How we score

Every number on this board is Surphmore's opinion, computed by an automated pipeline from publicly observable signals on a stated scan date. This page discloses the whole method: what we measure, how it is weighted, what we deliberately do not claim, and how to get anything corrected. Reasonable people using different signals or weights would reach different numbers. That is the nature of a score.

The two numbers

Each business gets a Digital Presence Score (0 to 100, the weighted sum of the six pillars below) and a separate AI-Readiness score (a re-weighting of the signals that determine whether AI assistants can find, trust, and recommend the business). AI-Readiness is suppressed entirely when we could measure less than 45% of its inputs, rather than shown as a misleading number.

Six pillars, weights disclosed

Grades and gates

Grades are fixed bands on the final score: A at 85 and above, B at 70, C at 55, D at 40, F below that. Two conditions cap the score regardless of everything else, because they undermine the whole footprint: no live, self-owned HTTPS website found in the scan (capped at 39), and a Google Web Risk safety flag (capped at 39). A Web Risk flag can be a false positive; we say "flagged as possibly unsafe at scan time," and a clean re-check lifts the cap.

What a missing signal does (and does not) mean

• measured — we observed the signal and scored it, with the evidence shown on the card.
• not found in scan — our automated scan looked and did not find it on the stated date. That is an observation about our scan, not a guarantee it does not exist. Tell us if we missed something.
• awaiting data — we could not measure it yet (a blocked crawler, a pending verification, or a data layer we have not run). It is removed from the math entirely, never scored as a failure.
• n/a — not applicable to this kind of business; removed from the math.

When too little was measurable to be fair, we publish "not scorable" instead of a number. If our crawler is blocked or a lookup fails, the signal is held for re-check, never counted against the business. When our automated matcher cannot confidently verify that a Google or Foursquare listing belongs to a business (we require a website match, or a name-and-locality match with phone corroboration), the signal is held as unconfirmed rather than scored against a possibly wrong entity.

Data sources and what we publish

Signals are collected at scan time from public sources: Google Places, Google PageSpeed Insights, Google Knowledge Graph, Google Web Risk, YouTube, Foursquare, direct fetches of the business's own website, live web-search sampling, and consumer-style AI-assistant test queries we run ourselves. Powered by Google. Powered by Foursquare. We publish our derived scores, bands, and observations, not the providers' raw data: ratings and review counts appear as banded ranges, review dates as age brackets, and we never republish review text. Each provider's data remains its own.

Scan dates are printed on every scorecard. Signals from paid pulls refresh on a rolling cadence, so a card reflects the web as our scan saw it on that date, not necessarily today.

Corrections, re-checks, opt-out

Every scorecard has a "Is this your business?" panel. Report a factual error and we will review it within 5 business days and correct anything we got wrong. Request a re-check after you make changes and we will re-scan. Ask to be removed and we will honor it. These are commitments, not courtesies: hello@surphmore.com.

The short version: disclosed inputs, disclosed weights, dated observations, hedged claims, held-not-punished missing data, and a fast correction path. If you think any of this is wrong or unfair, we want to hear it.