Methodology
Figures on this site as of 2026-07-20.
The population figure ("X million people without internet access")
This number is a sum of township populations, not an estimate. We only count a township
toward this figure if it is explicitly named in the scope of a currently
ongoing event whose type is full_blackout or mobile_data
(a platform block or throttling event does not mean the population has no access at all, so
those event types are excluded from this figure specifically — they still appear as their own
events elsewhere on the site).
Where an event's documented scope is a region but does not enumerate specific townships, we do not estimate a share of that region's population. That event contributes zero to this figure. This makes the number an undercount by construction — the honest direction to be wrong in, given the alternative is guessing at a region-wide population that may not be accurate.
Township population figures come from UN OCHA HDX — Myanmar Subnational Population Statistics (COD-PS), reference year 2023. Township name matching against event records uses a small, manually verified list of spelling variants (e.g. "Bawlakhe" / "Bawlake") — never a nearby or similarly-named township standing in for one that doesn't match.
Currently excluded due to no population-dataset match (not itself a township, or an unrecognized spelling variant): Payanthonzu.
"Longest ongoing blackout" (day count)
The number of days since the date_start of the ongoing event with the earliest
start date in the corpus. Computed client-side from that fixed date, so it advances daily
without requiring a site rebuild.
Event confidence tiers
Every event card and event page carries one of four confidence badges. The colors match throughout the site — the same badge color on a homepage card means the same tier on that event's full page.
- confirmed
- 3 or more independent technical detection layers (BGP, Cloudflare, IODA, OONI), or 2 technical layers plus an Access Now STOP record.
- corroborated
- Exactly 2 layers, at least one technical.
- verified (ground truth)
- An Access Now STOP record and/or media reporting only, no independent technical detection layer.
- signal
- A single unconfirmed technical detection. Excluded from the homepage entirely, not just visually de-emphasized.
What this site does not claim
Country-level or region-level technical data (Cloudflare traffic, OONI measurements) is never presented as township-level evidence. A technical signal at the country or region level can corroborate an event, but the township-level scope of an event is always sourced to a human-verified record (Access Now's STOP dataset or media reporting), not inferred from technical data alone.
Regional speed trends (Ookla)
Per-region median download speed change, first available quarter (Q1 2021) vs. most recent (Q1 2026), shown on the map only where sample counts clear a defensible threshold — at least 100 test samples in both the first and last quarter for that region and layer (mobile vs. fixed broadband are tracked separately). Below that threshold, no stat is shown for that region rather than a shaky one; several regions fail this on the mobile layer specifically because of a nationwide sample-count collapse from mid-2024 onward (see the site's internal research log for the full analysis — the collapse's timing lines up closely with the junta's May 2024 VPN-blocking order).
Selection-bias caveat: Ookla's data reflects people who ran a Speedtest — a population skewed toward those with working internet access, motivation to test it, and a compatible device. It is not a random sample of the general population, and a rising median speed among people who can still test does not mean connectivity improved for people who lost access entirely (they wouldn't appear in this data at all). Read as a trend among the still-connected, not a population-wide connectivity measure.
Speedtest® by Ookla® Global Fixed and Mobile Network Performance Maps was accessed on 20 July 2026 from AWS (aws.amazon.com/marketplace/pp/prodview-breawk6ljkovm). Based on this site's analysis of Speedtest® by Ookla® Global Fixed and Mobile Network Performance Maps for Q1 2021 – Q1 2026. Ookla trademarks used under license and reprinted with permission. License: CC BY-NC-SA 4.0 — non-commercial use only. This site does not use Ookla data for any commercial purpose — noted here because it constrains any future monetization of this site's data products.
Data sources
- Access Now STOP dataset (human-verified shutdown records)
- IODA (Georgia Tech) — regional outage alerts
- Cloudflare Radar — country-level traffic
- OONI — network measurement anomalies
- RIPEstat — BGP routing visibility
- Ookla Open Data — regional speed trends (CC BY-NC-SA 4.0)
- UN OCHA HDX — Myanmar Subnational Population Statistics (COD-PS) (2023)