Viral Sourcing
This turns the Codex dropshipping workflow into a concrete NovaMart process: find active demand, extract the full product record, dedupe by family, and only then publish to the catalog.
How To Find Viral Products Without Polluting The Catalog
The job is not to import everything that looks exciting. The job is to find product families with demand, clean supplier data, and real buying paths.
Active ad sweep
Source: Meta Ad Library
What to pull: Products with active ads, repeated hooks, recurring offers, and ad families that have stayed live long enough to suggest demand.
NovaMart rule: Capture the product family and the angle. Do not copy the ad. Do not treat a single flashy creative as proof of a winner.
Short-form pattern check
Source: TikTok Creative Center + Instagram/Reels manual review
What to pull: Products that keep appearing across different creators or hooks, especially if the payoff is obvious in the first two seconds.
NovaMart rule: Use short-form platforms to validate visual payoff. Do not rely on scraping brittle social pages as the primary source of truth.
Trend confirmation
Source: Google Trends
What to pull: Search direction, not just peak volume. The goal is to confirm the product is not already cooling off.
NovaMart rule: A product needs real demand across ads and search before it deserves page work and creative time.
Supplier viability
Source: Official product page or reputable supplier fallback
What to pull: Buyable URL, clean title, real product image, clear price, and a product description that can be rewritten into NovaMart copy.
NovaMart rule: No supplier URL, no import. No real image, no import. No clear use case, no import.
Fields Codex should extract for every candidate
- - Do not import gifts, bundles, early-access variants, locale variants, accessories, or replacement parts as standalone products.
- - Do not import the same product family twice under different promo names.
- - Do not keep copied supplier text. Rewrite the description into a clean buying angle.
- - Do not promote a product without a real buyable source URL and real image.
- - Do prefer official brand pages first. Use generic supplier lanes only when the product is not a flagship trust-led item.
data/viral-candidate-template.jsonINPUT_FILE=data/viral-candidate-template.json DRY_RUN=true node scripts/import-viral-candidates.mjsINPUT_FILE=path/to/candidates.json node scripts/import-viral-candidates.mjsCodex Runbook
This is the exact flow NovaMart should run when turning ad evidence into a real product listing.
Find evidence of demand
Action: Search active ads and short-form patterns for products with visible traction. Save the product family, hook, and source URL.
NovaMart mapping: Research first, import second. Demand proof comes before catalog expansion.
Extract the full product record
Action: Pull name, supplier URL, hero image, price, and the core use case. Keep the raw data separate from the rewritten store copy.
NovaMart mapping: This maps directly to the product fields NovaMart already stores: name, description, shortDesc, imageUrl, price, supplierUrl, supplierName.
Rewrite for conversion
Action: Rewrite the description into a direct buying angle. Keep it clear, specific, and native to the product page.
NovaMart mapping: The store page should read like a recommendation, not like supplier export text.
Dedupe by product family
Action: Check slug, supplier URL, and the normalized product family so variants do not enter the catalog as fake new products.
NovaMart mapping: This is now enforced inside the importer and cleanup scripts.
Publish to the catalog
Action: Use the import scripts to upsert clean products into NovaMart, then verify a sample of live product pages.
NovaMart mapping: NovaMart uses import-official-sitemaps, purge-noise-products, and cleanup-duplicate-products as the current pipeline.
Support fulfillment routing
Action: Once a product proves itself, connect it to the right supplier lane and fulfillment workflow instead of treating every import as equal.
NovaMart mapping: Hero products stay source-led. Generic lanes are fallback support, not the main trust layer.
How This Maps To The Store
Viral sourcing is not separate from the store. It is upstream of the catalog, cleanup, and testing layers already running in NovaMart.
Tools: Meta Ad Library, TikTok Creative Center, Google Trends
Output: Candidate list with angle, source URL, and demand evidence.
Tools: Codex scripts + product importer
Output: Clean product payloads with real links, real photos, rewritten copy, and no duplicates.
Tools: Noise purge + duplicate cleanup
Output: Only distinct products survive. Variants and junk listings are removed.
Tools: Homepage winners, product pages, landing pages
Output: The strongest products get page priority and testing support.
Video and platform references
Use Codex to find active ads, extract product data, rewrite listings, import to store, and support fulfillment flow.
Public transparency surface for active ads. Good for angle and demand validation before importing a product.
Validate whether interest is rising, stable, or already dead before making a product a priority.
What changes because of this video
- - Product research now starts from active demand evidence, not from random supplier browsing.
- - Supplier data has to be complete before import: title, price, image, buyable URL, and a usable angle.
- - Dedupe is now treated as a hard requirement, not cleanup later.
- - Catalog growth should feed the winner pipeline, not create more homepage clutter.