SleepClinicFinder.com

The problem

Someone snores badly enough that their partner sends them to the doctor. The doctor says “get a sleep study”. In Ontario that raises five questions at once: is it OHIP covered, do I need a referral, can I do it at home, which clinics near me do it, and if it turns out to be sleep apnea, is a CPAP machine the only option or will a dentist-made appliance do?

None of that was answerable in one place. Sleep clinics, CPAP vendors, airway dentists, myofunctional therapists and children’s sleep programs each live in their own corner of the web, and most of their sites don’t say whether they bill OHIP or need a referral.

What it is

A directory of 221 sleep clinics across Peel, Halton and the GTA — Mississauga, Brampton, Oakville, Burlington, Milton, Halton Hills, Toronto, Etobicoke, North York, Scarborough, Vaughan and Markham — organised by the six types of care people actually search for:

• Sleep study clinics (in-lab polysomnography and home sleep tests)
• Sleep apnea treatment and CPAP providers
• Dental sleep medicine and airway dentists
• Myofunctional therapy providers
• Pediatric sleep clinics
• Plus ten patient guides on cost, referrals, symptoms and treatment choices

Each profile states what the clinic offers, who works there, whether it bills OHIP, whether it needs a referral and whether it does home sleep tests — and every one of those claims is quoted from the clinic’s own website, not inferred.

The second directory is the interesting one

SportClinicFinder was the first build: national, 12,770 clinics, and a hard lesson about programmatic SEO when a Google core update cut its indexed pages from roughly 15,000 to about 1,000.

SleepClinicFinder is what I built after learning that lesson. It reuses the same Next.js, Neon and Vercel architecture, but the strategy is inverted:

• Regional, not national. 221 clinics in one region instead of thousands across a country. Small enough that every profile can be verified by hand-equivalent rigour.
• Evidence-first profiles. The enrichment pipeline does a deep scrape of each clinic site with Firecrawl, then writes the profile with the Claude API under a rule: no service, practitioner or coverage claim without a quote to back it. Profiles came out at a median of 383 words, ranging from about 150 to 650 depending on how much evidence existed. Thin where the evidence is thin, deep where it’s deep.
• Only 40 profiles are indexable. The other 181 clinics are searchable on-site but kept out of the sitemap until they have enough verified content to deserve a page in Google. The sitemap is the quality gate, and enrichment reads its target list from the live sitemap so nothing can slip in mid-run.
• Guides answer Ontario questions. “How much does a sleep study cost with OHIP”, “home sleep test vs in-lab”, “CPAP vs dental appliance”, “snoring vs sleep apnea” — each guide passed a scored content gate before publishing.

What went wrong, and what it changed

Five days after launch the site went down. The free Postgres tier had a monthly data transfer quota, and every page load was pulling clinic data straight from the database. Five days of traffic burned the month’s allowance.

The fix was architectural, not a bigger bill: public pages now render from a committed JSON snapshot of the clinic data, refreshed weekly by a workflow that commits only when something changed. Reads cost nothing, the database can be asleep or down and the site stays up, and only claims and admin writes touch Postgres. A monitor checks uptime and 5xx errors and emails an alert from the site’s own verified domain.

Timeline

• August 9, 2026 — live, submitted to Google Search Console and Bing, IndexNow wired.
• August 14 — rich-profile rebuild: profiles go from a median of 140 words to 383.
• August 15 — outage traced to the database quota; snapshot architecture shipped the same day.

Built in under a week, hardened within one.