Skip to main content
Research and methods

Paid-media methodology

Measured activity supports a campaign hypothesis. It does not predict advertising performance.

What is measured

Snapshot rates describe source events over a stated time window. Provider, timestamp, sample count and units travel with the observation. They are not counts of unique people, reach or conversions.

From signal to controlled test

Review freshness and relevance, inspect the source, formulate one audience or creative hypothesis, preview the platform payload, and compare a bounded test with your current control. Campaign changes remain a separate authorized action.

Evidence completeness

Recommendation confidence uses evidence_completeness_v1. It measures available evidence fields and freshness, not a probability of success, calibrated certainty or expected uplift.

What the literature supports

Chae and colleagues report conditional results for real-time marketing. Their study does not establish that posting at the right time alone increases engagement, and it does not evaluate TrendsAGI.

Read the cited study →