Methodology
Methodology
FinQuarry reads headlines and short feed descriptions. It never reads full articles. It writes its own headlines and summaries, then links to the original reporting. A model estimates three scores from 0 to 100. These are editorial ratings, not facts or predictions. They are the same for every reader.
Truth estimates how well a story's claims are supported. Its current breakdown is 47% reporting quality, 31% outlet reliability and 22% confirmation from independent outlets or direct financial data. Primary sources can add a bonus. A score is not a fact-check.
Importance estimates how much a story could matter. It weighs the model's impact rating at 62%, breadth at 13% and independent coverage at 25%. Opinion, promotional and personal-finance stories receive a discount.
Actionability is a general editorial estimate of how much a story could matter to markets and how time-sensitive it is. Its displayed breakdown is 55% expected impact, 20% direction clarity, 15% urgency and 10% how much may already be priced in. The score is also limited by the truth estimate. It is not a trading instruction and is not personalised.
Outlet reliability begins with an editorial baseline. Automated assessments compare reporting quality and later corroboration with other covered outlets. The estimate moves toward the evidence as more stories accumulate. Small samples are uncertain; fewer than 20 assessed stories show Insufficient data.
Most important and Most actionable list a story for 28 hours after it was first reported, however many follow-up reports it gets. On Saturdays and Sundays (New York time), when news is light, that is 48 hours. Breaking lists stories first reported in the last 8 hours, and with a free account you can change that window in Your view. This week shows each day's five most significant stories and keeps them for seven days. Where a Past month tab is shown, it keeps them for 30 days.
The public snapshot updates about every five minutes while the publisher runs. Feeds can be late, incomplete or unavailable. A headline or short description may omit context. The model can misunderstand a claim, miss a correction or overstate confidence. Read the linked reporting before relying on a score.