What Bus Bunching Teaches Us About Complexity and Control

I've often questioned why product teams don't take on more ambitious goals in transit tech. One explanation might be that tech teams tend to gravitate toward projects they can fully control with a technology-only approach—endeavors that don’t require the unpredictability of working with policymakers, operations leaders, elected officials, and the public. The irony is that the most transformative work can't happen without engaging in a meaningful way with these stakeholders and aligning everyone on a shared vision. This includes using actual data to gauge the effects of our activities—so we know when to persist or when to rethink. Sadly, in many transit agencies, it's all too easy to produce technology outputs that appear impressive internally without delivering results to the public. In this post, we'll explore how to make real change, drawing on proven strategies from the social impact sector.

More than fifty years ago, before the term “bus bunching” had entered the vocabulary of transit experts, Denos C. Gazis was quietly laying the foundation for our understanding of traffic flow—work that would later shape how we think about public transportation. At General Motors Research Laboratory, Gazis studied vehicle movement, exploring how minor delays could ripple through a traffic system, creating congestion, disorder, and unpredictability.

In 1972, Gazis authored a study introducing a novel way of thinking about following behavior—how individual cars “fight for survival” in a line on highways and how feedback loops control their movement. When the lead car slows, the vehicles behind it react in ways that aren’t always predictable. But Gazis’ work wasn’t just about cars; he had uncovered a fundamental truth about transportation systems that would also apply to bus routes. His findings would eventually help explain why buses tend to bunch up, even when their schedules are perfectly timed.

What made Gazis’ work so remarkable wasn’t just his ability to describe these phenomena mathematically; it was his insights, which have remained relevant for over half a century. Today, when two buses arrive together after a long delay, we witness the cascading effect of the principles Gazis described. Ironically, Gazis also theorized about using real-time data and computer systems to regulate traffic flow. We now have these systems in public transit, yet bus bunching remains a persistent frustration.

Bus bunching seems like the kind of problem engineers were born to solve. You have buses. You have a schedule. And you have the inevitable reality that they cluster no matter how well you plan. A technical problem, right? A problem just begging to be eliminated by machine learning.

So why can’t we solve it?

Ron Heifetz, author of Leadership Without Easy Answers, would say that bus bunching isn’t a technical problem at all. It’s what he calls an “adaptive challenge.” A technical problem has a straightforward solution—engineers can build a model, optimize a route, or create better software. However, adaptive challenges are different. They require people to change how they think, work, and interact, even when there is a significant technical aspect.

For example, brilliant transit researchers like Oded Cats have described holding control strategies and even how real-time passenger counts could reduce or eliminate bus bunching. However, to apply these strategies, transit agencies would need to allocate resources to the problem, use a workable change management approach, technologists would have to navigate uncertainty, and bus operations leaders and policymakers would need to lend their support. Of course, bus operator perspectives and buy-in would ultimately make or break any potential intervention.

And that’s the thing: technical solutions can only go so far in addressing the complicated problems that remain unsolved in public transit. The real challenge is adaptive: getting people to work differently. It requires coordination across various parts of the system, from on-the-ground staff to leadership. It demands a shift in how transit agencies operate, make decisions, whether municipalities support bus lanes, and even how riders engage with the service.

Adaptive challenges can be especially difficult for government agencies because they require a culture that enables bridging and risk-taking, which is risky in an overly rule-based or power-oriented setting.

Thankfully, there is a framework for agencies with the courage to tackle adaptive challenges.

Once we acknowledge that technology can’t provide a silver bullet that will magically solve everything, we can look for other approaches to solving systemic problems, and that is where frameworks like “collective impact” come into play.

John Kania and Mark Kramer, who first introduced the concept of collective impact, saw the same pattern emerging in complex social impact challenges like improving the education system in America. Their Stanford Social Innovation Review article described how organizations like Teach for America—well-funded, passionate, and resourceful—couldn’t make the kind of large-scale changes necessary to move the needle on education outcomes in America. Why? They were tackling a piece of the problem using an isolated impact approach where success depended on collective action—just like what we see with bus bunching, where single-intervention approaches have not worked.

Kania and Kramer go on to describe StriveTogether, another nonprofit that took a different approach. Instead of going it alone, they unified multiple stakeholders—schools, nonprofits, government agencies, and funders—to work together. They shared data, aligned resources, and created a common agenda. And the results? StriveTogether made and continues to make measurable improvements in education outcomes across entire communities—improvements that Teach for America, operating alone, couldn’t achieve.

This is the essence of collective impact: solving complex problems requires bringing together a variety of actors to work in alignment. It’s about recognizing one’s own “isolated impact” mindset and shifting to one that prioritizes collaboration across teams, departments, organizations, and the public. Like in education, where StriveTogether aligned stakeholders to create real change, public transit agencies must bring together technologists, policymakers, bus operators, advocacy groups, and even riders to develop a common agenda. Solving bus bunching isn’t just about better algorithms—it’s about creating the conditions for everyone involved, especially the public, to contribute to a shared solution.

That’s the power of collective impact. It turns individual efforts into collective progress. It’s the difference between Teach for America and StriveTogether. And maybe, just maybe, it’s the difference between a bus system that works and one that’s constantly falling behind.

To understand collective impact in a way that resonates with technologists and business leaders today, Let’s explore each of the five critical components through a business and tech lens:

1. Common Agenda: The Shared Vision

This is your North Star—the clearly defined problem and the collective goals everyone is working toward. It’s like setting a product roadmap for a cross-functional team. Without a shared understanding of the mission, departments (or organizations) could spin off into their own priorities, leading to misaligned efforts. For collective impact to work, all stakeholders—whether engineers, managers, or external partners—must agree on the common agenda.

2. Shared Measurement Systems: Data-Driven Accountability

Shared measurement systems, such as setting up a dashboard to track progress to ensure everyone's on the same page. For a team to work together effectively, everyone must agree on the data to track and what success looks like.

3. Mutually Reinforcing Activities: Complementary Collaboration

In collective impact, every stakeholder group focuses on what they do best, so product teams collect data, and advocacy groups engage the community. The key is that all efforts are connected and reinforce each other, driving toward a shared goal.

4. Continuous Communication: Seamless Cross-Team Coordination

Open communication ensures that the collective stays on the same page and promptly addresses problems. This is not just about impressive status updates; it’s about sharing insights, solving problems, and adapting to new information when it comes up.

5. Backbone Support Organization: The Dedicated Team of Facilitators

Think of this as a large-scale project’s product manager or program management office (PMO). Collective impact efforts require a backbone organization—a dedicated team that ensures coordination, keeps the shared agenda on track, and manages day-to-day operations. Just as a product manager facilitates a process to ensure the right technology gets built, the backbone organization orchestrates the collaboration between multiple stakeholders. They aren’t there to dictate strategy but to ensure that the right resources, communications, and systems are in place so all players can perform at their best.

Let's walk through each of its five components to understand how a profoundly complex problem like bus bunching could be solved using the principles of collective impact. The goal is to show how various actors—ranging from experts in machine learning to bus operators, empowered product teams, and government leaders—can collaborate to address the issues. As usual, we’ll emphasize Marty Cagan's product management principles and Jennifer Pahlka's policy engagement strategies.

1. Common Agenda: Creating a Shared Vision of Success

We will eliminate bus bunching to improve service reliability for all bus riders by working together from the policy and technology to the service delivery level.

Key Players:

• Government leaders and policymakers set high-level goals.

• Bus operations management and operators share their perspectives on the topic and align their organizations on making future changes.

• The public and advocacy agencies bring in the rider perspective, ensuring the agenda aligns with the needs of all bus riders and there is public accountability.

• Machine learning experts and product teams work to translate those operational goals into potential technical solutions they can measure and iterate on with their partners in bus operations.

From a product management perspective, the common agenda becomes the product vision. The initiative’s leaders ensure there is an emphasis on policy so the solution reflects regulatory realities and civic goals and is not hindered in cases where policy or policy interpretations block progress.

2. Shared Measurement Systems: Data-Driven Decision Making

Let’s agree on what data we will measure and how. For example, if our machine learning expert develops a model to optimize bus service adjustments, everyone will know if it’s working.

Key Players:

• Bus operations management and technology experts must agree on metrics that measure success, such as frequency of bunching incidents, average delay times, and overall system efficiency.

• Empowered product teams need real-time data to test and refine their solutions continuously.

• Government leaders and policymakers require transparent metrics to assess the broader impact and progress based on transparent information rather than creative stakeholder management.

• Bus operators and the public provide feedback on how these metrics align with the on-the-ground experience.

To really apply product management in transit tech, we must iterate based on what we see in the data and avoid assumptions whenever possible.

3. Mutually Reinforcing Activities: Cross-Functional Collaboration

Below are a few ways each potential stakeholder can help the initiative succeed.

Key Players:

• Technology learning experts develop algorithms to predict and prevent bus bunching by analyzing historical data and real-time conditions.

• Bus operations management may adjust schedules and dispatch strategies.

• Bus operators provide critical insights from the field, helping refine the models to reflect real-world conditions better.

• The product team coordinates between these groups, building user-friendly tools for operations staff.

• Program leaders and funding agencies ensure financial and regulatory support exists to make these changes possible.

• Advocacy groups ensure the proposed solutions benefit all riders, especially underserved communities.

Each stakeholder group must feel empowered to take ownership of the part they are responsible for, or the collaboration will not work.

4. Continuous Communication: Keeping Everyone Aligned

In government, we have all seen projects that were always “green” with an “on track” status until one day, they suddenly go “red” and fail as communications catch up with reality. For this initiative, we will confront issues, not hide them.

Key Players:

• The product team bridges the technical experts (machine learning, data science, software engineering, UX design, user research) and the operational experts (bus operations management and bus operators).

• Machine learning experts regularly share updates on algorithm performance and suggest adjustments.

• Bus operators and management provide continuous feedback on the implementation of these tools in real-time conditions.

• Program leaders and advocacy agencies communicate progress to the public and maintain transparency.

• The public is engaged through feedback channels, which helps refine service improvements and builds trust.

5. Backbone Support Organization: Coordinating the Whole Effort

Sometimes, a layer of “program” emerges in transit tech teams, but precisely what they do can be unclear. Do they provide solutions? No, this is what empowered product teams do. Are they just funders? No, they are deeply engaged in the work. Do they provide oversight? Sometimes. When applying collective impact, the role of program can become more well-defined as the backbone team; it manages resources, facilitates communication, and ensures progress is made across all fronts. It does this not by enforcing top-down output pressure but by course-correcting the implementation of collective impact.

Key Players:

• Bus operations policymakers ensure that the regulations support the operational changes.

• Government leaders provide the long-term vision and funding needed to sustain the effort.

Let’s evaluate the London Congestion Charge initiative. In this case, multiple stakeholders came together to address the city's worsening traffic congestion, which was causing delays, pollution, and inefficiencies in public transit.

Here’s how it reflected a collective impact framework:

Common Agenda: The goal was clear—to reduce congestion in central London and improve air quality. The initiative aimed to shift behaviors, reduce car use, and promote public transit, walking, and cycling.

Shared Measurement System: They tracked reductions in traffic volume and improvements in bus reliability, air quality, and overall efficiency of public transit services.

Mutually Reinforcing Activities: Different sectors coordinated: the transport authorities implemented the congestion charge, public transit agencies increased bus capacity, and the city promoted cycling and walking infrastructure. Advocacy groups and businesses educated the public and shaped policies.

Continuous Communication: Regular meetings were held between government agencies, advocacy groups, and the public.

Backbone Support: Transport for London (TfL) was the backbone organization, coordinating efforts with stakeholders.

The result was a dramatic reduction in traffic congestion (by around 30%), improvements in bus service reliability, and a boost in public transit use. While not without its challenges, the London Congestion Charge is widely regarded as a success and has inspired similar initiatives in cities worldwide.

When the MBTA announced its podcast, I was skeptical. But I listened. In Secretary Monica Tibbits-Nutt’s episode, she made a striking declaration: “I’m not here to keep my job; I’m here to do my job.” She was talking about being able to explain to her children that when she had the chance to make a difference, she took it—even though it would have been easier to maintain the status quo.

When buses clump together, riders experience delays, uncertainty, and frustration. This isn’t just a nuisance; it undermines the very promise of public transit. And despite all the hype around AI, it’s not going to save us from this. Like Secretary Tibbits-Nutt, those of us who have the privilege to work on these challenges must be willing to do whatever it takes to make a measurable difference—even if it means negotiating, aligning interests, resolving conflicts, and navigating the risk-averse culture of government. As Nelson Mandela once said, “It always seems impossible until it’s done.”