Browse documentation
Metrics and dimensions
Metrics overview
Metrics are computed at platform and tracker-day levels from completed runs. Read them as observed results for your configured prompt set—not as a universal measure of all AI search activity.
Use the right denominator and reporting scope for every Digraph metric.
Core measurement rules
Always record the tracker, date range, platform selection, and prompt count before comparing a metric. A change in configuration can change a metric even when AI responses have not materially changed.
- Compare equivalent prompt sets and platform scope.
- Distinguish completed runs from scheduled or missing runs.
- Use answer-level evidence to explain a daily change.
- Treat missing values as unavailable data, not as zero.
Metric families
| Family | What it describes | Useful dimensions |
|---|---|---|
| Coverage | How often the configured run set produced a tracked brand or entity result. | Date, platform, prompt |
| Recommendation | How often the tracked brand was classified as recommended. | Date, platform, prompt, competitor |
| Narrative | Sentiment derived from eligible mention records. | Date, platform, prompt |
| Competitive | Observed share within the tracked competitor set. | Date, platform, competitor |
| Evidence | Citations and source occurrence in tracked answers. | Date, platform, source domain or URL |
Read the definitions
The detailed definitions page documents the available calculation contracts. If an interface labels or rounds a value differently, the data contract and selected scope remain the authoritative interpretation.
Was this page helpful?