What's in a Queue? An Experimental Study of Job Ordering, Autonomy and Queue Visibility

How Queue Design Shapes Worker Performance

How should a manager organize a stack of incoming tasks? In service operations—ranging from warehouse fulfillment to IT helpdesks—the way a queue is arranged and presented to workers is a fundamental design choice. Traditionally, companies rely on First-in-first-out (FIFO) protocols. In FIFO, jobs are processed in the order they arrive. Alternatively, companies may grant workers the autonomy to choose their own sequence.

However, a recurring tension exists in the field. When workers are given freedom, they often "cherry-pick" easier jobs first. Previous studies suggest this behavior leads to lower overall productivity. A new study from Johns Hopkins University investigates the driver of this performance drop. Is it caused by the ordering itself, or is it a symptom of which workers choose to cherry-pick? The authors report that while easy-first (EF) ordering is associated with higher error rates under worker discretion, imposing an EF sequence externally reduces error rates by approximately 30% relative to the discretionary (Endog) condition.

The breakdown of discretionary ordering

In many modern service environments, workers operate with high autonomy. In a warehouse, this might mean choosing a picking route from a digital list. In a call center, it might mean selecting which ticket to resolve next. Many assume worker discretion is a neutral convenience. Others see it as a tool for managing variable workloads.

Existing literature suggests this assumption is flawed. The authors note that in settings like hospitals, workers who deviate from FIFO are often slower. A common interpretation is that worker discretion should be limited. This prevents "cherry-picking"—the act of skipping harder, older jobs to complete simpler, newer ones. As shown in, cherry-picked jobs in the study's experimental setting exhibited significantly higher error rates (17.2%) compared to those handled in FIFO order (9.0%).

This creates a dilemma. Is the problem the act of picking easy jobs, or the type of worker who prefers to do so?

Decoding the selection mechanism

To isolate the cause, the researchers developed a real-effort experimental task. Participants fulfilled virtual grocery orders of varying complexity. These included Small, Medium, and Large jobs. Jobs arrived dynamically over time. This allowed the authors to manipulate three levers: job ordering, worker autonomy, and queue visibility (the amount of information workers see regarding the queue).

The core investigation aimed to distinguish between a causal mechanism and a selection mechanism. A causal mechanism implies the act of deferring a hard job causes errors through fatigue or distraction. A selection mechanism suggests that error-prone workers are simply more likely to choose an easy-first sequence.

The researchers used several protocols. In the "Endog" condition, workers chose their own sequence. In the "Pre-commit" condition, workers committed to a rule before the round began. This removed their ability to react impulsively to new arrivals. Finally, in the "Exog" conditions, the researchers imposed the sequence. By comparing these states, the authors could determine if the performance gap was a property of the workflow or the workforce.

Evidence for imposed structure

The results reveal a divergence between worker behavior and imposed rules. In discretionary experiments, workers were "load-adaptive" (adjusting behavior based on workload). As workload increases, they speed up and cherry-pick more frequently. As shown in, completion times drop and cherry-picking rates rise following recent job arrivals.

Crucially, the authors find that the link between cherry-picking and errors is not necessarily causal. When they applied subject fixed effects (a statistical method to control for stable, individual differences), the relationship disappeared. This supports the selection mechanism. The error-prone workers are the ones opting for easy jobs.

The most significant finding for operational design involves the exogenous treatments. The paper finds that imposing an EF sequence reduces error rates by about 30% relative to the discretionary mean . This reduction is largely concentrated among lower-ability workers. They appear to benefit from a gradual increase in job complexity. Interestingly, imposing a sequence does not improve average speed. It can actually slow down the most capable workers. These experts lose the efficiency of managing their own queues.

Limits of visibility and personality

The study also investigated whether "visibility"—showing workers the full queue and arrival alerts—serves as a motivator. While the authors observed short-term productivity bursts after a new job arrival, long-term performance remained unchanged. Removing queue and arrival information did not affect average performance. This suggests that the "urgency" of a visible backlog is a transient effect. It does not translate into sustained aggregate gains.

Regarding worker screening, the authors explored if personality traits predict these behaviors. Most traditional measures, such as the Big Five, were weak predictors. However, the authors report a correlation between risk appetite and error-prone behavior. Workers with a higher risk appetite were more likely to both cherry-pick and submit jobs with errors. These workers likely gamble that a single error won't trigger a mandatory "redo" of the job.

The verdict on queue design

The decision on queue design depends on the organization's objectives and staff skill levels. If the goal is maximizing the speed of experts, maintaining autonomy is essential. However, if the goal is stabilizing quality, the authors' evidence suggests a different path. Moving from FIFO to an imposed, easiest-first sequence can be a powerful lever.

For practitioners, the takeaway is twofold. Treat flexible workflows as "earned autonomy" for top performers. For newer or less experienced staff, implement structured, complexity-increasing sequences. Regarding hiring, the study suggests that assessing risk appetite could be a useful addition to worker screening tests.

Figures from the paper