Pricing
Amazon Price Tracker
Amazon Price Intelligence Dashboard
Does where you live change what you pay on Amazon? Tracks 500 Amazon best-sellers across 10 US zip codes with hourly price snapshots for 4 days — 479,231 total observations. Answers two questions: does Amazon price dynamically by geography, and how much does any given product’s price move in 96 hours?
Inputs
Outputs
What you get after a full run.
- Price history for 500 products across 10 zip codes and 96 hourly snapshots
- 163 products that moved in price during the 4-day window (33% of catalog)
- Biggest single price swing: $90.00 — Beats Solo 4, +82%
- Geo pricing analysis: 97% of products priced identically nationwide; 12 exceptions identified
- 11 products caught running live title A/B tests with measurable BSR impact
- 81 products with MSRP changes during the collection window
- 2,150 multi-signal cascade events — price, MSRP, and Subscribe & Save changing in the same run
- Full product detail page for every ASIN: price timeline, zip bar chart, BSR trend, availability strip, MSRP history
How it works
A 6-phase pipeline. Read the blog here for a deeper explanation.
- Discover amazon_best_sellers agent identifies 500 top-selling ASINs across 10 product categories: Electronics, Grocery, Health & Vitamins, Beauty, Toys, Pet Supplies, Sports, Tools, Kitchen, and Office Supplies. Each category query returns a ranked list of best sellers with ASIN, title, BSR rank, and price. The top 50 ASINs per category are selected, giving a catalog weighted toward high-volume, frequently repriced products.
- Configure 5,000 (ASIN × zip code) input pairs uploaded to a Nimble Managed Agent Job with an hourly cron schedule. 10 zip codes: 5 major cities, 5 rural towns. The job is configured once and runs autonomously — no per-run orchestration code required. Zip codes were chosen to span a wide geographic and demographic range: New York, Los Angeles, Chicago, Houston, and Seattle for major cities; rural towns in Montana, Mississippi, Vermont, New Mexico, and Wyoming.
- Collect amazon_pdp runs every hour for 4 days. 480,000 total agent calls. 60+ fields captured per product per run. Each agent call returns the full product detail page as structured data — including fields Amazon only shows in certain contexts, like Subscribe & Save eligibility and dynamic badge text. The managed job handles retries automatically, achieving a 99.2% success rate across the full collection window.
- Process Raw Parquet data decoded into price history, per-ASIN price swings, zip code medians, BSR trends, availability timelines, title change logs, and multi-signal event records. Processing runs in phases: first a price normalization pass to strip currency formatting and handle null prices from out-of-stock runs, then a per-ASIN aggregation to compute swing magnitude and first/last observed timestamps. The final step joins all signal streams to identify runs where price, MSRP, and Subscribe & Save changed simultaneously.
- Surface 8 data stories extracted: algorithmic repricing, geo pricing exceptions, live title A/B tests, BSR demand patterns, out-of-stock gaps, MSRP shifts, Subscribe & Save as an inventory proxy, and multi-signal cascade events. Each signal type has a dedicated detection rule: geo exceptions require a price deviation of >5% from the national median for at least 3 consecutive runs; title A/B tests require the same ASIN to show two distinct titles across zip codes in the same run. Signal records include the full observation context — timestamps, zip codes, and raw field values — for drill-down in the dashboard.
- Dashboard 4-page Streamlit app with every data point linked to a product detail page showing price timeline, zip comparison, BSR trend, availability strip, and MSRP history. The product detail page is the core navigation unit — every chart, table, and signal card links to it. It shows the price timeline across all 10 zip codes and 96 runs, a zip comparison bar chart, BSR trend with volume estimate, availability strip showing every out-of-stock window, and MSRP history with change markers.
Stack
Nimble primitives plus the full runtime stack.
APIS & AGENTS
What it does
- amazon_best_sellers Discovers top-selling ASINs by category and page — used to build the 500-ASIN tracking list across 10 categories.
- amazon_pdp Full product detail page — 60+ fields per product including web price, MSRP, availability, BSR rank, and Subscribe & Save. Runs as a Managed Agent Job at 5,000 calls/hour.
RUNTIME STACK
Role
- Nimble Agent Jobs Managed job runner — handles scheduling, retries, and concurrency for 480,000 agent calls over 96 hours.
- streamlit 4-page dashboard with URL-based routing and product detail drill-down for every ASIN.
- plotly Dumbbell charts, price timelines, availability heatmaps, BSR trend lines, and scatter charts.
- pandas + parquet Processing 4.6 GB raw data into approximately 60 MB of structured Parquet files across 8 datasets.
- python 3.9+ Phase scripts for ASIN collection, data processing, and signal extraction
- MIT license Fork, modify, ship — no restrictions.
Reach out if you have any questions.
Talk to an Expert