105,000+
places on the platform
89.9%
carry a computed, confidence-capped score
83.7%
of on-the-ground community verdicts confirm our record
Same day
median time from community submission to human review
01The question this page answers
The most common skeptical question we receive is also the best one: you cannot have visited 100,000 places, so how can you score them? The premise is correct. Every map ever made is an inference machine built from records, reports, and corrections. The difference between a trustworthy one and an untrustworthy one is whether it tells you how much evidence sits behind each claim, and whether its scoring is allowed to express confidence the evidence does not support. ROLLIN is built on that distinction, and this page reports exactly where the evidence stands.
02The lifecycle of a location
Every location carries a label saying which stage its data has reached, in the app, on the web, and in the API. Movement is one way and evidence-gated. Nothing skips a stage, and nothing can be purchased.
01
Documented
Every listing begins as documentation that a place exists: identity, category, coordinates, and physical context from a multi-source pipeline.
02
Scored
A deterministic engine computes 0 to 100 from the evidence on file. Missing evidence caps the score. Unknown is never presented as good.
03
Refined
People who navigate these rooms confirm and correct the record, weighted by each contributor’s accuracy track record. Influence is earned by being right.
04
Verified
A fact graduates from reported to verified only through independent confirmation: a second observer or an instrument-grade capture. Nobody can verify themselves.
03The census, by evidence depth
22% of listings carry documented accessibility features. The remainder are documented places whose scores stay capped until evidence arrives. The cap is the design: the size of the dataset cannot manufacture certainty it has not earned.
Feature-documented · 22% Awaiting feature evidence · 78% · scores capped
| Evidence depth | Operating since | Coverage | Status |
|---|
| Human-verified on the groundIndependent confirmation, instrument capture, on-site review, or certification audit | Dec 2025 (community), Jul 2026 (instrument) | first-year program | tracked each edition |
| Accessibility features documented from recordsAt least one of six features on file | 2026 | 22.0% | |
| Awaiting feature evidenceExistence documented; scores capped by design until evidence arrives | | 78.0% | scores capped |
| Places on the platform | | 105,000+ | |
04Growth against a published baseline
The verified programs are young: community verification since December 2025, instrument capture since the summer of 2026. This edition is the baseline. Every future edition reprints the same census and the changes will be visible here, in both directions. Improvement claims that start from a flattering number can never be proven. Ours can.
Community contributions, cumulative
Reviewed community submissions since launch. Complete months through August 2026.
05How the record gets stronger
Each limitation maps to work under way at ROLLIN Labs, the applied research arm of the company.
The instruments
ROLLINSIGHT turns the sensors already in a modern phone into calibrated field instruments. Lighting and sound measurement is operational in production today. Doorway geometry and live photographic evidence are in field testing. Every reading is tagged with how it was taken.
ROLLIN Labs Why this compounds
A reading publishes as reported. An independent confirmation promotes it to verified. The verified core grows with the number of people carrying the instrument, not with staff headcount. It is the only verification mechanism we know of that scales honestly.
The goals we accept being measured against
Grow the human-verified core every edition, deltas printed here. Extend instrument capture to doorway geometry and photographic evidence. Close the restroom evidence gap. Scale documentation toward one million US places under the same caps and labels. Publish this report on schedule, whatever the numbers say.
06What we refuse
We do not sell scores.
No payment, sponsorship, or certification fee can raise a score, release a cap, or move a fact from reported to verified. Certification is a failable, on-site standard.
We do not guess.
Unknown is displayed as unknown. The engine penalizes missing evidence instead of papering over it.
We do not show unearned certainty.
Unverified critical features cap the score, structurally. The 78% of places without feature evidence cannot score like verified ones.
We do not let anyone verify themselves.
Verified status requires independence: a second observer or an instrument capture, never a self-confirmation.
We do not publish numbers we cannot stand behind.
Every figure on this page is measured and dated, and each edition refreshes it. The next edition prints the changes, in both directions.
We would rather show no data than wrong data.
This has been the operating rule since the first location entered the database.
07The reports
Both documents are public editions: complete on method and evidence, silent on thresholds, weights, and source agreements.
08Check our work
Sample the API
The public REST API returns the same provenance labels and scores the app shows. Pull any set of locations and test our claims against the ground.
Get a free key Correct the record
Every location screen offers a correction path, and the median review is same day. If we are wrong about a place you know, make us fix it.
Find a place Hold this page to account
This is edition v1.0. The next edition reprints the same census from the same definitions, and the deltas will be visible here.
ROLLIN Labs 