Improve Workflows by Managing Bottlenecks

Bottlenecks are a common source of frustration in organizations. A company’s legal review process may delay the execution of high-value contracts, or a shortage of computing resources may slow progress on a new digital initiative. Such constraints can cause tasks to pile up and hinder teams’ abilities to move forward with their work, costing companies time and money. When leaders encounter a bottleneck, they may dedicate resources to addressing it only to find that the process in question is still stalled by other bottlenecks. Our research has found that organizational bottlenecks can be best managed or avoided not by addressing them piecemeal but by taking a holistic view of work systems and resource portfolios and aligning them in ways that improve organizational performance.

Bottlenecks manifest as tasks that are stalled for one of two reasons: because they depend on the output of other tasks that have not been completed, or because the resources required to complete the task are not available. Task bottlenecks frequently occur as teams wait for approvals from legal or compliance departments, for example. Resource bottlenecks happen when there is a lack of resources necessary to complete a task or process — say, if a construction project has only one crane available and there are competing demands to use the crane.

Task bottlenecks may emerge when more time is needed to complete the activity than has been budgeted, even if all necessary resources are available. While resource bottlenecks can be solved by investing in more resources, task bottlenecks cannot necessarily be solved by simply throwing more resources at the problem. Let’s explore the nuances of these two forms of bottlenecks.

Task Bottlenecks: Centralization and Complexity

Task bottlenecks are dependent on two factors in a work system: centralization and complexity. Centralization refers to the degree to which a work system contains tasks that serve as connecting mechanisms between other tasks in a sequence, creating dependencies between activities, or, more specifically, the extent to which many activities within a system flow into or out of a few other tasks in the system.1

A hospital in which every specialty has its own blood lab is an example of a decentralized system. In contrast, a hospital with a single blood lab that handles all specialties is a more centralized system. Longer processing times could be a downside of having a centralized blood lab, but there could also be coordination benefits, since running only one test and then sharing the results with multiple groups would be more efficient than having each specialty request the same test separately and could help facilitate more holistic patient care.

Although centralization tends to create more bottlenecks, it can have benefits such as the ability to deploy resources efficiently, ensure consistency in quality, and coordinate efforts across multiple groups to achieve a shared goal. To evaluate these trade-offs and choose the appropriate level of centralization, managers also need to take the complexity of the activity system into consideration.

Complexity is determined by the number of tasks in a work sequence before it repeats.2 For example, a manager comparing multiple bids on a project completes a set of tasks to evaluate one bid before starting the process again and repeating the sequence of tasks to evaluate the other bids; in comparison, a manager evaluating a potential acquisition completes a much longer sequence of tasks before repeating it for the next acquisition-target appraisal. As sequences become longer, there is a greater chance that any given task in the cycle may fail to be completed, resulting in a greater ripple effect that will impact the entire cycle of activities by lengthening the time it will take to restart the cycle. Complexity rises when there is more variability in the tasks within the cycle, with some more difficult than others, resulting in a varied likelihood of task completion.

In the health care sector, treatment for a common ailment such as appendicitis would be a less-complex work sequence compared with diagnosing and treating a rare infectious disease, since the former involves a shorter sequence of tasks, each of which is well understood and likely to be successfully completed.

As companies organize their work, they not only decide what activities each cycle of tasks will comprise (that is, its complexity) but also which tasks will be run in parallel versus in sequence (that is, its level of centralization).3

Our research has found that rather than managing task bottlenecks only by decentralizing the system or reducing its complexity, an optimal work system aligns them — that is, it matches low complexity with centralized workflows, and high complexity with decentralized ones.

When companies have highly complex work involving many similar tasks, there is greater efficiency if tasks can occur in parallel (in other words, be more decentralized) so that there is less risk of blockage in the system. If a potential acquisition is being stalled because the manager tasked with evaluating acquisitions is still working on a prior deal, the system can be decentralized by having another manager evaluate acquisitions in parallel. On the flip side, when there is low complexity with shorter cycles comprising similar tasks, there is less that may go wrong before the cycle concludes and then starts over. This simpler, less-complex system requires less coordination and can remain centralized, which is less costly than running redundant tasks in parallel.

Resource Bottlenecks: Fungibility and Slack

Resource bottlenecks occur when tasks cannot be completed either because necessary resources are not available (there is not enough slack) or available resources are not applicable (there is not enough fungibility).

Fungibility is the degree to which a resource can be applied for multiple uses and contexts sufficiently well to complete the task.4 Low fungibility — when a resource is limited in its applicability to other tasks — is often referred to as specificity. For example, in the health care context, a dialysis machine, which is primarily used to treat kidney disease, is less fungible than an infusion pump, which can be used to deliver a variety of fluids to patients, such as medications and nutrients.

Slack is the extent to which resources are available to be used at any point in time.5 When there are slack resources, the organization has a buffer to draw on during unpredictable demand surges. Slack may ebb and flow: It may be more difficult to gain access to a crane in summer, when there are many construction projects, whereas in winter, when there is less construction, cranes may be readily available.

While the benefit of having slack is obvious, there are downsides to having too much in a company’s resource portfolio, because idle resources still incur costs. So how should companies manage slack in their resource portfolios?

Companies need some degree of both fungibility and slack to avoid resource bottlenecks. An organization is not helped by having either an available resource that is not applicable to the task at hand or a resource that could be used in another context but is unavailable.

For example, a consulting firm may have various client projects across a range of sectors. The firm’s resources are the consultants it employs with varied specializations. If it needs to assign a consultant to a banking case but the only available consultant is a specialist in the energy sector, that person is not fungible to be allocated to this task. The firm needs someone who is both available and a banking expert. If a second consultant has expertise in both automotive and banking sectors, this person would be considered fungible across these sectors and a suitable candidate for this assignment. However, if this second consultant is not available because they have been assigned to a different case, their fungibility is not helpful in this situation.

Our research finds that as companies have more slack resources, their performance improves only when they also have resources that are more fungible. Further, we find that organizations with fungible resources improve performance only when these are also slack resources.

Aligning Task and Resource Characteristics

We have highlighted that managers need to consider the alignment of task systems’ decentralization and complexity, as well as the alignment of resource fungibility and slack. Because resources are needed to accomplish tasks, and task completion makes a resource available for a subsequent task, task bottlenecks and resource bottlenecks affect one another.

Even when organizations are well aligned on centralization and complexity, any resource bottlenecks in the system will result in task bottlenecks, since the resources that are needed to complete some tasks cannot be allocated to them. Similarly, even in cases of alignment between resource fungibility and slack, a misalignment of decentralization and complexity will result in resource bottlenecks, since the lack of task sequence completion keeps the incomplete tasks’ assigned resources occupied and prevents these resources from being reallocated to other tasks.

Thus, in our study, we also considered how these four characteristics interact and how they should be managed to improve performance. Recall that companies perform better when the characteristics of their task system are aligned, with either low decentralization and low complexity (a low-low condition) or high decentralization and high complexity (a high-high condition). But we found that companies perform even better when the characteristics of both tasks and resources are aligned. Our results show that better-performing task systems (those with low-low or high-high conditions) are further improved by having applicable and available resources.

For example, in December 2022, the U.S. experienced a severe weather event that significantly affected air travel. All major carriers experienced weather-related flight cancellations during this period. Unlike its rivals, Southwest Airlines experienced a sizable ripple effect that extended beyond the weather event, resulting in almost 3,000 carrier-related cancellations on Dec. 26, followed by three more days of cancellations in the thousands. (See “The Ripple Effect of Bottlenecks.”)

Although all of the airlines faced inclement weather in late December 2022, Southwest faced the greatest difficulty recovering from this setback. Unlike its rivals, which operated on a hub-and-spoke model, Southwest operated on more of a point-to-point model at the time. Southwest’s planes were distributed across many airports, so some locations that should have had planes and crews in position had none, causing severe bottlenecks. Because cycle lengths were long (that is, complexity was high), getting these resources back into position to resume their cycles (without the availability of slack resources) resulted in an increased recovery time that affected its entire activity system.

How Companies Can Manage Bottlenecks

What can be done to manage the different types of bottlenecks that may arise within organizations? Based on our analysis of tasks and resources, here are our recommendations.

1. Locate and identify specific bottlenecks in the system. Managers should be continuously scanning for bottlenecks within their purview. They should identify whether the observed bottlenecks are task bottlenecks or resource bottlenecks and determine their sources. In most cases, a task is unable to start either because it is waiting for the output of another task (a task bottleneck) or because the appropriate resource needed to complete the task is not available (a resource bottleneck).

2. Evaluate task bottlenecks in the work system. Managers should evaluate whether a task bottleneck, while slowing things down, has the benefit of aggregating information for better coordination, decision-making, or quality assessment. In cases where there are no such benefits, managers should determine whether the bottleneck resulted from poor communication regarding the length of time needed to complete prior tasks; if so, they should facilitate better information sharing between the people responsible for different tasks in the sequence to improve timely coordination.

3. Determine complexity, and redesign centralization. After evaluating each individual task bottleneck, managers should evaluate the entire work system to identify the cycle of sequential tasks within which the bottleneck is occurring. The longer the cycle of sequential tasks, the greater the complexity, meaning that there is room for failure at more points within the system that could cause a slowdown or potential stoppage. Short task sequence cycles should work with a high degree of centralization. If the task bottleneck is within a sequence that is long and often experiences delays, managers should consider adjusting the work system to have more tasks run in parallel, thus making the system more decentralized, or shorten the sequence to reduce complexity. Running activities in parallel by dividing workers into subgroups (where the subgroups could be either working on different tasks or performing duplicate tasks) allows one subgroup to complete its work if the other gets stalled. An example is when software R&D projects are modularized to work independently of one another before integration.

4. Evaluate resource bottlenecks in the work system. Having addressed task bottlenecks, managers should next turn their attention to the resources needed to accomplish tasks. Specifically, for those tasks that were observed to have delays, managers should determine whether the task system design is to blame (in which case, they should follow the steps above) or missing resources are impeding task completion.

5. Determine why resources are missing, and then design more slack or fungibility into the system. Once delayed tasks that are missing resources have been identified, managers should investigate whether these necessary resources are unavailable or not applicable. If a critical resource is unavailable, then managers should include more slack in the system so that these types of resources are available to mitigate uncertainty. If companies design operations too lean, with very little slack, they run the risk of severe disruptions. For example, during a major ice storm on Valentine’s Day in 2007, JetBlue experienced an operational meltdown that affected over 130,000 customers and ultimately resulted in the airline’s customer bill of rights. That example illustrates the danger of not having enough personnel available to handle both thousands of simultaneous rebookings and the reassignment of flight crews in response to cancellations, delays, and diversions.6 However, managers should be aware that excess slack — or too much redundancy — increases costs and should also be avoided.

On the other hand, if managers find that the resources at hand cannot be applied to the task, this indicates a lack of adequate fungible resources. In that case, managers should add resources to their portfolio that can serve more than one purpose. For example, in the airline industry, low-cost carriers such as Southwest and JetBlue have chosen to have fewer aircraft types to simplify pilot training; this increases the fungibility of applying plane-crew pairings to completing their tasks of providing air transportation to customers. In the food industry, many restaurants now cross-train employees on both customer-facing and non-customer-facing tasks and have seen improvements in productivity, job satisfaction, and customer service.7

It is important that managers not only observe bottlenecks in their organizations but also empower their employees to identify and report them. Keeping the principles pertaining to task bottlenecks and resource bottlenecks in mind, companies can move toward more effectively designing their work systems and resource portfolios for improved performance.