Nextdoor·Software Engineer·Onsite - System Design / Architecture
May 2026
Data engineering interview at Nextdoor centered entirely on one massive analytical modeling question. It was a multi-part beast covering table design, ETL strategy, retention, rollups, and timezone handling. Felt more like a take-home that got crammed into a live session.
- Design the derived table schema that powers a fast daily KPI dashboard for a large photo-sharing app. Cover grain, keys, what gets precomputed versus computed at query time, and how you'd handle partitioning, clustering, and incremental loads.
- Walk through the daily ETL process to populate these KPI tables from the raw source tables. High-level steps are fine, SQL or pseudocode is acceptable.
- How would you add week-1 retention to this dashboard? For each signup cohort date, what fraction of users are active exactly one week later? What table changes are needed and how do you compute it efficiently?
- The dashboard needs to support weekly rollups instead of daily. What changes in the table design and computation, particularly around non-additive metrics, partial weeks, and how 'change vs 7 days ago' translates to a weekly grain?
- A New York office wants the same metrics reported in Eastern time instead of Pacific. What needs to change across timestamp handling, ETL windows, backfills, and how do you support both timezones without double-counting events at day boundaries?
“This is where I spent most of my energy and still felt like I was only halfway there.” The rest of the author's notes on Software Engineer interview at Nextdoor, Onsite - System Design / Architecture round, covers how they worked through the question, what the panel pushed back on, and what they would do differently.
View Post