See where every carrier really performs — based on real-device measurements.
An independent, device-measured read on where every carrier actually delivers — built from the LTE Discovery app community, de-identified, and aggregated. This is a single-metro slice of a nationwide dataset of billions of real measurements.
How the networks compare in San Francisco
Different carriers lead on different things — broad coverage, raw signal, 5G maturity, and deployed capacity are separate questions. Here's how they stack up.
Built for teams who need real-device evidence
You've seen how the networks compare — here's your seat. Each jumps to the proof that matters most for you, and tees up a conversation scoped to your use case.
Is your market in the data?
Explore sample availability in selected US markets. Request a custom sample to assess whether the data fits your use case.
Availability indicates published observations, not complete network coverage. Sampling follows our app community; ask about density, recency, or an unlisted market.
Where service is strong — across the city grid
Each cell is colored by real-world signal strength (P90 RSRP). Cells are H3 res-8 (~0.7 km² each); finer where data is dense.
Strongest LTE signal by location
Across the multi-year measurement history, the share of sampled area — every mapped cell — where each carrier posts the strongest LTE signal. This is not a share of the city's total area.
Signal-quality mix
Coverage isn't binary. This is how each carrier's footprint breaks down by real signal quality — usable strong service vs. marginal edge-of-cell.
Full carrier × band matrix
Every band each carrier has actually lit up in San Francisco — generation, breadth of deployment, median signal, and reliability. 5G/NR bands highlighted.
Crowdsourced (not drive-test); aggregated with a ≥5-reading-per-cell minimum, so sparse areas read as no-data; density and carrier mix follow our app user base; San Francisco only — the full dataset is nationwide.
How this is measured
Based on de-identified, aggregated device observations. Signal statistics summarize readings; cell reach is a derived estimate.
Crowdsourced, passive
Built from real LTE Discovery app users’ devices, measuring continuously as they go about their day — not a one-time drive test.
Based on real-device measurements
Signal metrics summarize device readings, including RSRP and SINR. Cell reach is derived. Sampling follows our app users, not a controlled, equal-weighted census.
Aggregated, sparse-cell suppressed
Only cells with at least 5 device readings are shown; sparse areas are suppressed (a minimum-sample floor) — so an empty cell means “no data”, not “no coverage”.
De-identified & aggregated
Readings are aggregated to H3 cells before you see them. No account identifiers are attached — and we don’t sell individual device traces.
Historic + live
Measurements span 2017–present — a deep history plus an ongoing live edge, not a single current snapshot. The SF blend is density-weighted toward the early LTE years; the 5G figures reflect the recent edge.
Download the San Francisco sample
Evaluate the real data, not just a page. These are the exact figures behind every chart above — exported straight from this metro sample, ready for your own analysis.
Preparing exports…
The API is real — probe it yourself
The paid product is an API. This is the live, keyless catalog endpoint — no signup, no data exposure, just the real product schema, straight from the server.
curl -s https://ltediscovery.app/api/partner/v1/catalog
Data endpoints require an API key. Start a self-service 14-day evaluation in the commercial portal, or request a custom sample below. See current plans
Data dictionaryfield reference across all exports
- network
- Carrier / mobile network operator (e.g. T-Mobile, Verizon, AT&T).
- generation
- Radio generation of the band — LTE (4G) or NR (5G).
- band
- 3GPP frequency band number (B-prefixed for LTE, n-prefixed for 5G NR).
- hex_count
- Number of distinct ~0.7 km² H3 cells (res-8) the carrier covers on that band.
- samples
- Total deidentified device readings aggregated into the row.
- median_p90_rsrp
- 90th-percentile RSRP (dBm) per cell, then taken at the median across cells — a strong-signal indicator.
- median_rsrp
- Median of per-cell P90 RSRP values, not the median of all device readings. Higher (closer to 0 dBm) is stronger.
- avg_reliability_pct
- % of samples above the usable-signal threshold — how dependable service is, not just present.
- coverage_cells
- Distinct res-8 cells where the carrier is measured (its footprint size).
- nr_cells
- Distinct cells with measured 5G NR service.
- q_excellent / q_good / q_fair / q_poor
- Cell counts by signal-quality bucket (≥ −85 / −85 to −95 / −95 to −105 / < −105 dBm).
- minimum-sample floor (≥5)
- Only cells with ≥5 device readings are published; sparser cells are suppressed (a minimum-sample floor — not a guarantee of 5 distinct devices), so a gap means “no data”, not “no coverage”.
What full access includes
The SF sample is one slice of three products. Full access opens all of them, for the markets you choose — each proof stat below is real, pulled from this metro. See the full product & use-case overview →
Per-carrier, per-band signal quality on H3 cells — RSRP, reliability, quality mix. API, CSV, or GeoJSON.
Hex-cell coverage tiles (~0.7 km² res-8)Which bands each carrier has actually lit up, depth of carrier aggregation, and 5G SA/NSA maturity.
Per-band deployment by carrierThis page, plus CSV + GeoJSON downloads. No login, no cost — evaluate the real data.
You’re here ↑ try the downloadsSelf-service monthly plan — see the commercial portal for current pricing, units, and markets.
See current plansHigher-volume self-service plan — see the commercial portal for current pricing and limits.
See current plansNegotiated embedded and OEM rights, nationwide API and bulk delivery, custom markets and cadence, and optional attribution waiver.
Talk to usPricing above is live from the current plan catalog. Start with a 14-day internal evaluation; embedded use and bulk delivery require negotiated Enterprise terms.
Why this beats a consumer speed-test app
Speed-test apps tell you a number at a point. This is engineering-grade RF intelligence — per-band, per-cell — across an entire market.
Comparison is to the consumer speed-test category in general, not any specific app.
Start a free pilot
Tell us your market and use case and we’ll send a custom sample — free, scoped to you. No commitment: see the real data for your area before anything else.
Prefer email? Reach the team at support@ltediscovery.app.