling a new one. This keeps the stage limits from becoming individual lists.
- An honest answer for requesters. A requester whose work is waiting in the ready queue can be told where it sits and roughly what is ahead of it. The process hub explains how cycle timeThe elapsed time from an accepted request to live work, read stage by stage to show where work waits; a process measure, not a target for individuals. shows how long that wait tends to be, from the team's own history, and the tutorial on setting service levels for marketing requests shows the same queue from the requester's side.
Urgent work still happens. The prioritising guide's urgent route lets a genuinely urgent request jump the queue, and when it does, something in progress is paused or the limit is breached on purpose and recorded. Recording the exceptions matters: if they become routine, the urgent route is being used as a way round the queue, and that is a prioritisation problem to fix upstream. The troubleshooting page on a marketing plan that keeps changing traces that kind of churn to a plan that never fitted the team's capacity.
For a marketer working alone, the same logic applies at a scale of one. The guide to working as a marketing team of one covers choosing what not to do; a personal limit on how many pieces are open at once is the operational half of that choice.
Reviewing the limits
Limits are a starting point, not a rule set in stone. Review them at a fixed point in the team's routine, using the reports the work system already produces: items finished per period, time from request to launch, the age of open items and where queues form.
| What the records show | What it suggests | What to try |
|---|---|---|
| A queue keeps growing in front of one stage | That stage is the constraint | Lower the limit upstream; look for work the constraint does that others could do |
| People are idle while items wait elsewhere | Limits are uneven, or skills are too narrow | Rebalance the limits; let people help in the stage that is full |
| Many items blocked for long periods | Dependencies outside the team are the problem | Raise them with the owners; bring dependencies forward into the brief |
| The limit is breached often for urgent work | The urgent route has become the normal route | Revisit prioritisation and the urgent route, not the limit |
| Items are split or closed early to stay under the limit | The limit has become a target | Restate that it is a team agreement; review time from request to launch instead |
Change one thing at a time, and give each change long enough to show its effect before changing another. A marketing team retrospective is a natural place to read a limit change back, since it already checks each change the team tried against the evidence. Where the records cannot show where work waits, the tutorial on mapping a marketing process traces recent work step by step before any limit or stage is changed. Where a process change touches several teams, such as moving a review earlier, the marketing management guide sets out where those decisions sit in the team's operating rhythm.
A wildlife charity limits its open work (illustrative)
Snorvadi Wildlife Trust limits its open work
Snorvadi Wildlife Trust, an invented wildlife charity, has a marketing team of 4 who handle fundraising appeals, supporter emails, the website, social media and the printed magazine. Supporters and trustees kept asking why appeals took so long to reach them, and the team felt it was working flat out.
The marketing manager counted everything open, including requests held in inboxes and work sent to the magazine's freelance designer. The board showed 17 items; the full count was 38. Over the previous month the team had finished 9. Of the 38, 14 sat in build and 11 in review, most waiting for the head of fundraising or the conservation team to check facts.
The team agreed limits of 6 items in build and 4 in review, kept the regular supporter email and the magazine on the board as items, and moved 8 postponed projects to a parked list with dates to look at them again. New requests went into a ready queue ranked by the manager. When review was full, the person who had just finished a draft helped clear a review item or chased a blocked fact check instead of starting something new.
For the first 3 weeks it felt slower, because nothing new started while review was full. The head of fundraising agreed to a short review slot twice a week instead of reviewing when time allowed. After 2 months the count of open items had fallen to 19, the team was finishing 13 items a month, and the time from an appeal being requested to its first email reaching supporters had fallen from about 7 weeks to about 4. Nobody worked longer hours; less of their work was waiting.
Common mistakes
- Counting only what is on the board. Work in inboxes, personal lists and agency queues is still open, and leaving it out makes every limit look generous.
- Copying someone else's limit. Limits come from the team's own counts and flow, adjusted by observation, not from a number seen elsewhere.
- Treating the limit as a personal target. People then close items early, split them or keep them off the board, and the count stops meaning anything.
- Letting blocked items fall out of the count. Blocked work that does not count invites new work to start beside it, which is how half-finished piles form.
- Starting something new because the current item is blocked. Help clear the block or help a colleague finish instead.
- Leaving parked work on the board. It inflates the load and hides what is really in progress; move it to a parked list with a reason and a date.
- Using the limit to decide what is accepted. Acceptance is prioritisation at intake; the limit only decides when accepted work starts.
- Changing several things at once. When limits, stages and review slots all change together, nobody can tell which change helped.
