Monster Energy Deal Tracker

A small price-tracking pipeline that checks Amazon twice a day for Monster Energy multi-packs, works out the price per fluid ounce for each listing, and opens a GitHub Issue when something drops to $0.12/fl oz or lower. It runs entirely on GitHub Actions and costs nothing to operate.

I built this because I drink a lot of Monster and got tired of checking Amazon by hand. It has also turned into a decent exercise in keeping a scraper alive against a site that actively tries to block scrapers.

Live output: deal_report.md is regenerated on every run. dashboard.json is a compact feed built for the dashboard on my portfolio site.

How it works

GitHub Actions (09:00 / 18:00 UTC)
  └─ tracker.py
       ├─ headless Chromium fetches 3 pages of Amazon search results
       ├─ parses every Monster listing: price, pack size, $/fl oz, stock
       ├─ re-checks deal candidates on their product page
       ├─ appends to price_history.json, writes deal_report.md, dashboard.json
       └─ current_deals.json → workflow opens/updates an Issue only if a new deal appeared

Fetching. Amazon serves an empty anti-bot page to plain HTTP clients almost every time, so pages are loaded in a real headless browser (Playwright). Each request gets a fresh browser context, there are randomised delays between requests, and blocked responses are retried with back-off. If the browser is unavailable the tracker falls back to requests.

Pricing from search results. A search result card already carries the ASIN, full title, featured price, Amazon's own unit price and the stock or offer status. Reading those means every Monster listing on the results pages gets priced with three requests, instead of opening 100+ product pages. It also avoids the trap the first version of this project fell into: pulling a price from a product page and getting the "lower-priced alternative" widget instead of the buy box.

Pack size. Total fluid ounces are parsed from the title and its variant text, which comes in a lot of forms (16 Ounce (Pack of 15), 16 Fl Oz | Pack of 12, 473 mL Cans, 12 Pack, 16 Oz (24-Pack), bundles like Pack of 15 + Pack of 15). The result is cross-checked against Amazon's unit price. Amazon's figure is not trustworthy on its own (some listings show $1.80/count or a per-ounce price that is off by 10x), so it is only a fallback and a sanity check. Listings where the two disagree are flagged and never alerted on.

Verification. Anything at or below the threshold is re-checked on its product page for the buy-box price, seller, availability and Subscribe & Save price before it goes in the report.

Alerting. The workflow keeps one open Issue. It opens a new one only when a deal appears that was not there last run (or got at least 3% cheaper), closes the previous one, and otherwise just refreshes the body. If Amazon blocks the scrape, nothing is committed, no Issue is touched, and the job fails so it shows up red.

Setup

1. Fork or clone, then push to your own GitHub repo.
2. Settings → Actions → General → Workflow permissions: select Read and write permissions.
3. Trigger the workflow once from the Actions tab to confirm the runner can get through to Amazon.

Configuration

All settings are environment variables, set in .github/workflows/tracker.yml or in your shell.

Variable	Default	Meaning
PRICE_THRESHOLD	0.12	$/fl oz that counts as a deal
MIN_FL_OZ	64	ignore packs smaller than this (4 × 16 oz)
MAX_PAGES	3	search result pages per query
SEARCH_QUERIES	monster energy drink	|-separated Amazon searches
VERIFY_DEALS	1	0 skips product-page verification
USE_BROWSER	1	0 forces plain HTTP (usually blocked)
DEBUG_HTML	unset	1 saves every fetched page under debug_html/

Schedule (cron, UTC) is at the top of the workflow file.

Running locally

pip install -r requirements.txt
playwright install chromium
python tracker.py

A run takes three to five minutes because of the deliberate delays between requests. Unit tests cover the title parsing, search-card parsing, product page parsing and deal logic, and need no network:

python -m pytest -q tests

Output files

File	Purpose
deal_report.md	Human-readable report from the latest run
current_deals.json	Latest deals plus which ones are new; used by the workflow
dashboard.json	Compact feed for a website (see below)
price_history.json	Every priced listing from every run, append-only

dashboard.json

About 50 KB, regenerated every run, fetchable without auth from

https://raw.githubusercontent.com/Nalomun/MonsterTracker/main/dashboard.json

{
  "generated": "2026-09-28T04:44:00+00:00",  // UTC
  "ok": true,                                 // false if the scrape was blocked
  "threshold": 0.12,
  "reliable_since": "2026-09-28",             // earlier history used the old parser
  "listings_checked": 73,
  "deals": [ { "asin", "title", "price", "fl_oz", "price_per_oz", "seller_info",
               "availability", "offer_type", "sns_price", "verified", "link" } ],
  "best":  [ /* same shape; 10 cheapest listings this run */ ],
  "daily": [ { "date", "best_price_per_oz", "best_asin", "best_title",
               "deals", "listings", "reliable" } ],                 // up to 120 days
  "products": [ { "asin", "title", "link", "fl_oz", "latest_price",
                  "latest_price_per_oz", "latest_seen", "min_price_per_oz",
                  "series": [["2026-09-01", 0.1124], ...] } ]      // 15 most-seen, 90 days
}

Don't point a website at price_history.json; it grows by roughly 40 KB per run.

Known limitations

• Amazon's HTML changes. The parsers live in parse_search_cards and parse_product_page in tracker.py; run with DEBUG_HTML=1 to capture what was served when something breaks.
• Search results vary by session and region, so the set of listings differs a little from run to run.
• Listings with no featured offer are priced from the cheapest third-party offer and labelled as such. Out-of-stock listings are recorded but never alerted on.
• History before 2026-09-28 was collected by an earlier version with a buggy price parser and should not be trusted; dashboard.json flags those days.

License

MIT