Meta·Software Engineer·Onsite - System Design / Architecture·Senior

Senior

Apr 2026

Summary

Meta data engineer interview with a meaty system design question around video streaming metrics. One question, pretty open-ended, felt like they wanted to see how you think about observability end-to-end rather than just knowing the right answer.

Questions Asked (1)

Pick a video streaming metric and design a full monitoring and diagnostic system for it: define the time-series view with appropriate granularity and smoothing, break it down by dimensions like device, network, and geography, add diagnostic visuals like percentile bands and funnels, describe how a user would drill down through the data, and call out which guardrail metrics you'd show alongside it.

System DesignProduct Analytics & MetricsRoot Cause Analysis

Author's notes

I went with rebuffering ratio because I had something to say about it.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Choose a metric like Video Start Time (VST) or Buffering Ratio, then structure your answer around the metric definition, time-series design, dimensional breakdowns, diagnostic visuals, drill-down flow, and guardrails. Emphasize how each layer helps identify and resolve issues, and tie it back to user experience and business impact.

Pro tip: Start by clarifying the metric's definition and the user experience it represents, then consistently link every design choice to how it helps detect or diagnose real-world problems. This shows product sense and engineering pragmatism.

1. Define the metric and its importance

Clearly state the chosen metric (e.g., Video Start Time) and why it matters for user engagement and retention. Define how it's measured and its target.

2. Design the time-series view

Specify granularity (e.g., 1-minute for real-time, hourly for trends) and smoothing techniques (e.g., moving averages, EWMA) to balance responsiveness and noise reduction.

3. Break down by dimensions

Identify key dimensions like device type, network type (WiFi, 4G, 5G), geography, and content type. Explain how slicing by these reveals root causes.

4. Add diagnostic visuals

Include percentile bands (p50, p90, p99) to show distribution tails, and funnels (e.g., from video load to first frame) to pinpoint drop-offs.

5. Describe drill-down and guardrails

Outline a user flow: start from global view, drill into dimensions, then to individual sessions. List guardrail metrics (e.g., rebuffer ratio, error rate, engagement) to monitor alongside.

Key Points to Mention

Metric definition and its direct link to user experience (e.g., VST affects engagement).
Time-series granularity trade-offs: real-time (1-min) vs. trend (1-hour) and smoothing methods.
Dimensional breakdowns: device, network, geography, and how they help isolate issues.
Percentile bands (p50, p90, p99) to capture tail latency and outliers.
Funnel analysis from video request to playback start to identify drop-off points.
Guardrail metrics: rebuffer ratio, error rate, session length, and their thresholds.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.