The face of the product asks in plain words. The names live here — one page down from the read, where they help instead of impress.
Before anything can be argued, the argument has to be staked — is this a dispute about what happened, about what to call it, about whether it matters, or about who gets to decide? Hermagoras named those four questions, and rhetoric has used them since.
Throughline uses the last of them as the reader question: on whose standing is this page speaking, and to whom? Direct address is the evidence. A page that says you is naming its buyer whether it means to or not.
The count is of sentences, not clauses, and a sentence addressing both buyers is counted under the nearer one. The counting rule travels with the finding.
It reads address, not intent. A page can name the right buyer in every sentence and still misjudge what that buyer wants — that is a different lens's question.
It also cannot read an audience that appears only in an image, a video, or a price table. Untranscribed media is out of its reach and says so.
Minto's rule, learned writing for consultants: a piece of writing has one governing thought, and everything below it exists to support that thought. Ideas group; groups summarise upward; the summary goes first.
So the question is two-part. Is there a single thought? And does the reader meet it before the support for it?
A late thought is a structural finding, not a stylistic one: the fix is an order change, and it costs one sentence.
It cannot tell you whether the thought is true, or whether it is the right thought to promise. It reads the architecture, and architecture can hold up a bad idea perfectly well.
Every claim persuades from one of three places: what follows from what (logic), who is speaking (credibility), or what the reader feels (emotion). Aristotle's three, and the oldest counting frame in the set.
Throughline counts at the sentence, under the single strongest appeal present, and reports the mix and the order — because the order is usually where the opportunity is.
Fourteen sentences carry two appeals at once. Counting both moves the figures by about a tenth and does not move the finding — which is why the rule is printed beside the number.
It cannot rank the appeals. A logic-heavy page is not better than a credibility-heavy one; the question is fitness for the buyer and the moment.
On the face this lens says logic · credibility · emotion. Never logos · pathos · ethos — the Greek belongs on this page, not on a customer surface.
Adler's first demand of any reader was to state what the thing is about as a whole — in a sentence, in your own words. The Syntopicon then indexed the ideas themselves, so a passage could be found by the idea it belongs to rather than the author who wrote it.
Here it asks the uncomfortable version: what is the page about underneath what it says it is about?
This is the read a competitor makes first, and the one the page would least expect.
It cannot be counted the way an appeal can, and it will not pretend otherwise: the claim rests on named removals and one absence. A reader who treats the tooling as the replacement reads the page as complete, and that reading is available from the text.
It is prose-bound in part. The subject can be drawn as a two-point axis and no further — a diagram has no grammar for “underneath”.
Vonnegut drew stories as lines: fortune up the vertical, time along the horizontal. His point was that the shapes repeat, and that you can see a story's shape before you can say what happens in it.
A page has a shape too. Throughline counts beats — a beat is where a claim completes — and reads where the line rises, where it flattens, and what sits after the ask.
A beat after the ask can only reassure. Two of these three would reassure harder in front of it, where the hesitation happens.
It cannot tell you a shape is wrong. A flat close is a fault on a launch page and a virtue on a pricing page; the lens reports the shape and names where the ask sits.
Hoey's finding was that words carry the company they have kept: every word arrives primed by the positions, collocations and textual places a reader has met it in before. Endings are the most primed position in any text.
So this lens asks what the page leaves the reader holding — and whether a sentence anywhere gathers the argument before the page stops.
An empty finding stays in the read. A lens that disappears when it finds nothing makes every other finding harder to trust.
It cannot read a close carried in an image or a video, and it says which blocks it could not transcribe.
It is prose-bound: at every density it arrives as sentences, with at most an anchor marker beside them. There is no honest geometry for an absence.
Schriver's method was protocol-aided revision: put the people who wrote a document behind the glass and watch readers who had never seen it try to use it. The drafters were reliably wrong about where the reading would break — not careless, just unable to un-know what they meant.
So this lens asks the one question a writer cannot answer from inside their own page: what does it ask of a reader who is not you? It scores the page's own captured phrases and reports the school year they need at first pass.
A high grade is not automatically a fault. A page written for procurement counsel should read like one. The number sizes the ask so it can be a decision instead of an accident.
It cannot tell a necessary technical term from an avoidable long one. A page whose subject genuinely requires named things will score high and be right to — the lens reports the ask, never whether the ask is fair.
It is prose-bound: a reading level is one number over one corpus, and this read has measured no second axis to draw it against.
It is not Schriver's whole subject. Her research also grounds terminology consistency, which this product reads through Hoey and Fowler — that finding lands on how it's written, not here.
Fowler's guardrail: one idea, one word. Reaching for a synonym to avoid repeating yourself makes a reader wonder whether you meant something new — his name for the habit was elegant variation, and he did not mean it kindly.
Throughline reads the sentences themselves: mood, voice, and where the hedges sit. Hedges are the useful signal — they mark where the writer was least certain.
The hedge cluster is a map: where the sentences soften, the argument has not been settled.
It is prose only, at every density. There is no geometry for a mood, and inventing one would be decoration — a mark that carries no claim, no relation and no quantity.
At Mostly visuals it arrives as a tighter paragraph beside the others' geometry. If that reads as a gap, the layout is wrong, not the rule.