Our Methodology
How we turn raw data into an honest, sourced answer — and what we refuse to do.
Last updated .
1. Availability results
For a given location and service, we resolve availability into one of four plain outcomes:
- Available — supported by reliable, sufficiently specific data.
- Likely available — area-level or strong indirect evidence, not address-confirmed.
- Check with provider — signals exist, but precision or freshness is insufficient to assert it.
- Not enough data — no reliable source; we show guidance, not a claim.
We never upgrade a result because of a commercial relationship. ZIP/area results are exploratory; address-level claims require address-level data.
2. Confidence levels
Every availability assertion carries a confidence badge, derived by a single pure function from source tier × geographic specificity × freshness:
The exact “freshness” windows are per-source and are being calibrated as real datasets are integrated; until then we apply the qualitative model above and lean conservative.
3. Rankings & “best” lists
Any “best providers in [place]” list is computed from objective, weighted, published criteria: realistic availability at the location, technology tier (fiber generally beats cable/5G-home, which beat DSL/satellite, context-depending), typical performance from cited sources, reliability signals where reputable data exists, and value. Rules we hold to:
- Weights are explicit and published; changes are logged with dates.
- Commercial status is excluded from scoring inputs.
- If a place lacks sufficient sourced local data, we do not publish a ranking.
- Every figure shown is sourced and dated.
4. Composite scores
Scores like a “good for remote work” area score combine multiple records (is fiber present? cable? 5G home? cell coverage by carrier? typical speeds? redundancy?) through a transparent rubric. We show the inputs, cap precision at the weakest input’s resolution, display the score’s confidence and as-of date, and never imply a guarantee.
5. Data handling
- Source tier and freshness drive conflict resolution when sources disagree.
- Missing data becomes unknown — never fabricated.
- Modeled data (for example, mobile coverage) is labeled and caveated, never treated as ground truth.
6. What our methodology must never do
- Let affiliate or sponsor status change availability, confidence, or rank.
- Present precision beyond what the data supports.
- Publish rankings without sufficient, sourced local data.
- Use unlicensed data for user-facing claims.
Disagree with a result? That’s welcome — tell us and we’ll review it against authoritative sources.