Identifying superfans in unexpected places at Science Friday

Audience listening isn’t new for Science Friday. It’s a 35-year-old science news public radio show and podcast designed to engage audiences, but its leaders suspected they could make more decisions based on audience needs.

Jaci Hirschfeld, Science Friday’s director of audience strategy, used the Audience Data Commons (ADC) schema to analyze two forms of audience feedback and develop a real understanding of why someone would call or email the show. The schema helped Science Friday move toward a systematic understanding of audience interests, participation patterns, and engagement. 

Data in 

Hirschfeld said the schema, which she used in an NPAI Co-Lab pilot, opened her eyes to the potential of how insights from first-party data can power Science Friday’s technical, fundraising and editorial strategies. 

Science Friday used the ADC schema to analyze open-ended voicemails and emails from two inboxes. The show recently reimagined their live call-in show format to answer audience questions async. They now publish daily podcast episodes that regularly include audience responses to specific prompts, and the podcast rolls up into a two-hour broadcast show that airs on Fridays. The show also fields story ideas and audience questions in an email inbox. 

In the pilot, Hirschfeld entered 200 voicemail transcripts and 150 email messages into the schema. She had to download and process the files first — an iterative and onerous process made easier with Zapier, Google Apps Script automations, code-writing guidance from AI tools and support from the Newsroom Robots team that helped newsrooms in the pilot. 

She came out of the process with two organized spreadsheets — one for the email inbox and one for voicemails — and a set of Zapier automations that feed in new audience feedback. 

How the schema worked

The ADC schema is a standardized framework intended to connect different sources of first-party data to reach a more comprehensive understanding of who their audiences are. It also helps newsrooms translate qualitative audience feedback into measurable patterns or trends, especially across different data types.

Science Friday used the schema to categorize open-ended messages by topic and interest, which helped them identify subjects and themes that generate the most engagement from their audience. (It’s technology, animals, climate, environment, and health.) 

They also organized the voicemail transcripts and emails by audience intent or interaction type, differentiating questions, personal stories, feedback, story ideas, and responses to engagement prompts.

Hirschfeld said she was surprised to learn that some highly engaged listeners send several emails per day, usually in response to the two-hour broadcast on Fridays, a level of loyalty that “would have gotten lost in the shuffle.” Sentiment and tone analysis also helped her understand whether audience responses were generally positive, neutral, or critical.

Insight out

A clear understanding of topic and interest categories was the most useful insight from the schema. Hirschfeld passed the quantitative findings along to Science Friday’s editorial team, helping them understand subjects that resonate with listeners instead of relying on anecdotal observations.

For example, the inboxes filled with complaints when Science Friday did a show about the songs appliances play but didn’t identify the composer of a sampled washing machine tune. 

“We received a surprising number of calls and emails from people who were genuinely frustrated that we'd introduced this fascinating example and then didn't say it was Schubert's Trout Quintet,” Hirschfeld said. The team produced another episode specifically about that song. 

“Audience feedback is increasingly becoming a pipeline back into our editorial process rather than something that simply lives in an inbox or voicemail box,” she said. 

The tool ultimately helped a newsroom that is actively exploring where AI fits into workflows to see the value of automated processes, with careful privacy guardrails in place. When Science Friday’s board asked how the organization is approaching AI at a recent meeting, the pilot gave Hirschfeld specific, tangible examples to share. 

Data from the pilot will make its way into grant reports as impact statements. Hirschfeld said it’s difficult to quantify the impact of audio programs, even though they can often trigger an emotional response from audiences. Testimonials from deeply engaged listeners are one way to show the value that Science Friday provides to its most loyal audiences. 

The pilot helped Science Friday understand both the potential and limitations when connecting two disparate data sources. They had no way to link listener phone numbers to email addresses, or create unified audience profiles across platforms.  

Despite this challenge, Hirschfeld kept the system running, and since the pilot she has analyzed more than 600 voicemails, a scope that gives Science Friday a clear picture of call frequency, sentiment and topics that callers are interested in.

Eventually, Hirschfeld hopes that standardized data from qualitative audience interactions can be integrated into Science Friday’s Customer Relationship Management (CRM) software, CharityEngine. This would allow them to link calls and emails to existing newsletter audience and donor records and build a more holistic picture of engagement over time, or even a refined audience funnel. 

Detailed audience sentiment data can help support the new membership program that the show is launching soon. “We’re making big pivots and want to grow and sustain,” Hirschfeld said.

Where it got hard 

Hirschfeld said at first she was nervous about the technical nature of the pilot and the tool itself, which is accessed via a command line interface. But the process helped her “feel empowered as someone without a technical background.” 

The NPAI Co-Lab team is exploring developing a more accessible interface for the tool, making it accessible to newsrooms without technically adept data analysts on staff. 

The process also had more roadblocks and took longer than she expected, especially during the onerous data-cleaning steps. But now, the process is automated and new messages flow into the spreadsheets that Hirschfeld created. She said newsrooms will need time, patience and technical support to overcome roadblocks. “This experience has made qualitative audience data feel approachable rather than intimidating,” Hirschfeld said.

“The questions the pilot asked us to consider fundamentally changed how we think about our audience data and what is possible with it,” Hirschfeld said. “The process pushed us to identify overlooked data sources, think critically about structure and workflows, and imagine entirely new ways of understanding audience engagement.”