Separate help searches with no results from searches with no click

Classify one search event at a time

This original six-event log stores result count and whether a click was observed. Zero results is one class. With results, clicked=false is no-click and clicked=null is unknown click evidence. The no-click fraction uses every result-bearing event as its stated denominator; unknown clicks remain separately reported. This is a sample log, not Asteratek analytics.

Keep both denominators beside the counts

Zero-result searches are 2/6. Three of the four result-bearing searches have no click: 3/4. There is one clicked search. None of these counts establishes that an answer worked, why a reader left or whether the search came from Google.

1. Count unique recorded events
2. Separate zero-result events
3. Classify clicks only with evidence
4. Report unknowns and denominators

Unknown results or click evidence is not silently counted as success or failure.

Prepare a search-gap worksheet

The complete JSON material includes the baseline and changed-click case. Use the table/Markdown sheet to annotate a later observed export, retaining source window, event grain and unavailable fields. Do not paste customer query text into a public sample.

Results must be a whole count 0–1,000 or null; click evidence is true, false or null. A clicked zero-result event contradicts this contract and is rejected. A zero result-bearing denominator has no percentage. Repeated IDs or more than 30 events are rejected.

A zero-result gap may need a reader-permission check before new content is commissioned. Use this worksheet to prepare a search-analysis brief; measure actual task outcomes separately.