CONCEPT Cited by 2 sources
Rebuffering rate¶
Definition¶
Rebuffering rate is the frequency (interruptions per streaming session, per viewing hour, or per 100 view-hours) at which a video-streaming client runs its playback buffer empty and has to pause playback to refill. It is one of the two load-bearing QoE (Quality of Experience) metrics for streaming services — the other being visual quality — because a viewer who sees a spinner cares less about VMAF than about getting back to the movie.
Why rebuffering happens¶
Adaptive-bitrate streaming (see concepts/adaptive-bitrate-streaming-dash) tracks available throughput over a rolling window and steps the bitrate up or down per segment. When real throughput drops below what the current ladder rung needs, the player's buffer drains faster than it fills, and eventually hits zero. The player then stops playback, waits to accumulate enough buffered content to resume, and starts playing again.
Structural causes:
- Network throughput dropped below the current ladder-rung bitrate faster than the ABR algorithm could step down.
- CDN fill gap — the requested segment isn't in the local cache, has to traverse the origin or a peer, adds RTT.
- Device-side stall — decoder throughput or disk-write throughput couldn't keep up with delivered bytes.
- TCP / QUIC reconvergence after packet loss or congestion events.
A more efficient codec lowers the bitrate at which a given visual quality can be delivered, so the ABR algorithm's current ladder rung demands less throughput, so the buffer drains slower, so rebuffering interruptions become less frequent.
Netflix's 2025-11-13 datum¶
Netflix publishes the first public fleet-wide rebuffering-rate delta for AV1 vs AVC and HEVC (Source: sources/2025-12-05-netflix-av1-now-powering-30-of-netflix-streaming): "AV1 sessions use one-third less bandwidth than both AVC and HEVC, resulting in 45% fewer buffering interruptions."
The causal chain is the obvious one: AV1 achieves the same quality at 1/3 less bandwidth → ABR's effective headroom over real throughput grows → buffer empty events become rarer → rebuffering interruptions drop 45%.
Two features of the datum:
- Session-weighted, fleet-wide. Not per-title, not per-region, not per-device-class. Per-region rebuffering-delta would plausibly be higher in bandwidth-constrained markets.
- Point-in-time snapshot (2025-11-13) — Netflix flags this in footnote 1 of the post.
Netflix does not publish the absolute rebuffering rate, only the delta. Absolute numbers are held as competitive data.
Relationship to VMAF¶
Rebuffering rate and VMAF are orthogonal QoE metrics — a session can have high VMAF and many rebuffering events, or low VMAF and none. The industry standard is to report both:
| Axis | What it measures |
|---|---|
| VMAF | Perceived visual quality of the frames that were played |
| Rebuffering rate | Frequency of playback interruptions |
| Startup delay | Time from play-button press to first frame |
Netflix's AV1@30% post reports all three directions: +4.3 VMAF vs AVC (quality), 1/3 less bandwidth (efficiency), 45% fewer rebuffering interruptions (smoothness). The three together frame a better-on-every- axis move.
Why codec efficiency is the main lever¶
The other levers — better ABR algorithms, larger buffers, lower-latency CDN fills, better device decoders — are all effectively fixed at any given moment per-session: the player can only use the buffer space the device allocates, the CDN response is only as fast as the network lets it be, the ABR algorithm is whatever the client ships. Codec efficiency is the only lever the server side fully owns that shifts the throughput-vs- quality trade-off uniformly across the fleet. That's why AV1 → 30% of streaming translates directly into a fleet- wide 45%-fewer-rebuffering-interruptions outcome.
Seen in¶
- sources/2025-12-05-netflix-av1-now-powering-30-of-netflix-streaming — canonical wiki source; Netflix AV1 vs AVC/HEVC delta: 45% fewer rebuffering interruptions via 1/3 bandwidth reduction.
- sources/2026-04-02-netflix-smarter-live-streaming-vbr-at-scale — second canonical Netflix datum on a rebuffering-rate delta driven by an encode-pipeline change: ≈5% fewer rebuffers per hour on Live after the CBR → capped VBR cutover at matched quality. Mechanism: easier-scene bitrate drops give the player's buffer more headroom per unit of link throughput; hard-scene spikes are bounded by QVBR's cap. Shows rebuffering-rate is sensitive not just to codec efficiency (AV1 case) but to rate-control mode (VBR case) — two independent levers moving the same QoE metric.