What does the hourly analysis do?

When a measurement is published, not everything is ready at once. Some pages wait for the answers to be classified separately. That work is done by an analysis that runs every hour.

What does it do?

At the top of every hour, the system finds the oldest published measurement whose themes have not yet been extracted and processes its answers. Each run processes a single measurement; if no measurement is waiting in the queue, nothing happens.

What gets processed are answers that are already stored. The analysis does not ask ChatGPT or the other engines again. It reads the stored answers with a language model and extracts which themes and which brands appear together in each answer. The spend of every run is capped at a fixed ceiling.

What changes on screen?

Once theme extraction is finished, these pages fill in:

• Competitor Intelligence (Rakip Zekâsı) and the cards about rivals,
• Opportunities (Fırsatlar),
• the persona screens.

If extraction has not yet run for that measurement, these pages do not stay blank; they say "theme extraction has not run yet" ("tema çıkarımı henüz çalıştırılmadı"). This is not a result, it is pending work. It usually completes within the next hour.

Other jobs that run in the background

Two other jobs should not be confused with the hourly analysis:

• Retry. Questions for which an engine returned no answer because of a temporary error are asked again every half hour. If the retry succeeds, that answer stops being "could not be measured". We explain why this matters in The difference between not measured and not mentioned.
• Automatic tracking. If weekly or monthly tracking is turned on in the brand settings, the new scan starts on Monday morning. This is a paid scan that asks the engines new questions, and it is separate from the hourly analysis.