NPR Digital Services Didn’t Fail. The System Wasn’t Designed for it to Succeed.

A note before we begin: I joined NPR Digital Services in January 2012 and was promoted to Managing Director in January 2016. I was the MD who closed it down. I have a direct stake in how this story gets told—and a responsibility to tell it honestly. What follows is my attempt to do that.

NPR Digital Services did not fail because the people running it weren’t capable.

It failed because the system it was trying to serve didn’t trust it.

And public media is about to try again. This time, with almost $50 million, a new independent nonprofit called Public Media Infrastructure, and a coalition of five founding organizations. If the lesson from the last two decades stops at governance structure and funding model, PMI will run into the same wall. The governance improvements are necessary. They are not sufficient.

The real work is trust. And trust, in a system this fractured and this decentralized, doesn’t repair itself.

What NPR Digital Services Actually Was

In 2008, NPR acquired Public Interactive, a Boston-based organization that had been helping public radio stations build out their digital capabilities since 1999. It was renamed NPR Digital Services and kept as a separate division, headquartered in Boston, focused on serving the broader ecosystem. The intent was sound. The premise—that stations needed shared digital infrastructure they couldn’t build alone—was correct then and is still correct now.

NPR Digital Services did real work across both tools and systems: analytics services, digital content tools and training programs, membership and marketing technology, and infrastructure work, including the Public Media Platform, the NPR Story API, and an architectural shift away from monolithic code toward microservices. Stations used those capabilities to produce work that earned awards and national recognition.

But by 2017, NPR had decided to shutter the Boston operation and absorb Digital Services into its own technology organization. Part of the rationale was resource efficiency. Part of it was strategic—NPR was working through what it meant to redefine the relationship between its member stations and the national organization, trying to operationalize a local-national network model that had been the subject of years of scenario planning and advisory group work. Merging the two divisions was, in theory, a step toward alignment.

In practice, it put a single team in charge of two fundamentally different missions—NPR’s internal digital ambitions and the infrastructure needs of hundreds of independent stations. Those missions were never going to coexist cleanly inside the same org chart.

By 2019, the mandatory station fees that had funded the division were eliminated. NPR Digital Services, as a distinct entity, was over.

The conventional diagnosis points to three structural failures: no independent governance, competing financial interests as NPR expanded into podcasting and digital advertising, and mandatory fees that removed any accountability for quality. All three are accurate.

But they’re symptoms. The root cause was a trust deficit that started accumulating long before the division was created—and that no governance document alone was going to fix.

An Experiment, Not a Failure

Before going further, I want to reframe something.

NPR Digital Services is almost universally described inside public media as a failure. I understand why. But I don’t think that framing serves the system well—and I think it’s getting in the way of the honest reckoning that PMI actually needs.

DS was an experiment. A serious, well-resourced, genuinely motivated experiment in whether a distributed, independence-first ecosystem could agree to share digital infrastructure. It produced real value. It also revealed, with considerable precision, exactly where the system’s fault lines ran.

Experiments that surface fault lines aren’t failures. They’re data.

It also matters when DS was running. The early 2010s were not the landscape public media is navigating today. The financial cliff wasn’t yet visible. Stations had more runway, more tolerance for friction, more capacity to resist change they weren’t sure they wanted. The urgency that now exists—federal funding winding down, newsroom layoffs, audience fragmentation accelerating, organizations like CNN and CBS News visibly struggling under the same pressures—that urgency wasn’t present in the same way. The system could afford to reject the experiment. It cannot afford to reject this one.

What would it look like if we approached PMI not as an attempt to avoid the mistakes of DS, but as the next iteration of a long experiment in figuring out how public media survives disruption? That framing changes what we owe the history. Instead of cataloging DS as a cautionary tale, we treat it as a prior round of learning. The question becomes: what did we actually learn, and are we building toward those lessons or around them?

This Vision Has Been Tried Before

The core vision behind PMI is not new.

In the early 2010s, Kinsey Wilson—then NPR’s Chief Content Officer—articulated a vision for a reimagined public radio ecosystem built on shared digital infrastructure. The centerpiece was NPR One, a non-linear, human-curated mobile listening app that blended local and national content in a single experience. The promise was that stations and NPR could build innovative products together on top of shared APIs, pooling technical investment instead of duplicating it across hundreds of independent organizations.

It was the right vision. It was also early—and it was built in a context that made adoption nearly impossible.

The infrastructure layer beneath NPR One—known as the Public Media Platform—was chronically underfunded and lacked the governance and incentive alignment mechanisms that would have given stations a reason to build toward it rather than around it. Stations saw it not as shared infrastructure they co-owned, but as an NPR product they were being asked to feed. NPR Digital was widely known inside the organization for a culture that bordered on insular devotion to the vision, which had the unintended effect of souring collaboration with other divisions and deepening station skepticism about whose interests the platform actually served. In 2023, NPR One's features were blended into the NPR News app as a refresh of its NPR mobile app experience.

NPR Digital Services, operating as a separate division, often ended up in the middle. When any part of NPR—News, Programming, Digital—created friction with stations on anything digital-adjacent, DS got pulled in, whether it was involved or not. The fee was the tangible thing stations could point to. So DS became the surface where diffuse institutional frustration landed, absorbing a reputation it was never supposed to carry.

NPR Digital Services inherited a trust deficit it didn’t create. That distinction matters—both for how we assess what happened and for what PMI should expect as it steps into the same ecosystem.

There is a direct line from Kinsey’s vision to what PMI is attempting today. The difference—and it is a significant one—is that NPR is no longer at the center. That changes the politics substantially. Whether it changes the outcome depends on whether the system has genuinely learned from what happened before, or simply reorganized around the same unresolved tensions.

How Trust Erodes in a Decentralized System

Public media is often described as “married, but living separately.” That description captures the structure, but not the dynamic. In most functional marriages, even complicated ones, there’s a shared agreement about how decisions get made. In public media, that agreement has always been contested.

Local stations are independent 501(c)(3) nonprofits. They are accountable to their communities, their donors, and their boards—not to national organizations. That independence is legitimate. It is also the source of persistent friction in any effort to build shared infrastructure.

What eroded trust in the Digital Services era wasn’t a single decision. It was an accumulation of signals.

The fee structure created a particular problem at the larger end of the ecosystem. Major market stations paying over $100,000 annually often felt the tools and services didn’t deliver value proportional to what they were investing. DS capabilities tended to serve mid-sized and smaller stations well—those were the organizations that genuinely benefited from shared infrastructure they couldn’t have built independently. But large stations had the resources, the leverage, and the loudest voices. Their dissatisfaction shaped the narrative around DS in ways that didn’t always reflect the experience of the majority of the system.

Transparency about NPR’s own financial pressures was inconsistent. Stations found out about restructuring decisions—like the 2017 layoffs of station relations managers in Boston—through news coverage, not through direct communication. When you learn about decisions that affect your operations from a trade publication rather than your partner, you update your point of view of how much you’re actually seen.

And when NPR aggressively moved into podcasting and digital advertising—competing directly with the stations it was simultaneously positioning itself as serving. The mission and the business model diverged. Stations noticed. That divergence, more than any structural issue, was another moment the trust account went negative.

NPR bears real responsibility for why the trust inside this system is where it is. That responsibility should be named, not softened. What makes it complicated is that it didn’t all originate with NPR Digital Services. DS inherited a system that was already fractured. It was asked to build bridges across fault lines it didn’t create—and when it fell short, it absorbed blame that belonged to the broader institutional history.

The Data Question Nobody Wanted to Answer

Underneath the governance failures and the fee resentment was a more fundamental conflict that the system never resolved honestly: who owns the relationship with the listener?

Stations framed this as an ownership question. My audience. My donor base. My market. That framing was understandable—and flawed.

A listener who signs up for a station’s newsletter, downloads an app, or makes a membership donation isn’t transferring ownership of themselves. They are granting permission to be communicated with. That permission belongs to the person who gave it. It can be revoked. It travels with the listener, not with the institution.

When national organizations began building digital platforms that created direct listener relationships—the NPR app, NPR One, national email lists—stations experienced it as a threat to something they owned. But the harder question, the one that went largely unasked, is whether stations were actually managing those listener relationships in a way that justified the ownership claim.

Were stations using data to deepen local relevance? Were they investing in the tools and practices that would have made them irreplaceable to their audiences, regardless of what national platforms were doing? Or were some stations holding so firmly to a belief in ownership that they were unwilling to participate in any data governance framework that acknowledged the listener had a say?

That rigidity had consequences. A shared data infrastructure—standardized analytics, unified audience measurement, coordinated permission frameworks—would have given every station better intelligence about their audiences than they had operating independently. It also would have required agreeing on how that data was used, who could access it, and what the listener’s rights were within the system.

That conversation never happened at the scale it needed to. Not because it was technically impossible. Because it was politically uncomfortable.

PMI is building shared audience analytics and data infrastructure as a core part of its roadmap. The data governance question is coming back around. This time, the system will need to answer it—including the part about what listeners actually consented to, and who is accountable for honoring that.

The Enterprise Problem Nobody Named

There is one more tension from the DS era that deserves to be said plainly, because it shaped the collaboration failures as much as anything else.

Building technology for a single station and building enterprise technology for hundreds of stations are not the same problem. They require different skill sets, different architectural thinking, different assumptions about extensibility, scalability, and long-term maintenance. Enterprise infrastructure has to be designed for growth, for flexibility, for the use cases that don’t exist yet. Station technology tends to be designed for the immediate need in front of the team building it.

Neither is wrong. They’re just different.

What created friction is that stations would come to Digital Services with a specific request—a feature, a tool, a capability they wanted. And the response from the product and engineering teams wasn’t a flat refusal. It was a set of questions: What problem are you trying to solve? Is this a content problem or a revenue problem? What does success look like? How would you measure impact?

Those are good product questions. They are the questions that separate building the right thing from building the wrong thing very expensively.

Stations often interpreted them as an interrogation. As if their judgment was being questioned, their strategy second-guessed. What was actually happening was that a service division was trying to understand the business case well enough to build something durable—something that would work not just for one station in one moment, but for the ecosystem over time.

The trust deficit made that conversation nearly impossible. When you don’t trust the institution asking the questions, the questions themselves feel like obstruction. And so the collaboration that would have produced better outcomes never fully happened—not because either side was acting in bad faith, but because the relationship wasn’t strong enough to sustain the necessary friction of good product development.

PMI will face the same dynamic. The work of building shared infrastructure is inherently a negotiation between what individual stations want and what the system needs. That negotiation requires enough trust to have the hard conversation—the one where a station says “we need this,” and the infrastructure team says “tell us more,” and both parties stay in the room long enough to find the actual answer.

Stations Bear Responsibility Too

Here is what I want to say plainly, because it rarely gets said: the national organizations were not the only source of friction.

Stations bear responsibility too.

A decentralized system in which every node prioritizes its own independence above shared goals is not a system. It’s a collection of separate organizations that happen to share a satellite feed. Some stations resisted adoption not because the tools were wrong, but because adoption itself felt like a concession to centralized authority. Some withheld feedback that would have improved shared products because engaging felt like endorsing a direction they didn’t control. Independence, when it hardens into reflexive resistance, becomes its own kind of bottleneck.

The system rejected the work. Not because of what national organizations did, but because of what everyone in the system chose not to do.

What PMI Is Building Today

PMI doesn’t have to become another NPR Digital Services. The structural differences matter and should not be minimized.

PMI is an independent 501(c)(3), not a division inside a national organization. It was founded in November 2025 by a consortium of five organizations—APMG, PRX, NYPR, SRG, and NFCB—and is funded by a CPB grant to sustain public radio distribution through 2030. Its inaugural Board of Trustees includes station leaders from small, rural, Tribal, and bilingual stations alongside the founding organizations. Participation is opt-in. There are no mandatory fees. Core services are available to stations at no cost.

These are meaningful improvements, not cosmetic ones. The design of PMI reflects the lessons of what came before it.

But design is not destiny. PMI is launching into a system that is carrying years of accumulated distrust, a fractured relationship between NPR and CPB that only recently settled, the loss of federal funding, and stations that are managing through layoffs and strategic pivots with fewer resources than they had five years ago.

The question is not whether the structure is right. The question is whether the system can actually use it.

PMI also has something DS never had: it starts with some initial trust already in place, because it was founded by station groups. That is a real advantage. It doesn’t eliminate the work of building trust with small and mid-sized stations—but it changes the starting position. The goal isn’t to earn trust from zero. It’s to make sure the early decisions reinforce it rather than quietly erode it.

The Question PMI Has to Answer Before It Builds Anything Else

What does PMI look like in five years, when the grant expires?

That question is not premature. It is the most important design constraint PMI is operating under right now. Every foundational decision—what to build first, which services to centralize, how to structure governance, how to price participation—has a different answer depending on what PMI is trying to become.

A sustainable technology utility serving the public media ecosystem looks different from a convener of shared investment. A station-owned cooperative looks different from an independent nonprofit with a narrow infrastructure mandate. The board members who founded PMI almost certainly have different intuitions about this, and those differences will surface as resource allocation decisions become real.

The Kinsey Wilson era offers an instructive cautionary note. The Public Media Platform needed substantial capital to build the infrastructure layer that would have made NPR One’s vision achievable at scale. It received significantly less. The gap between the ambition and the capitalization was never honestly reconciled—and the initiative collapsed under the weight of that gap, not from a failure of vision.

PMI has more runway. But runway without a shared destination is not enough. The board needs to reach alignment on what PMI is optimizing for at the five-year mark—and then work backward from that answer to make sure every foundational decision is building toward it, not just toward the end of the grant period.

There is also a first principles question worth sitting with: if you could redesign public media from the ground up today—knowing what you know about technology, about audience behavior, about the economics of media at scale—what would you build? Not what would you migrate to. What would you actually design? That question won’t produce a single answer. But asking it honestly will surface what the system actually values versus what it has simply inherited.

The exit is not the end. It’s the design brief. And how PMI allocates resources in its first two years will reveal whether intentionally or not, what answer the organization has actually chosen.

What Building Trust Requires—From Everyone

Trust in a system this complex is not a feeling. It’s a set of behaviors that compound over time. It has to be designed, not hoped for. And it requires something from every part of the ecosystem.

What NPR and PBS need to accept and do:

The most important thing NPR can do for PMI is resist the impulse to compete with it. That means being explicit—publicly and internally—about where NPR’s role ends and PMI’s begins. The CPB-NPR settlement established that NPR continues to operate PRSS, with no interconnection fees for stations, while PMI focuses on modern digital infrastructure. That clarity is a foundation. It has to hold under pressure, including the financial pressure that will come as the grant runs out.

NPR and PBS also need to accept that their credibility as partners is not assumed. It has to be rebuilt through consistent, transparent behavior—sharing information proactively rather than reactively, not making decisions that affect stations without early and genuine input, and acknowledging, without defensiveness, the ways the prior era of shared infrastructure created real harm to station operations and trust.

Acknowledgment is not weakness. In a system built on public trust, modeling the behavior you want the ecosystem to adopt is leadership.

What station leaders need to do:

Stations have to decide whether they want PMI to succeed. That sounds obvious. It isn’t. The default posture in a decentralized system under financial stress is self-protection. Stations are managing their own crises. Investing time and attention in a shared infrastructure initiative that won’t pay off for 18 to 24 months requires a level of institutional generosity that is genuinely hard to sustain when you’re cutting staff.

But self-protection, at scale, is what fragmentation looks like. If every station waits for someone else to adopt first, no one adopts.

Stations need to show up in the governance structures PMI is building—not just as recipients of decisions, but as active contributors to them. The board representation from small, rural, and Tribal stations is a structural achievement. It only matters if those representatives are empowered to push back, ask hard questions, and hold the founding organizations accountable to the stations’ actual needs rather than their aspirational ones.

Station leaders also need to share what isn’t working. The temptation in a new initiative with significant funding behind it is to be diplomatic. Diplomatic feedback produces better-looking products that still miss the mark. If a tool doesn’t work for a rural station with two engineers and a $2 million budget, that needs to be said directly and early—not six months after adoption when the problems become visible anyway.

And stations need to revisit the data question with fresh eyes. If the listener granted permission—not ownership—then the conversation about shared data governance isn’t a threat to the station relationship. It’s an opportunity to serve that listener better than any single station can do alone.

What PMI itself has to do:

PMI is currently building the systems and governance that will determine whether this initiative holds together. A few things will define that more than any technical roadmap.

Value has to be visible early, and it has to be visible to the smallest stations first. The credibility trap in shared infrastructure initiatives is that the early wins tend to benefit the organizations with the most resources to adopt quickly. If the first measurable results flow to NYPR and not to KCAW Raven Radio in Sitka, Alaska, the narrative about who PMI is actually for will calcify before the pilot cohorts are complete.

Governance mechanisms have to function under stress, not just in optimal conditions. The five founding organizations will eventually face a decision where their individual interests diverge from the system’s interests. Binding dispute resolution and decision rights that don’t require consensus for every operational choice are not bureaucratic overhead. They are the architecture that keeps the organization functioning when alignment gets hard.

And PMI needs to communicate differently than its predecessors did. Not press releases and annual reports. Ongoing, honest reporting to stations about what is being built, what is behind schedule, what costs more than expected, and what the organization is learning in real time. Stations burned by a lack of transparency in the Digital Services era are not going to extend trust based on good intentions. They are going to extend trust based on evidence that this organization behaves differently when things are hard.

That includes being transparent about data governance from the start—how listener data is collected, who can access it, what stations control, and what the listener’s rights are within the shared infrastructure. That conversation will be uncomfortable. It is also overdue.

The Structural Logic That Has to Hold

There is a design principle underneath all of this that, if PMI gets it right, makes the trust work possible.

Standardize the backbone. Differentiate at the experience layer.

The underlying infrastructure—content delivery, data standards, distribution, cybersecurity, compliance, emergency alerting, monetization technology—should be centralized and shared. Anything with significant governance, compliance, or security requirements benefits from centralization, not distribution. Every station rebuilding the same foundation independently is an enormous waste of resources at precisely the moment when resources are most scarce. Shared infrastructure, done right, gives every station more capacity to serve its community. It also creates room for PMI to negotiate enterprise-scale agreements with technology providers—so that stations aren’t each going to market independently for tools they all need.

But the experience layer—how a station serves its audience, expresses its editorial identity, builds local loyalty—has to remain local and flexible. The failure mode in the Digital Services era was not just governance. It was also a tendency to standardize the wrong layer. Tools designed around the needs of large metro stations with deep technical teams don’t work for community stations. One size, in this ecosystem, never fits all.

PMI’s modular, opt-in approach might be the right structural answer to this problem. The real test will be whether the founding organizations can agree on which problems PMI is actually solving—and therefore which services get resourced and prioritized first. That agreement has to come from genuine stakeholder alignment, not from the founding coalition building what’s easiest to build and hoping stations find it useful. Get the right people in the room. Talk through the trade-offs. Define the experiments within the priorities. Then build.

On AI: Don’t Ignore It. Don’t Rush It.

Any honest conversation about building digital infrastructure today has to include AI. PMI should not treat it as a future consideration. It should be a question embedded in every initiative from the start: how might AI help here? How might AI transform how this system thinks about digital infrastructure altogether?

But that question requires a foundation before it can be answered well. AI’s potential in this ecosystem runs across three layers—and conflating them creates the same kind of misaligned expectations that have plagued previous infrastructure efforts.

The foundational layer is data. Before PMI can meaningfully leverage AI, it needs to understand what data exists across the system, where it lives, how it’s structured, and whether it’s clean enough to be useful. Most stations are operating with fragmented audience data, inconsistent measurement frameworks, and no shared taxonomy. You cannot build intelligence on top of that without first building the infrastructure that makes the data coherent. That assessment has to happen before the AI conversation gets specific.

The efficiency layer is where the near-term opportunity sits. Once a shared data foundation is in place, AI can reduce real friction in operations that currently consume enormous institutional bandwidth: stakeholder communications across a diverse and geographically dispersed ecosystem, governance documentation and decision tracking, dynamic resource allocation modeling, cross-functional coordination between founding organizations and station cohorts. These are not transformational use cases. They are the operational overhead that currently slows everything down—and AI can compress those timelines significantly once the data infrastructure supports it.

The transformational layer is about reimagining how the system thinks about its own mission. What does content distribution look like when AI can optimize delivery in real time based on audience behavior? What does audience development look like when a small station in rural Alaska has access to the same analytical intelligence as a major metro outlet? What new revenue models become possible when the monetization infrastructure is shared and the data layer is unified?

PMI should be building toward that layer. But the path there runs through the data foundation first. Skipping ahead to transformation before the backbone is in place is how you end up with impressive prototypes that the system can’t actually adopt.

The right question at every stage of PMI’s roadmap is not “what AI tools should we adopt?” It’s “what would have to be true about our data and infrastructure for AI to genuinely transform how this system operates?” Build toward that answer. Everything else follows.

The Honest Question

The question that will determine PMI’s trajectory is one that no governance document can answer:

Can the system trust that this is being built with it—not for it?

For NPR and PBS, the answer requires a demonstrated willingness to cede authority over the digital infrastructure narrative and genuinely accept a complementary rather than primary role.

For station leaders, the answer requires a willingness to engage actively, share feedback honestly, and extend some degree of good faith to an initiative that the historical record gives them reasons to doubt.

For PMI, the answer requires building the kind of operational transparency and early credibility that makes that good faith reasonable rather than naive.

None of this is guaranteed. The history is real, the resource constraints are real, and the timeline is compressed. Federal funding winds down, and PMI needs to demonstrate enough value to sustain itself before that clock runs out.

But the original problem—that local stations need shared digital infrastructure they cannot build alone, and that building it requires a level of system-wide collaboration that doesn’t come naturally to an ecosystem designed around independence—that problem has not changed since 2008.

PMI might be the best-designed attempt to solve it that public media has produced. Whether that’s enough depends on whether the entire system—not just the organizations running it—decides it wants this to work.

The vision has been right all along. What changes now is whether the system is finally ready to act like it believes that.

—

Stephanie Kord Miller is a business operations strategist and executive coach with over 20 years of experience building systems inside complex organizations. She joined NPR Digital Services in January 2012 as Director of Station Relations and Communications and served as Managing Director from 2016 until its closure in 2017. She is the founder of Empower Your Results.