Skip to main content
Build a repeatable week-over-week report from two complete date windows that this guide defines in UTC. This guide uses the reusable client from the Quick start.

Define comparable windows

For this report, construct inclusive UTC calendar-day windows. End the current window yesterday, then subtract six days. End the previous window one day before the current window begins.
This avoids a one-day overlap and excludes the partial current UTC day.

Compare score points

Mention Rate, Source Rate, Coverage, and Share of Voice are percentage values. Sentiment is a score from 0 to 100. Report an absolute point difference for all five rather than a relative percentage change, which becomes misleading near zero and is not meaningful for sentiment.

Synthetic output

These labels describe the selected API windows only. They do not establish business impact or cause.

Add diagnostics carefully

  • Use List answers to inspect response-level changes. The endpoint does not apply hasSelfSource=false; filter the returned flag in your client and paginate before calculating a total.
  • Use List query fan-out records for aggregated query text. Each record exposes prompts[] and models[], not singular prompt or provider fields.
  • Keep citation, mention, query fan-out, and competitor movements as separate observations.

Validate the comparison

  1. Confirm the windows contain seven days each and do not overlap.
  2. Compare methodology.promptsCount, responsesTotal, and providers between periods.
  3. Explain prompt, provider, topic, or tag changes before interpreting a score delta.
  4. Label every value with its REST performance formula.
  5. Send the report only to destinations approved for the data it contains.