Heavymetric: Auto-Generating Sales Prospects for Local Businesses

What Heavymetric Actually Is

The link that started this — Heavymetric — returned a 404 when we checked, so there's nothing to review on the page itself. What's worth writing about is the workflow behind it, described by the person who built it. It's a lead-generation experiment: find local businesses that lack a website, generate a demo site for each one automatically, and text them a link as a sales pitch.

The URL structure (`/prospects/cincy-fence`) tells you the shape of it. Each prospect gets its own auto-generated page. A fence company in Cincinnati that has a Google Maps listing and a Facebook page but no real website suddenly has a working demo site built from their own photos and copy, delivered to their phone. That's the hook.

This is a demo, not a product with published pricing or clients. The builder's own framing was "I just made this for fun to see if it could be done." Treat it as a pattern to learn from rather than a service to buy.

How the Workflow Fits Together

Broken into steps, the pipeline is straightforward and each piece uses a commodity tool:

1. Scrape Google Maps. Pull local business listings for a category and city using Apify. You get names, categories, phone numbers, and — critically — whether a website is listed.

2. Filter for no website. Businesses without a site are the prospects. This is the whole targeting logic: they have a clear unmet need and something to show them.

3. Rank or qualify with a vision model. The builder mentions using Gemini Flash's vision capability to score sites cheaply — "pennies to rank 100 sites." You can point the same idea at existing sites to judge quality, or at Facebook pages to check if a business is still active.

4. Gather content from Facebook. For businesses with no website, their Facebook page becomes the source of truth — logo, photos, hours, services, and copy.

5. Scrape and structure the content. Firecrawl pulls the raw content, images, and assets. Gemini Flash then fills a template with that information — text and images arranged into a finished-looking page.

6. Generate the site and send the link. Each prospect gets a hosted page, and the phone number from step one gets a text with the URL.

The claimed economics are the eye-catching part: roughly 20 cents to generate 500 sites, because Flash-class models are cheap and the templates do the heavy lifting. That number is the builder's own estimate, unverified, but it's plausible given how little compute a templated fill-in requires.

If you want to build something like this without hand-coding every step, an agent orchestrator or a visual automation platform handles the glue. Make and n8n can both call Apify, Firecrawl, and a model API in sequence and trigger an SMS send; Zapier works too if you prefer its connector library. The model calls could route through OpenAI or Claude instead of Gemini if you already have those set up — the vision-and-fill logic is the same.

Why It Matters, and Where To Be Careful

For anyone doing outbound sales to local businesses, this flips the usual pitch. Instead of describing what you could build, you show up with the thing already built from their own materials. A working demo is a far stronger opener than a cold email, and generating it costs almost nothing. The same skeleton applies to agencies, freelancers, and SaaS founders targeting SMBs: scrape a market, qualify with a cheap model, personalize at scale.

The honest caveats matter here, because this is exactly the kind of workflow that gets abused. Unsolicited SMS is heavily regulated — in the US, the TCPA carries real penalties for texting people who didn't opt in, and carriers filter aggressively. "Text 500 businesses a link" is trivial to automate and easy to turn into spam that damages your sender reputation and your brand. Scraping Facebook also runs against platform terms, and repurposing a business's logo and photos into a site you built for them without permission is a copyright and trust question, not just a technical one.

There's also the demo-vs-reality gap. A Flash model filling a template from scraped Facebook posts produces something that looks finished but may get details wrong — wrong services, stale hours, a mismatched logo. At 20 cents for 500 sites, the temptation is volume over accuracy, and a wrong site sent to a business owner reads as careless, not impressive.

Used carefully — smaller batches, accurate content, an actual opt-in or a warm channel instead of blind SMS — the underlying pattern is genuinely useful. The parts are all commodity now: Maps scraping, a cheap vision model, a scraper, a template, a send step. The skill is in judgment, targeting, and follow-through, not in the automation itself.

If you want help wiring a lead-gen and site-generation pipeline like this into a compliant, maintainable workflow, browse the provider directory to find someone who can put it to work.