Portolan Operations decision desk
Open Portolan
All guides
Guide · Capacity, flow and scheduling

Hire more support agents or fix ticket routing? Why queues grow when everyone is busy

Decide with three numbers, not an instinct about headcount: how fast work arrives, how long it sits in the system, and how much is waiting. A queue can keep growing while everyone is busy. Waiting time rises non-linearly as utilisation nears full load, and variability makes it worse. More people help where the capacity lands on the step that limits output, at the hours work arrives. Routing helps when the limit is how work is distributed.

Published · Updated

Start with three numbers

In a stable system the average amount of work in it equals the average arrival rate multiplied by the average time each item spends there. Any two of those numbers fix the third. If tickets arrive at a known rate and each waits a known time, the backlog is already determined, and cutting the wait means cutting the work in progress or raising throughput.

That makes the first task measurement rather than debate: arrivals per day, time from arrival to resolution, and the open count. The law holds for averages in a stable system and says nothing about peaks or about demand that is changing, so treat it as the starting frame, not the whole answer.

Why the queue grows while everyone is busy

Busy and keeping up are different things. Queueing analysis shows that waiting time rises non-linearly as utilisation nears one, and that variability in arrivals and in handling time amplifies the delay. A team running comfortably at 80% load can end up with long queues at 95% without any change in average demand.

Variability has to go somewhere. It is absorbed by some combination of inventory, capacity and time, and in a support queue the buffer that is left is the customer's wait. You choose which buffer to pay for: spare capacity, smaller or more predictable work, or longer waits. You cannot remove all three.

When routing is the fix, and when hiring is

Output is set by the constraint. Improving a step that is not the constraint does not raise output, and local efficiency can pile work up in front of the real bottleneck. So find the step that genuinely limits throughput first: a specific skill, an escalation hand-off, an approval, a shift gap.

If the limit is how work is distributed, for example senior agents saturated while others are idle, routing and escalation rules move work to where capacity exists. If every step is near full load, there is no routing change that creates capacity and hiring, or removing demand, is the lever.

Timing matters as much as total hours. In a before-and-after study in one emergency department, queueing analysis was used to reshape provider schedules around the pattern of arrivals within the day, and the share of patients leaving without being seen fell by about a fifth, including when existing hours were reallocated without adding any. That is a hospital, so apply it by analogy: staffing to the daily average can hide the peak that creates most of the waiting.

What running a team hot costs

Teams often speed up as load rises, so a hot queue can look efficient for a while. In hospital data, sustained overwork was associated with slower service and worse outcomes. The data are observational and from hospitals, so they show an association and may not transfer, but they are a reason to treat sustained overload as an operating risk and to protect recovery time rather than read it as a motivation problem.

The path, step by step

These steps come from the Capacity, flow and scheduling playbook. Each has a checkpoint you can observe.

  1. Map the flow and find the constraint

    Walk the process end to end and measure work in process, wait time and throughput at each step. The constraint is the step that gates the whole flow, not the one that complains loudest.

    CheckpointOne named constraint, supported by measured WIP, queue time and throughput at each step.

    Common failureImproving the busiest-looking team without checking whether it gates the flow.

  2. Quantify demand, capacity and variability

    Measure arrival rate, service rate, the variability of each, and utilisation at the constraint. Estimate the waiting implied by that combination before choosing a fix.

    CheckpointUtilisation, variability and predicted wait at the constraint, with the assumptions written down.

    Common failurePlanning from monthly averages and concluding that capacity is adequate.

  3. Cap work in process and control release

    Set a WIP limit at the constraint and release new work only when capacity frees up. Start with a limit you can defend, then tighten it while watching throughput.

    CheckpointWIP stays at or below the cap, and flow time falls without throughput falling.

    Common failureSetting the cap so high that it never binds, then declaring that pull did not work.

  4. Match capacity to the arrival pattern

    Plot demand and waiting by hour and day. Move shifts, breaks, cross-trained staff and batch work into the peak before adding headcount.

    CheckpointWait time by hour before and after the rota change, with hours held constant.

    Common failureAdding the same number of hours to every shift and expecting peak waits to fall.

What this guide does not settle

The evidence here is queueing theory, factory and hospital studies. The research library has no controlled trial of hiring versus routing in a customer support team, so treat the result as a way to ask better questions of your own numbers, then test one change at a time.

What the research found

Each finding below is quoted from Portolan's research library, with the limits the library records for it. Applying a result from another setting, such as a hospital or a factory, to yours is an analogy, not a proof.

Little · 1961Established

A Proof for the Queuing Formula: L = λW

In a stable system, the average amount of work in the system equals the average arrival rate multiplied by the average time in the system (L = λW). Any two of the three quantities determine the third, independent of the detailed distribution of arrivals and service times.

Where it stops. The law holds for averages in a stable system and says nothing about variability, service-level tails or what to do when demand is changing.

Kingman · 1961Established

The single server queue in heavy traffic

The heavy-traffic approximation shows that waiting time rises non-linearly as utilisation approaches one, and that variability in arrivals and service times amplifies the delay. A system running comfortably at 80% load can collapse into long queues at 95% without any change in average demand.

Where it stops. A single-server, heavy-traffic approximation. Multi-server, priority and network systems need their own queueing models; the exact shape of the curve depends on the arrival and service distributions.

Hopp & Spearman · 2001Established

Factory Physics: Foundations of Manufacturing Management

Variability must be absorbed by some combination of inventory, capacity and time—the variability buffering law. High utilisation amplifies queues non-linearly, so the last few percent of capacity utilisation are extremely expensive in waiting time.

Where it stops. The laws are analytical and widely validated in manufacturing, but the book's models simplify real systems and require reliable process data.

Goldratt & Cox · 1984Applied

The Goal: A Process of Ongoing Improvement

Throughput is set by the constraint—the bottleneck—and improving anything that is not the constraint does not increase the system's output. Local efficiency can actively hurt by building inventory in front of the bottleneck, so the system should be managed as a chain rather than a set of optimised departments.

Where it stops. The core evidence is case-based and the book is a teaching narrative; constraints can be policy or demand rather than a physical step, and naive bottleneck analysis can miss them.

Green et al. · 2006Applied

Using queueing theory to increase the effectiveness of emergency department provider staffing

A 39-week before-and-after study in one emergency department used queueing analysis to reshape provider schedules around within-day arrival patterns. Patients leaving without being seen fell 22.9%; reallocating existing provider hours on selected days cut the same outcome by 21.7% without adding hours.

Where it stops. A before-and-after design in a single emergency department, not a randomised trial. Other simultaneous changes and seasonal effects cannot be ruled out.

Kc & Terwiesch · 2009Established

Impact of Workload on Service Time and Patient Safety: An Econometric Analysis of Hospital Operations

In hospital transport and cardiothoracic surgery, workers speed up as load rises, but sustained overwork is associated with slower service and worse care. For cardiothoracic surgery, a 10% increase in overwork was associated with a 2% higher likelihood of mortality.

Where it stops. This is observational hospital data, so the associations are not proof of causation and may not transfer to other settings.