Research Guides
Name the Resource Size Before Reporting a Page-Weight Change
One website audit calls a resource 30,000 bytes; another calls it 120,000 bytes. The difference may be a measurement-definition mismatch rather than a broken export. Before reporting that a page became four times heavier, identify which size each tool recorded.
This worksheet focuses on browser Resource Timing measurements. It is a manual research method, not a Colquhoun.shop performance-monitoring feature or a claim about this site's current speed.
1. Preserve the exact field name
MDN describes transferSize as an encoded response body size plus a fixed allowance for response headers. Its documented values also distinguish revalidation and local-cache cases; a zero can additionally reflect a cross-origin timing restriction. MDN: transferSize
By contrast, decodedBodySize describes the response body after content encoding, such as gzip or Brotli, has been removed. It is not the same quantity as the transferred representation. MDN: decodedBodySize
Keep those field names in the working file instead of renaming both columns “size.” When summarizing for readers, explain the chosen measure in plain language and retain the technical definition in the method note.
2. Compare like with like
In an invented review, an earlier export lists 120,000 decoded body bytes for a resource. A later report lists 30,000 encoded body bytes. Dividing one by the other does not establish a 75 percent reduction caused by a deployment. The observations describe different stages of representation.
The reviewer retrieves comparable fields for both runs, if available. If they are unavailable, the conclusion remains unresolved. A plausible improvement story is not a reason to manufacture the missing measurement.
3. Record how the page was exercised
Document the browser, page address, run time and cache conditions. Describe the interaction sequence too: initial load, scrolling, opening a panel or playing media can define different review scopes.
Use the same agreed sequence for a before-and-after comparison. Keep repeat-visit observations separate from the first-visit test rather than averaging them together without explanation. If a resource appears only after an interaction, say whether that interaction was included.
Do not treat a Resource Timing sum as a complete accounting of every network byte or the browser's memory usage. Label it as a sum of the selected fields for the captured entries, with the capture boundary made explicit.
4. Investigate zero before calling it savings
A zero transferSize needs context. Check whether the record has a plausible cache explanation or a timing-visibility limitation. Keep “zero observed in this field” separate from “resource absent” and “measurement unavailable.”
Do not change cross-origin permissions merely to make a report's totals look complete. Record the limitation and ask the responsible implementer to assess any proposed instrumentation change within its security and privacy scope.
5. Copy this resource-size comparison card
- Question: what change the measurement should help investigate.
- Field: exact API or tool metric and its definition.
- Units: raw bytes and any display conversion.
- Run: browser, time, page and interaction sequence.
- Conditions: cache state and known visibility limits.
- Population: captured resource entries and exclusions.
- Result: comparable observations and unresolved differences.
Save the original observations before calculating summaries. A colleague should be able to identify which resources contributed to the total and why a particular row was excluded.
The useful conclusion is a defined, reproducible size comparison. It does not independently prove faster rendering, better user experience or improved search rankings; those require their own appropriate evidence.