The Problem With Ranking Fulfillment Centers by Units per Hour

Ranking fulfillment centers by units per hour is appealing because the metric is simple, comparable and easy to explain, but that simplicity can create false conclusions when different sites process different products, order profiles, service commitments, process paths, automation levels and operating constraints. A facility handling dense, high-velocity small items through automated zones may naturally produce far more units per labor hour than a site handling bulky, low-velocity, multi-line or exception-heavy work, even if the second site is equally well managed or better aligned with its customer promise. When executives use raw UPH as a cross-site league table, they risk rewarding easier operating conditions, penalizing strategic complexity and encouraging local behaviors that improve the denominator while degrading quality, cycle time, service or total cost. The more useful question is not which site has the highest units per hour, but whether each site is converting its specific workload and asset base into reliable customer outcomes at an economically appropriate level of productivity.

Why Units per Hour Is So Attractive to Management

Large fulfillment networks need normalized metrics because absolute volume alone tells executives very little about efficiency. A facility processing 120,000 units per day is not necessarily more productive than one processing 40,000 and labor cost cannot be interpreted without knowing how much work was completed. Units per hour therefore provides a convenient common denominator that appears to make different facilities comparable, and it is useful for observing trends within a relatively stable operating environment where the product mix, process path, service commitment and technology configuration do not change materially.

The problem begins when the measure is asked to explain more than it actually contains. UPH compresses the entire operating environment into two numbers: units completed and labor hours consumed. It does not describe how far employees traveled, how many order lines were handled, whether the units were each-picks or cases, how much replenishment was required, how many exceptions occurred, whether orders needed consolidation, whether automation performed part of the work or how much capacity had to be preserved for a late customer cutoff. Two sites can therefore report very different UPH values even when the difference is caused primarily by workload structure rather than management quality.

Work Mix Can Move the Metric Without Any Change in Execution

A fulfillment center can improve or deteriorate on UPH even when the workforce, process design and management discipline remain unchanged because the mix of work changes. A day dominated by single-line orders from dense forward locations will produce a different labor profile from a day containing larger baskets, more multi-zone orders, more oversized items or higher exception volume. If the metric is reported only as a network average, management can interpret mix-driven movement as performance movement and initiate corrective actions against a problem that does not actually exist.

This becomes more consequential in cross-site comparisons because facility roles are rarely identical. One site may carry fast-moving national inventory and process highly standardized parcel orders, while another supports regional B2B demand, special handling, returns or slow-moving inventory. A raw ranking effectively assumes that one unit represents the same amount of work everywhere, which is often untrue. A more defensible comparison normalizes the workload into meaningful operating segments or uses expected labor content so management can distinguish how difficult the work was from how efficiently the site executed it.

Automation Can Make One Site Look Better Before Management Has Learned Anything About Operations

Automation creates another source of distortion because the labor denominator changes even when the total operating system becomes more capital intensive. A goods-to-person system can raise direct picking UPH substantially by moving travel out of the employee's work content, while the facility takes on equipment depreciation, controls support, maintenance, software and technical labor elsewhere. Comparing that site directly with a manual facility on direct UPH can therefore reward the transfer of work from labor to capital without showing whether total cost, service resilience or return on invested capacity is actually better.

The correct interpretation is not that automated and manual sites should never be compared, but that the comparison must reflect the operating model. If the executive question concerns labor productivity, the automation difference is relevant and should be explicit. If the question concerns total fulfillment economics, the comparison must include the cost and utilization of the capital that made the labor metric possible. If the question concerns customer performance, service and throughput must be considered directly. One number cannot answer all three questions.

High UPH Can Coexist With Poor Fulfillment Performance

Units per hour is an activity-efficiency measure, not a complete measure of fulfillment performance. A site can improve UPH by increasing batch sizes, concentrating work in the easiest zones, deferring exceptions or releasing work in ways that keep employees continuously busy, while the operation accumulates work-in-process downstream and misses customer cutoffs. The metric can therefore rise at the same time as order cycle time, overtime, rework or premium transportation deteriorate.

This is particularly dangerous when managers are compensated or ranked primarily against the metric because rational teams optimize what the organization visibly rewards. Picking leaders may maximize dense work and leave difficult orders for later, packing leaders may favor easy cartons or sites may classify indirect work differently to protect the denominator. None of these behaviors requires deliberate manipulation; they emerge naturally when a narrow measure becomes the dominant definition of success. Executives should therefore treat UPH as one diagnostic layer within a broader operating scorecard rather than as a standalone league table.

Indirect Work Changes the Denominator More Than Most Rankings Admit

Cross-site rankings are also vulnerable to differences in what each operation counts as direct and indirect labor. One facility may include replenishment, problem solving, startup time, equipment checks or training inside the measured labor hours, while another allocates those activities to separate cost centers. Both sites can follow their local accounting rules correctly and still produce productivity numbers that are not operationally comparable. Even when definitions are formally standardized, differences in process design can move work outside the measured activity, making one site's rate appear structurally superior.

For governance, this means the metric definition must include more than the numerator and denominator formula. Management should know which labor categories are included, when the clock begins and ends, how downtime is treated, how training is allocated and whether rework is counted as productive output. Without those rules, network rankings can create false precision because the organization has standardized the label without standardizing the measurement boundary.

Service Commitment and Quality Must Sit Beside Productivity

A site serving later cutoffs, premium promises or higher-complexity customers may deliberately preserve more spare capacity than a site operating a more predictable service model. That reserve can reduce average labor utilization and UPH while increasing resilience when demand arrives late or variability increases. Ranking the site negatively for the lower rate can encourage management to remove exactly the flexibility that protects the customer promise. The same issue applies to quality: driving more units through a process is not valuable if mis-picks, damage, recirculation or rework rise enough to consume the apparent productivity gain.

A balanced view should therefore examine productivity together with service attainment, quality, flow and cost. If UPH improves while on-time shipment and first-pass quality remain stable or improve, the productivity gain is more credible. If UPH improves while downstream backlog, exception volume or premium transportation rises, management should investigate whether the operation has merely transferred cost or delay elsewhere.

Normalize for Work Content Before Comparing Sites

The strongest network comparisons begin by identifying the characteristics that materially change labor content and then controlling for them. These may include units per order, lines per order, travel distance, SKU velocity, product dimensions, process path, automation level, service tier, replenishment intensity and exception rate. The objective is not to create a model so complicated that no one can use it, but to separate the few structural drivers that explain most of the difference in labor requirement.

One practical approach is to establish expected labor content for major work profiles and compare actual performance with the expected requirement rather than comparing raw site rates. A site processing a difficult mix can then outperform expectation even if its absolute UPH remains below the network average, while an apparently high-productivity site can be identified as underperforming relative to a much easier workload. This changes the management conversation from ranking outputs to understanding operational efficiency.

Use Site Rankings to Ask Questions, Not to Declare Winners

Network rankings can still be useful because variation often reveals opportunities to investigate. A site that consistently performs differently from peers may have better slotting, stronger labor planning, superior process discipline, more favorable automation or simply a different workload. The ranking should therefore identify where management needs explanation rather than assume that the explanation is already contained in the number. Site visits, process analysis and normalized data can then distinguish transferable practices from structural differences.

This matters culturally as well as analytically. When rankings are treated as verdicts, local teams spend time defending their position and explaining why the comparison is unfair. When rankings are treated as a structured starting point for learning, sites are more likely to share process innovations, expose constraints and collaborate on common standards. The metric then supports continuous improvement instead of becoming a source of metric management.

Conclusion: Compare the Operating System, Not Just the Rate

Units per hour remains a valuable fulfillment metric because organizations need a way to understand how effectively labor converts into output, but its usefulness declines when executives treat it as a universal measure of site quality. Different order profiles, process paths, automation levels, service commitments and measurement boundaries can produce materially different rates without indicating that one management team is better than another. A raw network ranking can therefore reward structural advantage, penalize necessary complexity and create incentives to optimize local activity rather than customer outcomes.

The better approach is to preserve UPH as a transparent productivity measure while surrounding it with the context required for interpretation: normalized work content, common labor definitions, service and quality outcomes, flow measures and total economics. Executives can then use variation between facilities to ask why performance differs and which practices are genuinely transferable. The objective is not to make every site produce the same units per hour; it is to ensure every site converts its particular workload, assets and operating conditions into the strongest possible end-to-end fulfillment outcome.