Skip to main content
FirstCoast.ai
All case studies
Case study: Productized diagnostic

Business Gap Scanning Service and Deep Dive Audits

A business owner submits a name and a URL. Minutes later they get a report naming what a stranger, a search engine, and an answer engine can see about their business, what is wrong, and who should fix each item. Free, at about three cents of compute.

How the business gap scan and the paid deep dive workA business owner submits two things on a self-serve form: their business name and their website URL. Nothing runs until the submitter verifies their email address, and per-submitter and per-target rate limits sit behind that, so a scan can only be started against a target by someone who has proven they can receive mail. The free scan then runs 45 automated checks in five families: DNS, email authentication, crawler access, performance, and AI-engine visibility. All of them are deterministic except the AI-engine visibility probe, which is the only model call anywhere in the free scan. The check engine is a single deterministic library vendored into the public web app, and the build fails if the two copies ever drift, so the scan a visitor runs is the scan the tests ran against. A whole free scan costs about 3.4 cents of compute, which is why the free tier has no cost floor under it. Every reading the scan takes is written to an evidence store, and that store is the hinge: both products downstream read from it rather than from the site. The free report renders from stored evidence under rules the tests enforce. Disclosures render before findings. An unrun check reads as unmeasured and is never presented as a finding, and a clean verdict requires real evidence coverage, so the report will not tell a business it is fine on the strength of nothing. Recommendations name who should do the work, and what we would not sell renders before what we would: a licensed professional first, the owner themselves next, and us last, with roughly a fifth of all recommendations routing away from us entirely. Report links work as credentials, so the owner can return to their report without an account. The paid deep dive, bought through Stripe, adds what no probe can reach through a structured questionnaire, and re-scores the evidence already stored rather than going back to the site. It never re-scans: there is no second probe. On that paid tier the card processing fee costs more than the AI does, with compute under 4 percent of the payment fee.WHAT THE OWNER SUBMITSFREE, SELF-SERVEA name and a URLBusiness nameWebsite URLBEFORE ANYTHING RUNSEMAIL VERIFIEDVerified before it runsPer-submitter limitPer-target limitTHE FREE SCAN45 AUTOMATED CHECKSDNSEmail authenticationCrawler accessPerformanceAI-engine visibilityTHE ONLY MODEL CALLone deterministic engine, vendored into the appthe build fails if the two ever driftabout 3.4 cents a scanSTORED EVIDENCEEvery reading, storedkept, so a deep dive can re-score itTHE FREE REPORTDISCLOSURES FIRSTONLY WHAT WAS VERIFIEDAn unrun check reads as unmeasurednever as a finding, and no clean verdictwithout real evidence coverageWHO SHOULD FIX IT, IN THIS ORDERA licensed professionalThe owner themselvesUsabout a fifth route away from usTHE PAID DEEP DIVESTRIPEWhere no probe reachesa structured questionnaire coverswhat nothing can measureNEVER RE-SCANS THE SITEit re-scores evidence already storedcard processing costs more thanthe AI compute doesNO SECOND PROBE

The challenge

  • Small-business owners have no cheap, honest way to learn what is broken about their online presence
  • Most free audits are sales sheets where every finding routes to "hire us"
  • Automated scanners overclaim, presenting what they could not verify as if it were a finding

What we built

A self-serve scan of 45 automated checks covering DNS, email authentication, crawler access, performance, and AI-engine visibility, plus a paid deep dive that adds what no probe can reach, through a questionnaire that re-scores stored evidence without touching the site again.

  • Reports only what was actually verified: an unrun check is reported as unmeasured, never presented as a finding
  • What we would not sell renders before what we would: roughly a fifth of recommendations route to a licensed professional or the owner themselves
  • A clean verdict requires real evidence coverage; the scan refuses to tell a business it is fine on the strength of nothing

How it's built

A deterministic check engine vendored into a public web app, where the build fails if the two ever drift, with per-submitter and per-target rate limits, email-verified scan authorization, and report links that work as credentials. Checkout for the deep dive runs through Stripe, and the AI-visibility probe is the only model call in the free scan.

Results

  • Free scan compute cost: about 3.4 cents, so the free tier has no cost floor under it
  • On the paid deep dive, card processing costs more than the AI does; compute is under 4% of the payment fee
  • Report integrity rules are enforced by tests, not review: disclosures first, evidence named, and recommendations that describe work rather than outcomes

What can a stranger see about your business?

Run the free scan and find out in minutes, no sales call required.