Agency Work-in-Flight Limits for Client Delivery

The short answer

A useful agency work-in-flight limit is not “three projects per person.” Projects vary too much. Set limits at the stages where work actually waits, such as strategy review, production, founder review, client approval, and launch.

Start with an observed limit for one delivery cycle. When a stage reaches its limit, the rule is to finish, unblock, or deliberately swap work before starting another item.

What counts as work in flight

Count an item when the agency has accepted responsibility and someone expects progress. Include active production, internal review, rework, and work waiting on a near-term agency action.

Keep client-blocked work visible, but do not let it consume the same limit indefinitely. Give blocked work its own lane with an owner and next review date. Otherwise the board looks full while the team quietly starts more work around it.

Choose the stage that actually constrains delivery

The work-in-flight limit record

Copy the limit record

Delivery stage:

Items currently in this stage:

Observed completion rate per week:

Oldest item and age:

Provisional limit:

What counts toward the limit:

What happens when the limit is reached:

Who can approve a swap:

Blocked-work lane and review cadence:

Review date for changing the limit:

Call the first number provisional. A WIP limit is an operating rule to test, not an industry benchmark. Raise or lower it only after observing completion, queue age, quality, and missed commitments.

A practical starting method

1. Map five to eight real delivery stages

Use the states work already moves through. Avoid a detailed process map that nobody maintains.

2. Count current items and age

The oldest item often reveals more than the total. A queue of four that clears daily is different from a queue of four that has not moved for two weeks.

3. Limit the bottleneck, not every column

Start where work waits or switches most. One meaningful limit is more useful than a board full of decorative numbers.

4. Define the stop-start rule

When the limit is reached, the team must finish, unblock, reassign, or approve a visible swap. It cannot simply hide the next item in another status.

5. Review after one delivery cycle

Keep the limit if completion becomes more predictable without harming quality. Adjust it if the queue shifts elsewhere or the rule is routinely bypassed for a valid reason.

Worked example

A six-person agency has eight client deliverables in production and five waiting for founder review. The team keeps starting work because production appears available, but the founder-review queue grows.

The agency sets a provisional founder-review limit of three items. When the queue reaches three, the team finishes review-ready work, gives a qualified reviewer authority, or moves the promised date before starting another item that needs founder approval. After two weeks, it reviews queue age and missed dates. The number is not universal. The explicit stop-start rule is the useful part.

Do not use the limit as a performance target

A limit is a ceiling, not a quota. Do not keep a stage full to make people look utilized. Empty space can mean work is flowing. The useful outcomes are shorter queues, clearer tradeoffs, fewer hidden starts, and more reliable client commitments.

Related Qubit resources

Sources and operating boundary

This guide does not provide a universal WIP benchmark. The limit must be tested against the agency's work shape, skills, quality controls, and client commitments.

• Atlassian: working with WIP limits
• Planiflow: capacity planning for agencies
• Awork: workload planning in agencies