Results and outlooks are read out of earnings releases, call transcripts and filings within a minute of the print. Prior and new ranges are kept in the company’s own words, then differenced independently — so the label and the arithmetic can be checked against each other on every row.
{
"revision_id": "SN:guidance:20260910",
"ticker": "SN",
"company": "SharkNinja",
"metric": "Revenue",
"metric_family": "headline_financial",
"period": "FY26",
"direction": "raise",
"prior_guidance_text": "11.5% to 12.5%",
"new_guidance_text": "16% and 17%",
"prior_midpoint": "12.0%",
"new_midpoint": "16.5%",
"midpoint_change_pct": 37.5,
"range_relation": "clears_prior",
"words_numbers_agree": true,
"source_type": "press_release",
"released_at": "2026-09-10T11:02:14Z",
"published_latency": "38s",
"llm_model": "gemini-2.5-flash-lite",
"prompt_version": "guidance-v3"
}A sample of the schema that lets you filter on company size, metric family, period, direction of the change, the range relation, and how far the midpoint travelled — with prior and new figures quoted verbatim, so a screen result is readable without opening the filing.
Because every revision is stored as a differenced pair, the whole season aggregates. Distribution of the moves, where they came from, and which line companies actually touched.
Each bar is the change in the midpoint of the guided range. Around a third of all moves are under 2% — a company nudging the middle of its range rather than repricing the year.