The Weightedly Eight
the eight signals · every thread · read the same way
This is the standard Weightedly reads by. Every thread the desk considers is read against eight signals, and here they are in full — their names, their ranges, and the plain question each one descends from — set down so that anyone can see what the desk is looking for and hold it to exactly that.
None of these questions are new. They are the questions a working desk editor has always asked of a story before spending a reader’s attention on it: How do we know this? What is actually being said? Who is telling us? What Weightedly has done is not invent them but make them mechanical — eight separate readings, each with a fixed range, applied the same way to every thread and open to inspection.
We make no claim to that tradition’s authority beyond having inherited it. The integrity in these questions belongs to the editors who asked them long before this tool existed; ours is only the work of writing them down precisely and applying them without exception. That is the whole of what is published here — the questions, made answerable.
- 01
factual anchor
range 0–5“How do we know this?”
Does the thread rest on anything a reader could check — specific numbers, named tools, dated events, steps someone else could repeat? This is the difference between an account and an impression. A thread that only gestures at a feeling scores low here, however strongly the feeling is held.
- 02
narrative depth
range 0–5“What's actually being said?”
Beneath the headline reaction, is there a real account or argument — parts that connect, a sequence, a cause, a consequence? Depth is not length. A one-line observation can carry more of it than a thousand-word rant that never leaves the same spot.
- 03
position clarity
range 0–4“Where does the writer stand?”
Can you tell what the writer actually thinks? A discernible position — even an unpopular or mistaken one — reads clearer than a paragraph that hedges every clause into mush. This measures whether there is a stance to respond to, not whether the desk agrees with it.
- 05
engagement quality
range 0–4“Did it start a real conversation?”
Did the thread open a conversation or collect applause? Substantive replies — corrections, additions, disagreements, follow-up questions — count for something a wall of agreement does not. Volume alone means little; what the replies contain means everything.
- 06
content type value
range 0–4“What kind of thing is this?”
Some kinds of post earn an answer more readily than others. A genuine question, a field report from real use, a considered review — each opens a door a brand can honestly walk through; a passing rant usually does not. This signal names the kind before the rest of the reading weighs it.
- 07
originality
range 0–4“Have we heard this before?”
Is this the hundredth turn of a take that circulates every week, or something the feed has not seen — a genuinely new angle, an unusual case, a question no one has answered well yet? Familiar is not worthless. But the desk has read it before, and says so.
- 08
readability
range 0–2“Can a non-expert follow it?”
Could someone from outside the niche follow the thread without a glossary at their elbow? Plain explanation scores; a jargon wall scores against. The narrowest band on purpose — clarity matters, but it is the last thing weighed, never the first.
The eight are read; they are not tallied into a verdict. Three rules govern how they are used, and each exists to keep a hard thing from being made falsely easy.
- The eight are never averaged.
- There is no single blended number, and there will not be one. The disagreements between signals are the insight: a thread strong on lived authority but thin on evidence is a different object from one thick with facts and no point of view, and a tidy average would erase precisely the distinction that tells you how to answer. Eight readings stay eight.
- Classification comes before scoring.
- Before a thread is weighed at all, it is sorted by who is talking — buyers in conversation with buyers, or the industry talking shop among itself. These are two different rooms. A signal that means one thing in the buyer stream means another in the industry stream, so the stream is settled first and the reading follows from it.
- Most threads are let go.
- A thread the eight cannot support is not forced forward. The desk raises little and lets most pass — quietly, with no verdict written on the thread and no reader’s attention spent. Silence is the common case here, and it is not a failure of the thread or of the desk. It is the standard working as intended.
This document is the standard. It is not the machine.
The exact wording of the scoring prompts, the thresholds that decide where a thread is routed, the hard-block rules that stop one cold, and the calibration corpus the desk is tuned against are not published here, and they will not be. That is a deliberate line, not an oversight. A standard you can read and argue with is worth publishing; a set of tunings measured against a private library of examples is worth keeping.
The point of publishing the eight is that you can hold the desk to its own questions. The point of withholding the calibration is that the questions stay honestly hard to game. Both serve the same reader.
The standard is public; the calibration is ours.
- v1.0 · 2026-07-17First public statement of the standard.