Ecom Ad Radar

Brand watchlist + creative diff timeline for Meta Ad Library — not another generic ad dump.

Built for ecommerce media buyers, agencies, and DTC teams who will pay for change detection, not raw JSON.

Why companies pay (moat)

Most Apify Ad Library actors stop at “scrape ads once”.

Ecom Ad Radar persists a brand snapshot between runs and emits:

Signal	Meaning
new	Creative appeared since last monitor
stopped	Was active last run, gone now (killed / paused)
changed	Same ad ID, fingerprint drifted (copy / CTA / media)
isWinner	Still live + longevity + score threshold

That temporal layer is what agencies schedule daily and put on Slack.

Output

Each dataset row includes scrape fields plus:

• changeType, changedFields
• fingerprint / previousFingerprint
• firstSeenAt, lastSeenAt, daysSinceFirstSeen
• isWinner, offerSignals, creativeScore, longevityDays
• cleaned DPA/DCO copy (no {{product.brand}} when cards exist)

Key-value outputs:

• OUTPUT_REPORT / OUTPUT_REPORT.md — agency one-pager (headline, winners, new/stopped/changed)
• OUTPUT_SUMMARY — per-brand counts + alerts
• ALERTS — short human-readable change digest

Optional Slack / webhook URL input sends NEW/STOPPED/CHANGED alerts after monitor runs.

Dataset views: Overview, New & Changed, Winners, Stopped.

Input (agency setup)

1. Run mode = monitor (default)
2. Brand watchlist = list of competitor Page IDs
   Example: 559929417490965 (Kylie Cosmetics — demo only; use your competitors)
3. Proxy = Apify RESIDENTIAL + country US
4. Optional: Slack / webhook URL for change alerts
5. Schedule daily on Apify (Schedules → every 24h)

Optional:

• winnerMinDays / winnerMinScore — winner radar thresholds
• snapshotStoreName — named KVS for history (ecom-ad-radar-snapshots)
• scrape mode — one-shot dump without diff

Local

npm install
npx playwright install chromium
node src/test-extract-sample.js
apify run

Push

apify login
apify push

How it works

1. Playwright opens Ad Library (handles Meta rd_verify / 403 HTML).
2. Parses SSR + intercepts GraphQL while scrolling.
3. Enriches offers / score / platforms / startDate; cleans DPA templates.
4. In monitor mode: load previous snapshot → diff → save next snapshot.
5. Push ranked dataset (new → changed → winners → active → stopped).

Notes

• Do not paste personal Facebook cookies into the Actor.
• Diff quality improves after the second scheduled run (first run saves baseline).
• Deep pagination depends on Meta UI + proxy quality.