Step-by-Step Teaching

Step 1: The Core Problem

Why does this matter?

Imagine Netflix has millions of devices — smart TVs, phones, gaming consoles, streaming sticks. Each device has different hardware:

❗ Key Insight: You cannot push a 4K feature to a device that only supports HD. Doing so breaks user experience.

Therefore: Netflix needs a system to know what each device can do.

---

Step 2: The Solution — Device Capability Modeling

Netflix builds a data model that captures each device's capabilities, including:

• Screen resolution
• Supported video profiles
• Audio capabilities
• RAM size
• CPU cores

Think of it like a detailed profile card for every device model.

---

Step 3: First Data Strategy — The Cumulative Table

Purpose: Track the latest state of each individual device

How it works:

• Stores current capabilities per device
• Gets updated as devices change (software updates, new hardware detected)
• Ideal for per-device lookups

Example data structure:

{
  "Screen Height": ["720"],
  "Screen Width": ["1280"],
  "Video Profiles": [
    "playready",   ← HD capability
    "hevc"         ← UHD/4K capability
  ]
}

How to read this:

This specific device has a 720p screen (1280×720) and supports both HD and UHD video profiles

---

Step 4: Second Data Strategy — The Histogram Table

Purpose: Understand capability distribution across many devices

How it works:

• Looks at active devices over the past 28 days
• Groups devices by model and software version
• Counts how many devices support each capability

Example:

{
  "Video Profiles": {
    "playready": "100%",   ← ALL devices support HD
    "hevc": "20%"          ← Only 1 in 5 devices support UHD
  }
}

Real-world interpretation:

Out of all active streaming sticks, every device can stream HD content, but only 20% can stream in 4K Ultra HD

---

Step 5: How These Two Tables Work Together

Cumulative Table          Histogram Table
─────────────────         ─────────────────
"What can THIS            "What % of ALL
 device do?"               devices can do X?"
       ↓                          ↓
  Per-device               Aggregate-level
  detail                   distribution
       └──────────┬───────────────┘
                  ↓
         Analytics Products
         (Feature decisions)

---

Step 6: Real-World Application — Feature Decisions

Using this data, Netflix can answer questions like:

Example Decision:

If only 20% of devices support HEVC (UHD), Netflix knows that rolling out a 4K-only feature would exclude 80% of users — a data-driven reason to maintain HD fallback support

---

Summary

💡 Core Principle: Data about what devices can do enables Netflix to decide what features to build and deploy — maximizing both innovation speed and user experience quality.