Proxies for price tracking
Collecting prices once is easy. Collecting them on a schedule so the numbers stay comparable is where the work is.
Frequency follows the category
Fast-moving categories change through the day; slow ones barely move in a week. One schedule for everything either wastes requests or misses changes.
Split the catalogue by how often it actually moves and give each part its own cadence.
Size on the peak
When important items are checked more often, the request rate spikes in those windows. The pool has to cover the spike.
An average figure leaves you short at exactly the moment the data matters most.
Keep the series comparable
Price often depends on the store or region attached to the session. If that attachment drifts between runs, your series records your own configuration changes.
Fix the location, fix the currency, and record both alongside every value.
Verify by hand every week
Take a hundred random items and check them manually. Gaps in monitoring form quietly and show up only in comparison.
store attached fixed per run currency fixed per run timestamp recorded with every value