NPR Network Infrastructure
B2b PLATFORM DESIGN • DEVELOPER EXPERIENCE • CONTENT STRATEGYBackgroundTeams across NPR develop tools for member stations that ultimately power audience experiences.
But these tools were all built in isolation from each other, over a long period of time. Each one had its own data model with no shared standards or infrastructure.
As a result, station tools were frustrating to use and to maintain, making developing audience products unnecessarily difficult.
I led the design strategy that guided our approach to modernizing the foundation that powers NPR’s core digital services.
The teamVeronica Erb, research lead
Lauren Bracey Scheidt, product lead
Dina Nawas, product manager
Noah Carnahan, engineering lead
If a station staffer needed to get a new show on their air, it meant signing into and using 5 separate tools that did not connect to each other.
The challengeWe needed to unify NPR’s technical infrastructure, and a longstanding headache for station staff provided us with a clear starting point.
Station staff used NPR-managed tools daily, but the lack of common infrastructure meant they had to sign into each one separately. Every day.
No thanks.
Thankfully, NPR had recently committed to making full use of Salesforce, which has single sign-on functionality out of the box.
This meant we could start by solving a concrete problem for station staff in service of establishing the first piece of shared infrastructure that teams across NPR could easily implement.
DiscoveryThe team interviewed pairs of product managers and engineers for each station tool to understand how we could design an easy-to-implement and standardized sign-in process.
Our findings revealed the complexity and severity of problems caused by the lack of shared infrastructure.
Silos had been preventing knowledge sharing and collaboration. Everyone was frustrated and wanted more cohesion and connection, but no one had spare time, and oftentimes, it wasn’t even clear who they should connect with in the first case.
This is when we realized that we were dealing with a much more complex problem.
Sure, these systems needed to talk to each other, but we needed to figure out how to let the people who run them do the same.
[The station analytics tool] is like lots of different, over-engineered processes that are not easily maintainable and not very well documented. It’s Frankensteined and very brittle. It exists largely in isolation from other digital tools.
— NPR software engineer
building shared understandingWe weren’t interested in preserving silos, so we invited our research participants to hang with us.
Nothing more to it — just some snacks in the kitchen on the 4th floor of NPR HQ and a chance to get to know each other as humans before we started working more closely together.
After that, we invited our fledgling coalition to a workshop to create the first comprehensive map of NPR’s existing infrastructure.
This map illustrated how convoluted our systems were, which:
Visually conveyed the complexity of the problem to our stakeholders
Made it easier for engineers in different teams across NPR to help identify opportunities for improvement
ExperimentationOur mapping exercise helped us select a low-risk product to use as our guinea pig.
Engineers from our team and theirs worked closely together to build on top of Salesforce’s SSO product.
We all created technical documentation together, resulting in two separate teams with deep knowledge on how the new SSO product works, and how to best implement it.
Knowledge: shared. H*ck yeah.
This process also helped me gather requirements for a one-time password migration flow, which I worked on with my research colleague, Veronica.
We shared our progress with the full team often, and eventually tested it with station staff using paper prototypes.
Tackling the intractableWe still needed a unified data model in order to make station tools easier for NPR staff to maintain and station staff to use.
We’d outgrown our old data models, which had grown organically over time. They were disparate, unconnected, and broadcast-centric.
Veronica and I sprinted ahead along with our product colleague, Lauren, to deepen our understanding of the problem and figure out out how to structure all of this data.
We established a shared language based on our research. This got all of our collaborators on the same page, while also creating an easy-to-understand naming convention for the Salesforce solution design.
We also created a robust problem statement alongside our many collaborators, which you can read here.
SketchingWe were painfully aware of how easy it is to add things to a data model, and how hard it is to take things away.
I relied heavily on concept mapping to help me understand the relationships between organizations, the people who work there, the broadcast stations they operate, the services they produce, their collaborators… you get the idea.
We needed to get this right from the get-go. So as I began sketching the structure, I shared my work early and often with everyone involved and held quick, ad-hoc sketching sessions.
collaborative Stress testingI convened our coalition to stress test the model collaboratively to see how it could handle use cases that the current system struggled with.
Turns out, our research and collaborative design process was worth it!
We had an incredibly durable model that could accurately represent even the most complex relationships, prioritize digital use cases, and preserve information important to the broadcast side of the business.
Because each of the 250+ member stations in the network are independently owned, the data model had to accommodate a number of use cases. New York Public Media is an example of one that's pretty straightforward.
Arizona Public Media is a little bit more complex, but the model can still handle it. Here, you can see that the Arizona Board of Regents holds the member station's FCC license.
The model can also accommodate consortia — large-scale, long-term collaborations between many member stations.
One of my favorite examples! This shows one of the more complex relationships, where one overarching organization owns two member stations who operate varying numbers of broadcast stations. Before the data model existed, each broadcast station was treated as its own organization. Yikes!!
IMPACTThe implementation of the data model made a number of significant workflow improvements possible.
The new data model makes full use of Salesforce in order to more logically organize data about organizations, the services they produce, the brands associated with them, and any stations they may operate.
It has allowed NPR to:
Introduce audiences to regional collaborative newsrooms, which was once seen as a significant engineering burden
Centralize communication workflows for NPR staff within Salesforce, making that work more transparent, reducing duplicative efforts and conflicting information
Let station staff provide important information about their job, lessening the burden on NPR staff who were maintaining that data manually and improving our data hygiene
Target communication based on that information so that station staff can get the most relevant information at the right time
Create a modernized, more efficient content delivery API, making it easier to receive content from across the network and deliver it to our shared audiences
It also gave us our foundation for NPR Studio, a centralized workspace for member station staff where all of this work comes together.
You can read about that in the NPR Studio case study!
An example of an organization record in Salesforce containing reliable, complete data managed by the people closest to it: station staff!