A working content and lead-capture engine, built before we go looking for a network to run it through.

01 — The data moat

Real government data, not spun template copy

Every local page reads from an actual ingested dataset — not a paragraph with a city name swapped in. The numbers genuinely change city to city, because they’re queried, not written.

• roofingNOAA Storm Events (Hail + Wind + Tornado) — 5,770 real recorded events, 2015–2025, joined to city by location, not raw county match
• roofingUSDA Forest Service Wildfire Risk to Communities — real national risk-to-structures percentile per city
• hvacNOAA Climate Normals + EIA rates — 20 distinct weather stations, zero city collisions
• foundationUSDA SSURGO and Census ZCTA, true PostGIS area-weighted polygon intersection — 2,720 soil-composition rows across 108 ZIPs
• foundationU.S. Drought Monitor — real weekly current-conditions tied to the same shrink-swell soil mechanism
• plumbingEPA SDWIS — 575 real compliance/violation records across 92 water systems
• pestUSDA Plant Hardiness Zone, resolved per city to a real ZIP centroid
• solarEIA electricity rates (per-state payback economics) + Census ACS roof age (per-city) — a full vertical built entirely from data already ingested for HVAC and roofing, zero new pipeline
• cross-verticalHUD/Census American Housing Survey — real household-reported condition rates feeding roofing, plumbing, and pest control from one integration (honestly limited to the 11 of 20 metros AHS actually surveys)

02 — Programmatic scale

Built for every U.S. city and town — 20 seed metros is the starting point, not the product

This is the actual mission, not marketing framing: a system that can run for any U.S. city or any subset of them, on demand — not a fixed 20-city demo. The route layer already reflects it (ISR’s dynamicParams: true means a city outside the seed list still renders correctly on first request, cached 24h after), and 17 of the data-backed page types below read from sources that are already nationally complete (NOAA, Census, USDA, FEMA) — widening coverage past the seed set is an ingestion-scope decision, not new engineering. Where a source has a genuine structural ceiling (AHS’s federal survey design covers roughly 50 metros nationally, full stop), that’s disclosed honestly on the page itself rather than implied as a temporary gap.

The same holds for adding whole verticals, not just cities. Solar & battery — the 6th vertical — shipped as a complete, data-backed vertical built entirely from data already ingested for HVAC and roofing (electricity rates + roof age), with zero new pipeline. That’s the architecture thesis in one move: the expensive part is a data integration done once, and everything downstream — new cities, new page types, new verticals — reuses it.

03 — Technical SEO, done, not deferred

The infrastructure a real launch needs is already in place

Native XML sitemap and robots.txt (602 URLs, machine-verified), Service schema.org structured data on every city page — deliberately typed as a referral service, not LocalBusiness, so the markup doesn’t misrepresent what ClearPick is to search engines.

• isrISR rendering with 24h revalidation — new city pages render on first request, no rebuild required
• testing198-test Vitest suite plus a 26-case Playwright end-to-end suite (real-browser, mobile + desktop) covering navigation, animation, and timing behavior
• contentEvery displayed metric ranked against the other tracked metros — no unverifiable “national average” claims
• ux48 original, AI-generated photos (one per vertical + one per service page) — no stock-photo licensing, no fabricated testimonials anywhere on the site

04 — Compliance-grade capture

Consent logging built for an audit, not just a checkbox

TCPA consent is enforced server-side, never trusted from client state. On every submission, the exact disclosure text shown, IP address, user agent, and timestamp are logged in the same database transaction as the lead itself — a lead can never exist without its matching consent record, or vice versa. Every consent record also carries a disclosure version number, so a disputed lead can be answered with exactly which version of the disclosure was live at the moment that consumer submitted, not a reconstruction.

05 — Table-stakes competitive parity

A full service & symptom catalog — the baseline content every ranking competitor already has

42 dedicated service pages across all 6 verticals (Ant Control, Dripping Faucets, Roof Repair, Battery Backup, and every symptom in between) — original, fully-rewritten copy in ClearPick’s own voice, cross-linked so no page is a navigation dead end. Correctly represents ClearPick as the referral platform matching homeowners to contractors, never as the contractor performing the work.

• pest22 individual pest pages, each citing a distinct real biological/practical fact — not one template with the pest name swapped in
• plumbing6 symptom pages — Dripping Faucets, Running Toilets, Clogs, Low Pressure, Water Heater Issues, Leaking Pipes
• roofing / hvac3 pages each — replacement, repair/symptom diagnosis, and maintenance
• foundation2 multi-section pages — 15 named warning signs, 3 real soil-mechanism causes
• solar6 pages — rooftop panels, battery backup, solar-plus-battery, retrofit storage, repair, and maintenance

06 — Moat candidate

Same-day storm alert pages — built on radar data, not the 75–90-day-lagged bulk file everyone else uses

A daily Vercel Cron job scans NOAA’s near-real-time NEXRAD radar hail-detection feed nationwide, auto-generates a time-sensitive page for any county hit above a real, citable severity threshold (1″ hail / 50% detection confidence — the insurance industry’s own damage-likely threshold), and pushes it to search engines via IndexNow the same day, before organic crawl discovery would ever catch up. Honestly labeled as radar-estimated, not human-confirmed — a genuinely weaker, but much faster, evidence tier than the bulk historical data every competitor is still working from.

• verified liveA single real detection run created 266 pages across counties from South Dakota to New York in one pass — genuinely national, not clustered near the 20 seed cities
• expiryPages honestly deindex after 60 days and hand off to the matching city’s evergreen hail-history page, if one exists

07 — Conversion & attribution architecture

Built for a pay-per-call relationship, not just pay-per-lead — before we have one

Every vertical/city/topic page ends in a quote-request form (§04). Built and tested ahead of any real network relationship — retrofitting attribution onto historical leads isn’t possible once they exist, so this shipped before it was needed rather than after:

• attributionFirst-touch click ID, campaign ID, traffic source, and landing page captured the moment a visitor arrives and stored with every lead — no reconstruction after the fact
• one ID, two outcomesThe same click ID a visitor is assigned on arrival is the identifier space for both a form submission and a tracked phone call — reporting on campaign/session performance is one query across both conversion types, not a manual reconciliation between two disconnected systems
• postback-readyA secret-gated postback receiver exists and is tested, matching leads/calls back by that shared click ID — the exact field-name mapping adapts to whichever call-tracking platform a network confirms (Ringba is the common case among networks like this), rather than guessing one in advance
• not yet liveDynamic call-tracking numbers are not turned on — that’s a real, open item (which platform, where a call option should appear on the page), not something we’re representing as finished

Live proof

See it running, not described

Dallas–Fort Worth is the one seed metro with live data across all 6 verticals — the fastest way to see the full breadth in one place.

Proposed pilot

One market, one or two verticals, terms open to discussion

Dallas–Fort Worth is the natural starting market — it’s the one seed metro with live data across all 6 verticals, so there’s no ramp-up gap between signing and real pages going live. A lead-post relationship to start, with pay-per-call following once the attribution and postback infrastructure above (§07) is connected to a confirmed platform. Which verticals, what volume ramp, and payout structure are genuinely open — this section exists to start that conversation, not to presuppose its outcome.

• marketDallas–Fort Worth, TX — all six verticals already live
• modelPay-per-lead to start (form-based, fully consent-logged per §04); pay-per-call to follow once §07 connects to a real call-tracking platform
• volumePre-launch — first-30/60/90-day ramp based on organic performance, or a paid-traffic test if that’s preferred
• askPayout structure and terms discussion, and your lead-intake API spec so integration work can be scoped on our side