Methodology Chapter 02 of 08
Why this Story appears in your feed.
Five signals score every article and Story, you decide how much each signal counts, and your own preferences adjust the result. Any card in the app will show you its own math.
Every card explains itself.
The homepage protagonist article opens its relevance explanation and shows the visible topic, coverage, and publisher controls behind its feed placement.
Trump Asks Congress for $88 Billion, Mostly for War With Iran
You can change the ranking rules.
The Feed Formula shows the complete signal allocation and demonstrates how changing Publisher Reach changes the visible allocation.
What the score is made of
Five signals produce a score that ranks each article and story in your feed. There is no personalization or behavioral tracking that guides how each is ranked. Each asks a different question, and each answer is a number from 0 to 10. Open any of them to read the table that decides that number, exactly as the app prints it.
- Source Credibility How trustworthy is the source?
-
Media Bias/Fact Check's factual-reporting rating for the publisher.
How it's scored
SCORING RUBRIC (0–10)
RATING WHAT IT MEANS SCORE
Consistently accurate, minimal errors 10
Generally reliable reporting 8
Accurate with some sensationalism 5
Mix of factual and inaccurate content 2The rating comes from Media Bias/Fact Check, an independent organization that reviews newsrooms against their accuracy record. It describes the publisher rather than the article, so everything from one newsroom starts on the same footing, and Mosaic News does not re-rate anyone. It reads the label MBFC published and looks up the matching number. The card lists the top four bands; the one it leaves off is a Low factual rating, which scores 0.
- Trending Speed How fast is this story growing?
-
How much new coverage a Story is drawing right now, measured against every other Story being tracked today.
How it's scored
SCORING RUBRIC (0–10)
RATING WHAT IT MEANS SCORE
Among the fastest-growing stories right now 9–10
Growing faster than most of today's stories 7–9
Picking up new coverage at a steady pace 4–7
Little or no new coverage right now <4Each article in a Story carries a weight that halves every six hours, so coverage from two hours ago counts for far more than coverage from yesterday. Repeat filings from one newsroom are damped: an outlet's second article on a Story counts half as much as its first, its third a third as much, so a single busy newsroom cannot make a Story look like it is spreading. Those weights add into one number, and that number is ranked against every other active Story. The score is the position in that ranking, which is why the bands read “most of today's stories” instead of naming a fixed count.
- Coverage Breadth How many outlets are covering this?
-
How many independent newsrooms are on the Story, also measured against every other Story today. A wire report reprinted across many sites counts once.
How it's scored
SCORING RUBRIC (0–10)
RATING WHAT IT MEANS SCORE
Among the most widely covered stories right now 9–10
More outlets than most of today's stories 6–9
Mid-pack coverage so far 3–6
Fewer outlets than most stories <3This counts newsrooms rather than articles. Ten pieces from one outlet count once, and a wire report reprinted across a chain counts once for the whole chain. That count is then ranked against every other active Story, the same way Trending Speed is. Ranking rather than fixing a threshold is what keeps the bands honest as sources are added: a standing rule that twenty outlets means wide coverage would quietly stop being true.
- Freshness How recently was this published?
-
The article's own publication time. For a Story, the time of its most recent coverage.
How it's scored
SCORING RUBRIC (0–10)
RATING WHAT IT MEANS SCORE
Published within the last hour 10
Published 1–4 hours ago 8
Published 4–12 hours ago 5
Published over 12 hours ago 2The one signal computed from a single timestamp. It starts at 10 the moment a piece publishes and halves every six hours, so it reads about 5 at six hours old, 2.5 at twelve, and 0.6 by the end of the first day. Nothing is ever cut off. An older article keeps a small score and can still outrank a fresh one when the other four signals are strong enough.
- Publisher Reach How widely read is the source?
-
The publisher's rank among the world's most-visited websites.
How it's scored
SCORING RUBRIC (0–10)
RATING WHAT IT MEANS SCORE
The web's biggest newsrooms 8.5–10
Large national audiences 6–8.5
Solid but smaller audiences 3–6
Small or specialist audiences <3The rank is the publisher's position among the world's most-visited websites, taken from the Tranco list, which averages several independent traffic measurements over thirty days. The score follows that rank on a logarithmic scale, so every tenfold drop costs about the same 3.2 points: a publisher ranked near 10,000th scores about 6.5, one near 100,000th about 3.2. Everything inside the top 800 scores a flat 10, so the household names sit level with each other and the other four signals decide between them.
Not every publisher carries a rating, so the formula has to decide what an absence means, and it answers differently for the two signals that depend on one. An unrated publisher scores 3 out of 10 for credibility rather than 0, because not knowing whether a newsroom is careful is different from knowing that it is not. Reach works the other way. Almost every publisher in the feed has a traffic rank, so not having one is itself evidence of a small audience, and it scores 2.
How five numbers become one
The five scores are not averaged. Each is capped by the weight you gave it, and those weights always total 100, so raising one signal lowers the others to make room.
Each signal is graded
Every signal scores the article or Story in front of it from 0 to 10, following the rubric printed in its own card.
Your weights set the ceilings
The five weights always total 100. A signal you weight at 30 can contribute at most 30 points, and one you set to zero contributes nothing at all.
Each earns its share
A signal scoring 7 out of 10 earns 70 percent of its ceiling. The five contributions add together into the base score, before any preference of yours applies.
What you can change
The weights decide how the feed ranks in general. Preferences say something narrower: more of this, less of this, or none of this. Every one of them uses the same five levels. Favorite adds 20 points, Like adds 10, Neutral changes nothing, Dislike subtracts 10, and Blocked takes the item out.
Topics
A favored topic lifts anything matching it, a disliked topic pushes it down, and a blocked topic removes it when the match is confident. Because a secondary topic match is less certain, blocking one pushes the item far down rather than deleting it.
Publishers
Publisher preferences act on individual articles. Favoring an outlet lifts its own reporting, and blocking one removes its articles, but the Story other outlets are also covering stays where it is.
Article types
You can set a level for news, analysis, opinion, editorials, features, and reviews. Stories are exempt, because one Story can hold several types at once.
The boundary
What this score does not mean
- A high score is not a claim that an article is true. It means the article scored well under the formula you can read.
- Political lean is used for perspective and coverage comparison. It is never a ranking reward.
- Two of the five signals are relative, so a Story's score can ease off when other stories heat up, without anything about that Story changing.
- Preferences are not a behavior profile. They are the settings you chose, sent with the request and not kept afterward.
- Blocking a publisher removes its articles. It does not erase a Story that other outlets are also covering.