Back-office outsourcing bottlenecks: WIP, work item age and flow time
The number of completed tasks alone does not show where a process slows down. Learn to track work in progress, the age of the oldest items and flow time, then agree corrective action with your provider.
A process can meet its monthly target while some cases wait for weeks. An average hides the oldest items; a closure count does not tell you how much work remains in progress or where a document is stuck. In outsourcing, the client and provider need a shared view. Otherwise each party optimises its own step while delays pass downstream.
Flow management makes that picture visible. The official Kanban Guide, May 2025 edition, defines work in progress (WIP), throughput, cycle time and work item age. These are not labels for a board; they help detect a growing queue before it becomes a complaint, missed deadline or costly overtime recovery.
Five measures that reveal flow
| Measure | Meaning | Decision it supports |
|---|---|---|
| WIP | Started but unfinished cases | Is the team starting more work than it can finish? |
| Work item age | Elapsed time from the start of an open case to now | Which active item needs attention first? |
| Throughput | Cases completed in an agreed period | How much work does the system finish against demand? |
| Cycle time | Elapsed time from a case starting to completion | How long does the customer wait from process start to result? |
| Queue age | Waiting time before work starts, preferably separate from active work | Is the constraint in intake, prioritisation or execution? |
Agree exactly when an item is “started” and “finished”. For correspondence handling, that might mean receipt into a register and hand-off to the named recipient; for an invoice, successful registration and routing for approval. If client and provider reports use different definitions, the metrics cannot be compared meaningfully.
WIP and throughput help locate the constraint
A large number of open cases is not automatically a problem if they are recent and have a planned next step. The warning is a trend: WIP rises over several weeks while throughput stays flat or falls. Check whether the constraint is an incomplete input file, missing approval, manual re-keying, absence of a decision-maker or too many parallel tasks.
Little’s Law relates average WIP, throughput and the average time work spends in a system: WIP = throughput × average flow time. In a stable process, using consistent units and a matching period, 30 cases in progress and an average of 10 completions a day correspond to roughly three days of average flow time. This is a relationship between averages, not a promised deadline for an individual case. John D. C. Little published the original proof in A Proof for the Queuing Formula: L = λW.
Limit started work instead of only asking people to go faster
When everything is “urgent”, teams often start many cases at once, interrupt one to begin another and increase waiting time for all of them. Set a WIP limit for stages where overload is visible. Once the limit is reached, a new case should not automatically be started: finish an item, remove a blocker or agree together which exception takes priority. This is a pull approach: work starts when capacity and a clear opening are visible.
Do not copy a limit from another company. Start with a baseline, run a short pilot and check whether fewer started items improve flow without reducing quality or exception handling. A limit should expose a constraint, not hide demand or reject agreed work.
Show open-item age as a distribution, not just an average
Report the count and value of cases by age band, plus the oldest case, median and a high percentile of cycle time. Choose bands that fit your SLA and process; there is no universal threshold. Separate active work from waiting for the client, courier, postal operator or an internal decision. Otherwise a provider may appear late because of a blocker outside its control—or the reverse.
Add blocker status, reason, owner of the next action and the next update date. A case waiting for client information stays visible but should not be counted as active provider work. Clock-stop and restart rules must be agreed before reporting begins.
A practical outsourcing setup
- Define a work item. Agree what one case is, when it enters the queue and what error-free completion means.
- Map the real stages. Include intake, data checks, execution, approval, dispatch and exceptions. Show waiting as its own state.
- Build a baseline. For several weeks, measure WIP, open-item age, throughput, cycle time and rework; segment different case types.
- Set priorities. Combine deadline, business impact and risk. FIFO works only when cases have comparable priority.
- Pilot a WIP limit. Start with an overloaded stage and agree who can change the limit and how urgent exceptions are handled.
- Set escalation triggers. For example, an item exceeds the agreed age, the queue grows for several periods, or a blocker has no owner.
- Review causes, not blame. A weekly client-provider review should end with an action owner, due date and a check of the result.
Separate client and provider responsibilities
The provider should maintain current statuses, follow agreed priorities, flag blockers and suggest improvements. The client owns decisions requiring its authority, complete inputs, approved exceptions and named contacts. The RACI and SLA should explain how waiting for a client decision is counted and what happens to an urgent case when its owner is absent. One shared queue with shared definitions is better than two conflicting reports.
Where can these measures help?
In correspondence and e-COK workflows, measure the time from registration to scanning, assignment and hand-off. In finance and accounting processes, track invoices waiting for review or approval. In Credit Controlling, track unresolved queries and payment promises. Measures should support diagnosis and decisions, not false precision or incentives to close cases too early.
If work is growing while the team keeps firefighting, first find where the queue forms and which decision blocks flow. BackOffice Outsourcing handles correspondence, document workflows and selected finance and administration processes; scope and reporting are agreed for each process. Talk to us about your work queues.
Sources
- The Kanban Guide, May 2025 — WIP, throughput, work item age and cycle time definitions
- Open Guide to Kanban, July 2025 — practical flow measurement
- John D. C. Little, A Proof for the Queuing Formula: L = λW, Operations Research
Frequently asked questions
How is WIP different from a queue count?
WIP includes cases that have started but are not finished. A queue may also include items waiting to start. Report the groups separately.
Does a WIP limit guarantee faster delivery?
No. A limit reveals overload and prevents too much work from being started, but the underlying blocker must still be addressed and quality protected.
How should we set a WIP limit?
Use actual flow data, constraints and a pilot. There is no single value that fits every process.
Does waiting for the client count in provider cycle time?
That depends on the agreed definition. Report total case time and separately show active work and waiting for client decisions; clock-stop rules must be explicit.