Problem
Spotify's Analytics Platform team builds notebooks, dashboards and metrics for people who work with data. The title "Data Scientist" now covers many different jobs, and many people who work with data do not have that title. The team used the broad term "data practitioner", but it was too vague to guide product decisions. The team first needed to understand who its users are.
Example from the article: "data practitioner" lumps together people with the same title but different needs, and people who work with data without any data title.
Solution in brief
The team built personas with a mixed-methods loop. User interviews produced the first quantitative classification rules, and further interviews confirmed or corrected those rules. The rules stayed simple (for example, "used dashboarding tools more than 90% of data scientists") so stakeholders could read and trust them. The result is six personas, and a daily dataset that assigns every internal user to one.
Questions this case answers
- How to define user segments that are interpretable enough for product decisions when the formal role label is too broad to separate needs
- How to keep a qualitative and a quantitative segmentation in agreement while the segment rules are still being built
- How to segment internal data users into groups that product teams can act on when job titles no longer say what people do?
- How to combine user interviews with usage data so that segments are both validated and measurable?
Tools and methods
- Mixed-methods research — in-depth interviews and analysis of employee usage data, run as an iterative loop
- Rule-based classification — threshold rules on usage compared with other data scientists, mapping each employee to a persona
- Daily persona dataset — every internal user classified each day, used for recruitment and for segmenting product-area metrics
Google or a one-time email code. No password.