Methodology

No black boxes.

Everything here is built from public GTFS feeds — the open schedule data agencies publish. Here's exactly how the numbers are made.

What "frequency" means

For every stop we count boardable departures during a set of time windows, then divide by the window length to get departures per hour. Because each direction is a separate stop, the number reflects what a rider standing at that exact spot actually experiences.

Which day we measure

To represent normal service, we take the Wednesday and Saturday within the feed's active range with the most scheduled trips — the busiest, most typical days — resolving exactly which services run from the feed's calendar and its exceptions. The dates sampled are shown on every page.

Cities with several agencies

Where a metro's transit is split across agencies (Seattle's trains and buses come from different ones), we merge every listed agency's official feed and pick the representative day over the combined schedule. Each page names the feeds and versions used. One honest caveat: when two agencies serve the same physical stop, each agency's stop record keeps its own counts — we don't guess at merging them — so a shared stop can read slightly less frequent than a rider standing there actually experiences.

The frequency bands

Stops are colored on a green → blue → red scale — green is frequent, blue is middling, red is sparse — and sized redundantly (more service, larger dot) so the ranking still reads even where the colors are hard to tell apart:

• Very frequent — 8+/hour (≤7 min)
• Frequent — 4–8/hour (8–15 min)
• Moderate — 2–4/hour (16–30 min)
• Infrequent — 1–2/hour (31–60 min)
• Sparse — under 1/hour (worse than hourly)

How the grade is computed

Each city's letter grade is a transparent weighted score (0–100), not a subjective call:

• 50% — share of stops with frequent service (every 15 min or better)
• 25% — weekend service retained vs. weekday
• 15% — daily service span (first to last departure)
• 10% — median midday headway

The composite maps to a letter (A ≥ 80, B ≥ 59, C ≥ 45, and so on). The verdict sentence summarizes the strongest and weakest of those four inputs.

Limitations

This is scheduled service, not real-time reliability — it doesn't capture delays, cancellations, or crowding. It counts departures across all routes at a stop, so a busy hub looks frequent even if any single line isn't. And it's a snapshot of one representative day. It's a strong first read on where useful transit reaches — not the last word. Each page notes its feed version and generated date.

The plain-English side — how to actually ride each system — lives at public-rides.com.