AI Price Drop Tracker — Monitor Any Product Price Across Multiple Sites and Get Notified the Moment It Drops Below Your Target
Built by SynapseHelix ·
About this build
A Python and Claude powered workflow that monitors product prices across AliExpress, Amazon and competitor websites on a daily schedule, compares them against your target price threshold, detects genuine drops versus temporary sale badges, and sends a Slack or email alert the moment a real price drop is confirmed — so
What was built
Built this after manually checking a product across 6 different supplier websites for 3 weeks straight before finally catching a price drop on a Friday afternoon. By the time I checked it again on Monday the sale was over. The whole thing felt completely solvable with a simple automated workflow. What started as a simple price checker turned into something more useful once Claude got involved in the analysis step. How the workflow runs: Step 1 — Product list setup in Airtable A simple Airtable base serves as the product tracking list. Each row contains the product name, the URLs to monitor across multiple sites, the target price threshold in AUD, the current known price and the date last checked. Adding a new product to track takes about 30 seconds — just add a row. Step 2 — Daily price scraping via Python Every morning at 7am a Python script runs through every product in the Airtable list. For each product it visits each URL and extracts the current displayed price. The script handles the 3 most common price display patterns — standard price text, strikethrough sale pricing and bundled quantity pricing where the per-unit cost requires calculation. The scraper runs with a randomised delay between requests and rotates user agent strings to avoid being blocked. For sites with aggressive bot protection a fallback uses Perplexity to search for the current price rather than scraping directly. Step 3 — Price analysis via Claude This is where the build gets more interesting than a standard price tracker. Raw price data alone is not enough to determine whether an alert is worth sending. Claude analyses each price point across 3 dimensions. Genuine drop versus manufactured urgency — is this a real price reduction or a site that has artificially inflated the original price to make a discount look bigger. Claude checks whether the previous prices in the Airtable history support the claimed original price or whether the original price only appeared recently. Trend direction — is this the lowest price seen in the tracking period or is it a small fluctuation in an otherwise stable price. A 2 percent drop on a product that fluctuates 5 percent daily is noise. A 15 percent drop on a product that has been stable for 6 weeks is a genuine signal. Alert priority — should this trigger an immediate alert, a daily summary mention or no notification at all. Not every price drop is urgent. Claude assigns a priority of High, Medium or Low based on the drop size relative to the target threshold and the historical price pattern. Step 4 — Price history logging to Airtable Every daily price check gets logged to a second Airtable table as a timestamped record. This builds a price history over time that makes the Claude analysis more accurate — after 2 to 3 weeks of daily logging the system knows what normal price fluctuation looks like for each product and can distinguish genuine drops from noise much more reliably. Step 5 — Tiered alert delivery High priority alerts — a product drops below the target threshold for the first time or drops more than 10 percent in a single day — trigger an immediate Slack message with the product name, previous price, new price, percentage drop, the URL and a one sentence Claude note on why this was flagged as high priority. Medium priority mentions get batched into a daily digest sent at 8am covering all products that had notable price movement in the last 24 hours but did not cross the high priority threshold. Low priority fluctuations get logged to Airtable only with no notification. What I built this for originally: Dropshipping product sourcing. Monitoring 15 AliExpress products daily to catch temporary flash sale pricing before it expired. After 3 weeks of running it caught 2 genuine price drops that I would have missed on manual checking — one a 34 percent drop on a high volume product that lasted 4 days. Secondary use case that emerged: Competitor price monitoring. Added competitor product pages to the tracking list and now get daily alerts when a competitor changes their pricing. For a small e-commerce business this is genuinely valuable intelligence that used to require manual checking or expensive tools. What broke during development: The fake sale detection was the hardest part to get right. Many AliExpress listings show a crossed-out original price that was never actually the real price — it is set artificially high to make the discount look bigger. Early versions were triggering alerts on these fake drops constantly. Fixed by requiring the original price to appear in at least 3 consecutive daily logs before Claude treats it as a genuine baseline. Dynamic JavaScript rendered prices did not work with basic Python requests. Fixed by adding a Playwright option for sites that require JavaScript execution — slightly slower but catches prices that basic scraping misses entirely. Results after 3 weeks: 15 products monitored daily across 6 websites. 2 genuine price drop alerts that led to action. 7 false positive alerts eliminated by the Claude analysis step that would have triggered under a simple threshold check. Zero missed alerts on products that genuinely dropped. What I would build next: A price prediction layer that uses the historical data to estimate when a product is likely to hit its lowest price — some products cycle on predictable schedules around pay periods or end of month. Also want to add a buy now button directly in the Slack alert that opens the product URL in one click.