Today, DealMonitor turns one year old. On March 21, 2025, the idea was born: a price tracker that works on any website — not just Amazon or the usual suspects, but everywhere. From niche pharmacies to private islands.
The Beginning
The first version was simple: enter a URL, an ML model detects the price on the page, and when it drops, you get an email. No manual CSS selector configuration, no copy-pasting prices. Just the URL and you're done.
What sounds simple was technically challenging. Every online shop has a different structure. Some render prices via JavaScript, others hide them in nested HTML, others load them via API. Our approach: a machine learning model that extracts the correct price from a page's HTML — regardless of how the shop is built.
What Happened in 12 Months
One year later, DealMonitor looks very different from the beginning:
- 145 users tracking prices across 146 different shops
- 380 active trackers monitoring products around the clock
- AI price detection v4 with an ensemble model — trained on real user data
- Chrome extension for one-click tracker creation right on the shop page
- 3 notification channels: Email, Telegram, and Web Push
- 9 languages for an international user base
- API integrations with AliExpress, Etsy, IsThereAnyDeal, and Shopify for instant price queries without scraping
- Dark mode, tracker groups, drag-and-drop, price history charts, and CSV export
The Biggest Challenges
If we learned one thing: reliably extracting prices is incredibly hard. Every shop is different, and the variance is enormous.
Amazon alone has dozens of price elements on a page — marketplace offers, strikethrough prices, unit prices per 100g, variant prices. Our model had to learn to find the one relevant price.
Then there's bot protection: Cloudflare, Akamai, DataDome. Shops invest millions to block automated access. We learned there's no silver bullet — sometimes you need HTTP, sometimes Selenium, sometimes an external API, and sometimes a specialized scraping service.
And of course variants: one product, five colors, three sizes, different prices. The URL is the same, but the price changes depending on the selection. This remains one of our biggest challenges to this day.
What We Got Right
The most important decision was to work on any website rather than just a handful of shops. This makes us more flexible than competitors and gives our users real freedom. Whether it's a bicycle on bike-discount.de, heating pellets on heizpellets24.de, or a private island on vladi-private-islands.de — DealMonitor tracks it.
Second: the ML model learns from every user. Every confirmed price improves detection for everyone. It's a flywheel effect: more users → more training data → better detection → more satisfied users.
What's Next
We have ambitious plans for year two:
- Wishlist import — import Steam wishlists and Amazon wish lists with one click
- Better variant detection — reliably distinguish every product variant
- Price forecasting — predict the best time to buy using historical data
- Mobile app — price alerts on the go
- Edge extension — not just for Chrome
Thank You
A huge thank you to all 145 users who have tried DealMonitor. You're the reason we get better every day. Every tracker you create makes the model smarter. Every piece of feedback shows us where to improve.
Here's to another year of deals! 🎂
