01
Meta Ads Library
Mines active competitor creatives — angles, hooks, offers — and reverse-engineers the products driving the spend. Daily delta on what's new vs yesterday.
// source
Meta Ads Library API
// researcher
Six lanes. One queue of winners.
The Researcher walks Meta Ads Library, AliExpress Trends, competitor bestsellers, cross-store frequency, Google organic + paid, and spy tools nightly. Candidate scoring + supplier matching land you a triaged queue of products to approve, not a 500-row spreadsheet to read.
6 discovery lanes Reverse-image suppliers Operator-reviewed
candidate queue · 11:14
Ribbed Bamboo Sleep Set
meta · organic
Linen Beach Wrap (Sand)
trends · frequency
Magnetic Sand-Free Mat
spy_tools · meta
Foldable Travel Kettle
competitor_bs
USB-C Beard Trimmer
trends
12 new today · 47 in queuenext cycle 11:00 tomorrow
// discovery lanes
Where winners come from.
Six independent lanes, each with its own signal source and its own bias. The candidate scorer combines them — a product surfacing in three lanes outweighs one with a single hot signal.
02
AliExpress Trends
Watches order-velocity curves on supplier-side AliExpress listings. New rising stars surface days before the spy-tool crowd catches them.
// source
AliExpress crawl
03
Competitor Bestsellers
Crawls the bestseller pages of your competitive set on Shopify, WooCommerce, BigCommerce. Cross-references with your own assortment to find gaps.
// source
Storefront crawl
04
Cross-Store Frequency
When a product appears in ten competing stores in a week, it's a market signal. The frequency lane reads the network, not the single source.
// source
Aggregate crawl index
05
Google Organic + Paid
Combines Search Console-style organic rankings with the live SERP ads. Rising queries + advertiser interest = a winning angle, not just a winning product.
// source
DataForSEO SERP
06
Spy Tools + Trustpilot
Hooks into the spy-tool feeds you already pay for, then layers Trustpilot review sentiment so a high-volume product with a 2.1-star average gets a hard pass.
// source
Multi-vendor
// candidate_scorer
Score a candidate like an operator would.
Each candidate is scored on five weighted dimensions — and the weights are tunable per store. Margin and uniqueness lead; market-fit and trend velocity shape the queue; review-health is the cap that keeps you out of trouble.
margin_potential
w=0.30
Estimated landed cost vs the going retail price across competitor set — wide gaps score high.
uniqueness
w=0.25
pgvector distance from existing catalog (yours + competitors). Saturated = penalized.
market_fit
w=0.25
Cross-references your brand_profile category, audience age, region — bad fit downscored hard.
trend_velocity
w=0.15
Slope of demand signal across lanes — fast risers outweigh established but flat performers.
review_health
w=0.05
Trustpilot, Trustindex, and aggregated review sentiment — a multiplier that caps high-risk products.
research_candidates · rc_4421
title : "Ribbed Bamboo Sleep Set" source_lane : meta_ads_library (also: organic) source_url : https://facebook.com/ads/lib/… candidate_score: 0.87 PROMOTE ├─ margin : 0.92 (avg retail €78, est. cost €11) ├─ uniqueness : 0.81 (cosine 0.34 from your catalog) ├─ market_fit : 0.88 (matches brand_profile) ├─ trend : 0.78 (3-week climb, 4 stores) └─ review : 0.91 (4.6 ⭐, 1.2k reviews) supplier_match : found via DataForSEO ├─ url : aliexpress.com/item/100501… ├─ moq : 1 (sample-ready) └─ ship_eu : 9–14d, available DDP queued_at : 2026-05-25T11:02Z awaiting : operator review
// supplier_matcher
Reverse-image to the supplier, in one step.
Pretty product, no supplier? Magistry runs the candidate image through DataForSEO reverse-image search and surfaces AliExpress / Alibaba listings with the same SKU — MOQ, shipping windows, DDP availability included before you ever open a tab.
Reverse-image search
DataForSEO Lens API matches candidate images to known supplier listings — same product, different listings, different prices.
Listing aggregation
We rank suppliers by price, MOQ, EU/US ship windows, and complaint volume from prior orders.
Auto-link on approve
Approve the candidate, the top supplier match becomes the linked source — Catalog Specialist takes it from there.
supplier matches · rc_4421
▼ ranked by composite score #1 aliexpress.com/item/100501… price: €10.80 moq: 1 ship: 9–14d EU (DDP avail) review: 4.8 ⭐ (2.1k orders) → recommended #2 alibaba.com/product/72441… price: €7.20 moq: 200 ship: 28–35d EU review: 4.1 ⭐ (gold supplier) → MOQ too high #3 aliexpress.com/item/103328… price: €11.20 moq: 1 ship: 12–18d EU review: 4.4 ⭐ (340 orders) → backup
// operator review
A queue, not a spreadsheet.
The /dashboard/candidates queue is the only place a researched product touches a human before it's a draft. Approve → onboarding fires; reject → the row is held so future cycles don't resurface it; hold → flagged for next week's research review.
01
Candidate lands
Six lanes write into research_candidates. Each row has source_lane, source_url, raw evidence blob, candidate_score, and supplier_match if found.
02
Operator triages
/dashboard/candidates is a ranked queue. Hit approve, reject, or hold-for-research. Approvals create a draft product in Shopify with the discovered supplier pre-linked.
03
Onboarding kicks in
Approved candidates enter the AUDIT lifecycle state. The Catalog Specialist takes over from there — copy generation, image pipeline, classify cycle.
04
Track the bet
Every researcher-sourced product gets tagged in evidence chains. You can read 'how did Meta-Library-sourced products perform' as a single SELECT.
Onboarding is a state, not a script. An approved candidate lands in AUDIT— and the Catalog Specialist's sync / classify / decide / execute loop owns the rest. Every move from there is a decision_log row you can read.
// researcher
Stop scrolling spy tools at midnight.
Six lanes run nightly. Candidates are scored, suppliers matched, queue ranked. You spend the hour approving, not the week researching.
6 lanes · Reverse-image matched · Operator-reviewed