Size from volume and window. Thread count follows afterwards. The formula holds on any platform, only the safe rate changes.
required rate = pages / window length in minutes minimum = required rate / safe rate per address working number = minimum x 3
Run 200 requests from one address, raise the speed in steps until the first refusal, then take roughly 60 percent of that figure. The measurement happens once per platform and lives in a table afterwards.
| Section of a platform | Typical safe rate |
|---|---|
| product page | about 60 requests per minute per address |
| search and filters | about 30 |
| reviews and ratings | about 15 |
| live price and stock | about 8 |
pages 3000 window 180 minutes required rate 17 requests per minute safe rate 10 per address minimum 2 addresses working number 6 addresses
The triple reserve covers retries, addresses resting after a burst of failures, and one address held aside for manual verification. Sizing to the exact minimum breaks on the first heavier day.
The arithmetic stays predictable only on private addresses used for data collection. On a shared pool your rate adds to somebody else's and the formula stops forecasting anything.