How we collect, and what we estimate.
We use clean sessions. We use a fixed location. We make one record for each product in each period. We read the measured fields from the marketplace. We produce each estimated field with one named method.
Access
We reach each marketplace on its own terms. We use a public interface when one exists. We use the storefront when the interface is closed. We use a home-network exit when the limit is the network, not the client.
- 01Use the public interface when the marketplace gives one. This surface is the most stable.
- 02Use the storefront data route when the interface is closed to third parties.
- 03Match the session and the browser when a marketplace needs it.
- 04Use a home-network exit when the limit is the network, not the client.
The demand model
We estimate units from review speed when the marketplace does not publish sales. We use a review rate of 2 to 4 percent. We calibrate the rate against client sales. We measure growth as the change between snapshots. We do not add past history.
Record rate
We take a snapshot each day at a fixed hour. We add a weekly summary. The rank, the review count, and the price then stay comparable from period to period.
Limits we state first
- —No party can add past demand history. A record starts on the first day of collection.
- —The full category sale value is public on one marketplace. For the others, we estimate it until a partner login exists.
- —Some sale data needs an account. This is an access agreement, not a collection block.