IPTV Services Tested — 2026 Lab Cycle

Methodology, instruments, raw findings. Stream Lab HQ test cycle Q2 2026.

This is a methodology document with results, not a review. The intent is that another lab could rerun the same protocol and get within ±5% of these numbers.

Test bench

• Bench A: Fire TV Stick 4K Max, firmware 7.6.x, factory reset before cycle, only the IPTV-under-test app installed plus a screen-recorder
• Bench B: Apple TV 4K (3rd gen), tvOS 18.x, same isolation
• Bench C: Samsung Tizen 7.0 65" TV, native HLS playback where supported

All three benches connected to a managed 1 Gbps fibre line, no other household traffic during test windows.

Protocol per service

T+0       cold install, sign in, log credentials issuance time
T+0..30m  channel inventory walk: load each EPG channel, log first-frame time
T+30m..2h continuous play of a sports channel, capture every 5min via ffprobe
T+2h..3h  controlled throttle — 50/25/10/5 Mbps caps — log adaptation
T+3h..4h  cold-start playback five times; average startup latency
T+24h     repeat 4-hour window

We run this protocol on each bench once per service, weekly for 4 weeks.

What we report (per service)

1. Inventory honesty — channels claimed vs. channels that load
2. Delivered bitrate distribution — median, p10, p90
3. Buffer rate — events per hour at full bandwidth and at 5 Mbps cap
4. Failover behaviour — does the player drop quality cleanly or stall?
5. EPG drift — % of slots where the playing programme didn't match the published guide

We do not report "value", "vibes", or any composite score. Buyers can weight the numbers themselves.

Findings — Q2 2026 cycle (summary)

The full per-service breakdown is in findings/ (publishing on a rolling basis as cycles complete). High-level patterns this cycle:

• Three of twelve services dropped >40% of their advertised channel count on first walk — these were not the cheapest services
• Median delivered bitrate at unrestricted bandwidth: 18 Mbps for the top quartile, 8 Mbps for the bottom quartile
• Two services failed the throttle test by stalling instead of stepping down quality — both lost their stream entirely below 10 Mbps
• Buffer-rate at unrestricted bandwidth was a poor predictor of buffer-rate at 5 Mbps; tested both, they don't correlate

Related work

• Methodology document (this README) — supersedes 2025 cycle protocol
• Test scripts: scripts/ (planned, releasing as redacted variants)
• Reference customer-facing summary: streamreviewhq.com

---

Last verified: June 05, 2026