Methodology

How it works

WifiRank exists because ISP reviews in Bangladesh contradict each other. We replace opinions with measurements, and we show exactly how those measurements are made.

What the test measures

The test runs in your browser against Cloudflare’s server in Dhaka, using Cloudflare’s open-source speedtest engine. We record download and upload speed (Mbps), ping (round-trip latency, ms) and jitter (how much ping varies, ms). Packet loss is measured when available.

This is the domestic path. It reflects how you reach content served from inside Bangladesh, which includes the local caches most big sites use. It does not measure speed to servers abroad, so a provider that looks fast here can still be slow for overseas games, calls or downloads.

The engine ramps up through larger and larger transfers. We report sustained speed: the median of the transfers long enough to load your line (a quarter of a second or more). Short bursts on small files are left out, because many connections start fast and then settle, and the settled speed is what a download or video call actually gets. This reads lower than tools that report peak or 90th-percentile speed.

What we store

• Your results, the provider and area you chose, and how the device was connected (cable or Wi-Fi band).
• Optionally, your package speed and monthly bill, and a map pin rounded to about 500 metres.
• The operator name of the network the test came from, and a hashed IP used only to limit abuse. We never store your raw IP address.
• A random ID kept in your browser, so your repeat tests count as one person. No account, no name, no phone number.

How rankings are calculated

• One person, one vote. All tests from the same device on the same provider are reduced to that device’s median first. Running 50 tests does not count 50 times.
• Medians, not averages. The published number is the median across devices, so a few extreme results cannot move it.
• Last 90 days only. Providers change, so old results expire.
• Minimum sample. A provider needs at least 5 testers to be ranked. Below that it is shown dimmed as an early signal.
• Star ratings use a Bayesian average: a provider with one 5-star review is not ranked above one with a hundred 4.6-star reviews.

How we keep it honest

• Tests run over mobile data or from an IP outside Bangladesh (usually a VPN) are detected from the network they came from and excluded.
• Reviews can only be posted from a device that completed a test on that provider. Each device gets one review per provider, which it can edit.
• Tests are rate-limited per device and per network.
• Providers users add themselves stay hidden until checked.

Known limits

• A browser test measures your whole path, including your Wi-Fi and router. That’s why we ask how you are connected. Cable results are the truest picture of the ISP.
• International speed is not measured (see above). Reviews are the best signal for it today.
• Speed tests cannot see outages. That is what the star rating and comments are for.
• We compare against the package range you report, which we cannot verify.

Open data

All aggregated results are available as JSON at /api/stats and /api/map, free to reuse with attribution (CC BY 4.0). AI agents can start at /llms.txt.