Starts alone cannot tell you how quickly jobs are delivered
Moving every new request into progress can feel like a way to avoid keeping customers waiting. Yet if the available working time stays the same, each job receives a smaller share, and the first delivery can happen later. Count starts and deliveries separately.
Work in progress, or WIP, counts jobs that have started and have yet to finish. A job waiting for a customer reply or internal approval still counts if it has started and remains unfinished. The official Kanban guide defines work in progress using explicit start and finish points. (The Kanban Guide (May 2025))
A hypothetical example with one worker shows the difference between starting early and delivering early.
Finishing four jobs in sequence brings the first delivery forward
Suppose four equally sized jobs, A, B, C, and D, arrive at the same time. Each requires one full day of one person's work, and only one worker is available. Days here mean available working time expressed in day units; holidays are excluded. Assume no switching overhead, approval delays, or revisions.
For the shared approach, use an idealized model in which the worker's time is allocated equally across all four jobs and all finish together after four days. For the sequential approach, the worker delivers A after one day, followed by B, C, and D. Delivery happens as soon as a job is finished.
| Measure | Share time equally across four jobs | Finish one job at a time |
|---|
| Delivery of A | 4 days after receipt | 1 day after receipt |
| Delivery of B | 4 days after receipt | 2 days after receipt |
| Delivery of C | 4 days after receipt | 3 days after receipt |
| Delivery of D | 4 days after receipt | 4 days after receipt |
| Time to finish all four | 4 days | 4 days |
| Mean time from receipt to delivery | 4 days | 2.5 days |
Both approaches deliver four jobs in four days, averaging one delivery per day across the whole period. Nevertheless, the first delivery happens after four days in one approach and after one day in the other. In the sequential approach, customers A, B, and C receive their deliverables earlier, while D receives theirs at the same time.
The mean is (4 + 4 + 4 + 4) / 4 = 4 days when time is shared, versus (1 + 2 + 3 + 4) / 4 = 2.5 days in sequence. The same time to finish the whole batch can produce different waits for individual customers. These are calculations from stated assumptions, rather than measured improvements in a workplace.
Little’s law connects unfinished work with elapsed time
The average number of jobs inside a process relates to the rate at which jobs pass through it and the time they spend there. Little’s law states this relationship as average number = flow rate × average elapsed time. Little's own paper explains that the equality also holds over a finite observation period when the process is empty at both its beginning and its end. (Little’s Law as Viewed on Its 50th Anniversary)
Apply that relationship to the process from actual work starting to delivery in our example. Starting all four together keeps four jobs in progress throughout the four days, so time-averaged WIP is four. Working in sequence keeps one started, undelivered job in progress at any moment, so average WIP is one.
Both approaches deliver four jobs over four days. Dividing WIP by that delivery rate gives a mean time from work starting to delivery of 4 jobs / (1 job per day) = 4 days, versus 1 job / (1 job per day) = 1 day.
That one day excludes waiting before work starts. B waits one day after receipt, C waits two, and D waits three before starting. Even when using Little’s law, the event used as the starting point changes what the time means.
Include the wait before work starts in the customer’s wait
A shorter time after starting does not necessarily mean earlier delivery for customers. In the sequential example, the mean after starting is one day, while the mean after receipt is 2.5 days.
Suppose every start in that same sequential schedule is delayed by another two days. Each job still takes one day after starting, but deliveries now happen three, four, five, and six days after receipt. The mean from receipt to delivery becomes (3 + 4 + 5 + 6) / 4 = 4.5 days. The post-start measure remains one day, while customers wait two days longer on average.
Moving a job from an in-progress column back into a not-started column therefore does not establish that delivery has improved. Record receipt, actual start, and delivery timestamps to calculate both the wait before starting and the elapsed time after starting.
Assess delivery counts and customer waits alongside starts
In this example, starting every job together increases neither the number delivered over four days nor the speed of the first delivery. It delays the first delivery and increases the mean customer wait. Asking what an existing job needs for delivery is more useful than increasing the number reported as started.
Real teams can have different people responsible for different stages. With separate designers and proofreaders, stages can run in parallel, so a company-wide limit of one job need not be best. If proofreading is backing up, examine whether proofreaders have the required copy and whether answers to revision requests are delayed, alongside decisions about starting new design work. The Kanban guide describes pulling in work when capacity exists and removing obstacles from blocked work. (The Kanban Guide (May 2025))
When limiting starts, compare delivery counts and receipt-to-delivery times before and after the change, using comparable job sizes and the same time measurement. If idle time rises and deliveries fall, reconsider the limit or the allocation of work across stages.
When publishing production schedules on a website, specify whether the stated time begins at application or after the required materials arrive and work starts. Customers can then understand which waiting period the schedule describes.
SpreadSite is a no-code website builder that lets you create, publish, and update a website from Google Sheets using ready-made templates.
To explain how customers can request your service and what your production schedule means, give SpreadSite a try!
Build a website with SpreadSite →
