Internet Quality Ranking Methodology
Learn how Toolora crowdsourced internet quality rankings use medians, minimum samples, grid diversity, confidence, stability, and rank change.
Speed test and sampling
Tests are initiated voluntarily in a supported browser. Download, upload, ping, and jitter are measured against same-origin endpoints. Results can vary with the device, Wi-Fi, server path, local traffic, and test time, so one test is an observation rather than a coverage claim.
Aggregation, averages, and percentiles
Public map points reduce coordinate precision and combine nearby measurements into grid cells. Regional summaries include count, median, average, P25, P75, variation, active days, provider diversity, and connection diversity. Heatmaps use aggregated cells within the requested bounds and period.
Data quality and coordinate privacy
Input ranges, UUID uniqueness, rate limits, minimum samples, active-day coverage, and location-grid diversity support data quality. Public responses never include raw IP addresses or precise individual histories. Coordinates are optional and reduced in precision in public output.
Data source and privacy
Results come from browser tests voluntarily submitted by Toolora users. Public rankings contain aggregate values only. They do not expose test UUIDs, individual coordinates, IP addresses, sessions, or full user agents.
Median, percentiles, and variation
Median is the primary value. P25 and P75 describe the middle distribution, while standard deviation describes variation. Statistical outliers are not automatically treated as invalid measurements.
Minimum eligibility
Entities must satisfy configured test counts, active days, and unique location grids. Default thresholds are 30 tests for provinces, 20 for cities, 50 for national providers, at least three active days, and two grid cells.
Ranking formulas
- Download/upload: highest median, then P25, sample count, active days, and lower variation.
- Ping/jitter: lowest median, then P75, lower variation, and sample count.
- Stability: lowest download coefficient of variation.
Rank change and update frequency
Rank change compares an equal-length preceding period. A missing eligible previous result is shown as New. Rankings are generated by scheduled maintenance scripts and are never calculated during a public page request.
Known limitations
Toolora data is crowdsourced, may be uneven between locations and providers, and is not an official government statistic or a complete representation of provider coverage.
Frequently asked questions
Does a first-place result mean the provider is always best?
No. It means the entity ranked first for the selected metric, filters, period, and eligible Toolora samples.