Summary
In upstream mode, totals measure bytes on the socket to the provider, TLS included. Direct sizing measures the destination connection and estimates an added CONNECT exchange. Read each figure’s label; none is a bill.
Hosts
Bytes per host with the CONNECT exchange: measured in upstream mode, estimated in direct sizing. The chart shows the twelve heaviest hosts; the rest are grouped. Hostnames are shown exactly as the report stored them.
Resource types
Allocated: reported bytes plus a share of tunnel overhead. Each host's tunnel bytes are shared over its reported requests in proportion to their sizes, so per-type figures add up to the host totals but the split is an allocation, not a measurement. Bytes no reported request explains appear as the type unreported.
Attribution buckets
Where the with-CONNECT bytes went. Tunnels of hosts with no matching request are labelled by rule: a catalogued background component, a tunnel opened before the helper attached, an idle preconnect, or unattributed. The with-CONNECT total includes an estimated exchange in direct sizing; the bucket is a label.
What-if and generated fixes
What-if figures are modelled, not measured: they read the allocated bytes of the types or hosts they would remove and never count unreported bytes. A second, compared run is the measurement. Fix code is shown as stored in the file, as plain text; review it before use.
Find results
Which response of one page load already carried the searched values, ranked by the bytes a standalone request of it would move (billed basis: body, headers and one TLS handshake estimate). The searched values are never written to the report, so this viewer neither shows nor asks for them.
Cost at your rate
Estimated billable transfer at your rate; providers meter differently — check your provider's billing definition. scrapescope has no default rate and this page suggests none: type the price your own provider charges.
Per GB of 10^9 bytes, or per GiB of 2^30 bytes when the report was written with --gib. Leave empty for no estimate.
Run comparison: A against B
Differences are B minus A. Two runs of the same job can differ by timing alone (an extra idle connection, a component download that came or did not), so read small deltas as noise and large ones as a lead to check with a third run.