October 21, 2026
Narratives retirement complete
The narratives retirement announced on October 5 is now complete. Use the upgraded Landscapes product instead.October 5, 2026
Narratives retire on October 21, 2026
We’re retiring the narrative capabilities in favor of the upgraded Landscapes product. On Wednesday, October 21, 2026 your assistant will no longer be able to search narratives, chart a narrative’s volume over time, or pull a narrative’s mentions.September 21, 2026
Geography as a first-class capability
Your assistant can now reason about where coverage is happening using PeakMetrics’ normalized geography: region, subregion, country, state, county, and city. Ask for a place by name (“Ohio”, “the Bay Area”, “West Africa”) and the assistant scopes questions to it without you handling any codes or IDs. Geography composes with everything else the assistant can do:- Any question can be narrowed to a place. Narratives, mentions, top accounts, and top journalists all accept a location filter. Picking a country automatically includes every state, county, and city inside it, so a “coverage in the US” question doesn’t require enumerating states.
- Analytics can be broken down by geography. Workspace analytics can report volume by country, or drop to state, county, or city when you need finer resolution, to answer “where is this story loudest?” directly in conversation.
- Individual mentions carry their location. Quotes and links can be grounded in a specific place at every level the mention resolved to, with coordinates for city-level results.
May 8, 2026
Randomized mention sampling
Mentions search now supports randomized sampling, so your assistant can pull a representative read of a large mention pool instead of only the top-ranked slice. All standard workspace filters still apply, and samples are drawn from across the full time range you specified. Use this when you want an unbiased view of what’s being said rather than the loudest or most recent voices. Typically useful for:- Qualitative research and briefings, where cherry-picked top results would skew the narrative and you need a cross-section of how people are actually talking about a topic.
- Sentiment and theme spot-checks on high-volume workspaces, where reading the top 50 mentions overweights a handful of viral posts.
- Auditing coverage for balance across authors, domains, or regions before quoting mentions in a report.
- Sampling for downstream analysis (tagging, clustering, LLM summarization) where a random sample generalizes better than a ranked one.