S2N Gauge v2.1
News is sold on coverage; intelligence is sold on consequence
Making that move means being able to say which of your output a subscriber could act on, and being believed. The Gauge measures it sentence by sentence, for a named reader, against the norms of the genre, and keeps the reasoning so the answer holds up under challenge.
The part of the business with no instrument on it
Every other part of a high-trust information business is measured. But your editorial quality is judged, by whoever is in the room, in the language of taste, one piece at a time. So perceptions of both improvement and decline are invisible until renewal, and you set strategy and direction without evidence.
Volume, page views and editor intuition cannot answer continuously what output a subscriber could act on, or how confident they can be in it.
But the Gauge makes quality something you can benchmark, price against, and show a buyer.
What it asks of every sentence
Does it tell this reader anything?
A sentence that is news to a generalist can be old news to a specialist, so the reader is recorded before scoring starts. (Signal)
Is the claim grounded?
Whether the reader could check it. The most direct read there is on sourcing discipline. (Confidence)
Does it change what they would do?
Reporting that does not frame or affect a decision scores low here; reporting that shapes choices scores strongly. (Actionability)
What it looks like
Three sentences from one of my own pieces, HTIBs Are Losing Their Best Sourcing When It Matters Most, scored by the Signal-to-Noise Gauge in a single run over its 53 sentences. Each score comes with the reason the Gauge gave it.
Scoops look the same as rumours, and at renewal, content that cannot be told apart from a cheaper alternative gets priced as one.
- Signal
- 8Connects the provenance failure to a commercial pricing consequence at renewal; non-obvious and decision-relevant framing.
- Confidence
- 5Reasoned commercial inference; no source; plausible and coherent but asserted rather than evidenced.
- Actionability
- 8Names the revenue consequence directly and supplies the comparative frame: scoop priced as rumour at renewal.
Watch the price reporting agencies, which already sell grading and provenance and know what they are worth, start to apply this in areas where their coverage faces more competition.
- Signal
- 8Names a specific market actor class and a specific competitive dynamic; interpretive novelty in identifying PRAs as the leading signal.
- Confidence
- 4Forward-looking prediction; no sourcing; evaluative claim about PRA competitive response is unverifiable.
- Actionability
- 7Directly shifts reader’s monitoring calculus: PRAs as the leading indicator to watch — decision position changes.
The corpus produced several findings.
- Signal
- 1Pure connective tissue; no content delivered; entirely anticipatory.
- Confidence
- 5Structurally true given the preceding sentence; verifiable in principle from the article itself.
- Actionability
- 1No decision position shifted; reader learns nothing actionable.
The first has consequence for a reader at renewal. The second is sharp but unsupported: it is a forecast with no source behind it. The third is connective tissue and scores 1, which is why the 8s mean something.
Where Signal and Actionability are high and Confidence is low, the flag tells you to verify, to push for better attribution, or to say plainly that your sourcing is thin.
How a document gets scored
Read
Your article is broken into sentences, and who it was written for is put on the record.
Judge
One pass by a language model. It scores each sentence and writes a reason for each score.
Verify
Encoded rules check that work against the standards for this kind of writing.
Aggregate
Sentence scores roll up into four figures, and one file that shows the working.
One pass of judgment, then no more: a language model does what it is good at, reading a sentence closely and saying what it is doing. Everything around that pass is editorial expertise written into rules and tested.
What comes out
Signal
How much of your output carries Signal: the share of sentences that tell your reader something.
Confidence
How much is grounded: the share your reader could check for themselves.
Actionability
How much is consequential: the share that changes a decision.
- One composite, combined so that strength on one dimension cannot pay for weakness on another.
Alongside them, a file holding every sentence, every score, every reason, and the record of the run that produced them. Any figure can be walked back to the sentence it came from.
It gives your title a benchmark: a baseline against the norms of its genre, and against your own past. Make changes, then measure again.
Who reads it, and for what
- Commercial
- Quality is asserted in every renewal conversation and proved in none. The Gauge provides you with evidence.
- Directors
- Your title’s characteristic strength or weakness, named across your body of content and tracked over time.
- Editors
- Which of your sentences and pieces need attention, and why: where Signal, sourcing and Actionability pull apart.
- Investors
- An independent read on your investment: is your product improving, and how does it stand against comparable titles?
“We could do this with Claude”
A summary is not the product. A model can summarise an article. The Gauge measures what a summary leaves out: what is new to a specified reader, and what it tells them to do.
Every score can be challenged. Each score comes with its reason and a record of the run that produced it, so it can be inspected, questioned, dismissed or re-run.
The scoring standards are written down, versioned and locked. Only the reader profile changes from one publisher to the next; the standards stay the same.
What I can show so far
I have run the Gauge on 4,091 sentences of high-trust B2B content from one specialist sector.
Against a written reader profile (ten articles, 436 sentences), two separate runs agreed on 91% of sentences about whether each bore on the reader’s decisions. Where the Gauge said a sentence did, an expert reading blind agreed four times in five (16 of 20). Its errors were mostly misses: consequences left implicit.
It describes pieces and titles, never individual writers.
Scored examples from three public documents are in the Observatory: The Gap Is Now Measurable.
Start with a first run
A first run is 15–20 of your articles. How it starts.
Start the record now. A measured history against independent, documented standards gives sales teams ammunition and buyers something to check.