Context
LinkedIn had a wealth of workforce data — the kind talent leaders at Fortune 500 companies would pay dearly for. The company had been producing custom insight reports through a small analyst team, helping clients answer questions like "Where are the AI researchers located across the country?" or "How competitive is the engineering talent market in Austin?"
But clients kept asking for more. They wanted a self-service platform — and it made business sense, since a small analyst team couldn't scale. That became LinkedIn Talent Insights. The problem: the team didn't know who the day-to-day users would be, what they needed, or what the product should look like.
How I Worked With the Team
I was the sole researcher, but this was never a solo effort. PMs, designers, and product leaders were involved at every step — observing interviews, participating in synthesis workshops, and co-creating artifacts like the journey map and proto-personas. I ran debriefs after every research round so the final readout was never a surprise. By the time I presented, the team had already been building the insights alongside me for months.
Research
Over 4–6 months, I ran three to four rounds of research. Each round consisted of 8–10 in-depth interviews (90 minutes each) with talent executives, analysts, data scientists, recruiters, and sourcers at large enterprises. I combined foundational and evaluative approaches — some sessions focused on understanding their workflows and pain points, others evaluated early design concepts and prototypes.
Key outputs included proto-personas mapping the different roles involved in talent strategy, a journey map tracing the full talent forecasting workflow, and a co-creative workshop where heads of talent from client companies validated our personas and maps against their own organizations.
The Turning Point
Partway through, the team got stuck in a black hole of feature debates: Should we build dashboards? A presentation builder? Notification systems? Each PM had a different priority and they were going in circles.
Instead of scoping individual studies for each question, I decided to go deeper. I spent several weeks re-reading every set of notes and re-watching interview videos from all our prior studies, asking a single question:
"What are participants organically asking us — unprompted, without our prodding?"
Two powerful themes emerged. First, participants across every role and company were intensely focused on data — not collaboration features, not dashboards. They wanted to know: Where does your data come from? Can I trust it? How was this metric calculated?
Second, this anxiety wasn't abstract. It was driven by their need to share data with highly analytical hiring managers and executives. If a recruiter couldn't answer basic questions about the data's provenance, it would undermine their credibility in front of the people they needed to influence.
Core problem: Talent professionals are worried that their stakeholders will misinterpret or undermine their data, creating misalignment or hurting their credibility.
The T.R.A.C. Framework
I translated this core problem into a "true north" for the team: How might we empower talent professionals with robust, relevant data that earns them trust and credibility with their hiring managers?
Then I broke that into four actionable product principles:
Transparency
Show users exactly how every data point is calculated. No black boxes.
Relevance
Help users narrow down to the specific data that answers their question.
Accuracy
If the data isn't trustworthy, nothing else matters. This is table stakes.
Control
Let users curate, filter, and export data on their own terms.
Impact
LinkedIn Talent Insights became the fastest product to hit $50M in revenue in LinkedIn's history.
The team built detailed explanations of data formulas directly into the product interface (Transparency).
3–4 additional data scientists were hired specifically to clean and validate data sets (Accuracy).
More robust search filters and data export capabilities were added (Relevance + Control).
The T.R.A.C. framework became the team's decision-making lens for every subsequent feature debate.
LinkedIn
ServiceNow
Netflix
