No number without a source

1. Named in the sentence. The publisher and the year sit in the same sentence as the figure, so a reader never has to hunt for who said it.
2. Primary only. The origin gets linked — the study, the official statistic, the organisation's own reporting — never an article about it.
3. The population is stated. A figure that does not say exactly who it describes is not a fact, it is an impression with a decimal point.
4. A quantity in words counts as a number. "Most people", "one in five", "for months afterwards", "a growing share" — the same claim in other clothes, and refused on the same grounds.
5. Untraceable means unpublished. There is no category for "widely accepted". If it cannot be traced, it does not run, and the sentence gets rewritten without it.

Every draft is read against the fact file before it can be published, and the build itself refuses to complete if a figure appears in a page that is not in that file. A draft that needs a number to make its point does not get the number — it gets rewritten, or the section is dropped. Several sections on this site were dropped that way. The mechanism is the same one that keeps the list below honest: the only claims recorded there are ones that were written, checked, and then taken out.

Nothing. No figure has passed this process, so no figure appears on the site, and this page has no list to show. If that changes, the source will be named here with its year and its method, and it will be named again in the sentence where it is used.

• Loneliness in retirement often arrives months later, once the novelty of free time has passed.Onset-duration claim about a population. No primary source exists in this file. Note: the phrasing originates in content-spec §2 item 14 (directAnswer + description) and must be fixed there as well as in the draft.
• The same place, months on, is full of people who know your name.Duration-to-familiarity claim. Contradicts the paragraph immediately above it, which correctly refuses to put a period on it.
• Most people judge the attempt during the first stage and conclude it failed. / Most people decline that role by withdrawing."Most people" is a prevalence claim, banned by name in content-spec §0.4.
• The hours that feel worst, which for most people are not the middle of the day."For most people" is a prevalence claim about when loneliness peaks. Untraceable here.
• Losing most of a shared circle is an ordinary outcome of divorce rather than an unusual one.Prevalence claim ("ordinary rather than unusual") with no source. The mechanism can be described without ranking how common it is.
• Most of what people do about loneliness in a marriage falls into three groups. / Commonly three things, in some order.Prevalence claim plus an enumerated taxonomy presented as fact — content-spec §0.4 bans a stated count of types.
• Plenty of marriages spend a season in this shape and come out of it; some stay there for years.Prevalence ("plenty of") plus duration ("for years"). No source.
• Being alone all day and loneliness get confused in retirement more than at any earlier stage.Comparative frequency claim across life stages. Would require evidence this site does not hold.
• Retirement usually removes several people from the circle at once.Quantified loss ("several people at once") stated as a pattern. No source.
• After sixty, several of those structures end within a fairly short stretch of life.Quantity plus timing claim about the life course. No source.
• A modest fixture attended weekly produces more company than a better one attended once.Comparative outcome claim presented as a mechanism ("that is the whole mechanism"). Rephrase as a description of how recognition builds, not as a measured comparison.
• Search results for how to make friends after 60 are mostly written by organisations that want something in return.Prevalence claim about a SERP. Internal phase-2 measurement is not a published source a reader can check; soften to "much of it".
• The left column is what people report saying.Imitates reported/survey data. content-spec §0.4 bans formulations that simulate evidence even without a number.
• Build the sentence out of three parts / a fourth part that nobody needs.Stated count on a numberless site; the same page's WhatWeDont promises "no rule with a number in it". Rephrase to name the parts without counting them.
• This is why recommendation lists fail so consistently as a genre.Frequency claim ("so consistently") about outcomes across readers. No source. Safe rephrase: "This is why recommendation lists are a weak genre for this — the list names activities, and the activity was never the mechanism."
• Attention on the feeling is the one input that enlarges it.Causal mechanism stated as an established, exclusive finding ("the one input"). §0.4 bans formulations that simulate evidence. Safe rephrase: "an evening spent trying to feel less lonely tends to go badly, because the feeling is the thing being attended to."
• Loneliness without a triggering event is not a stranger case than the other kind.Comparative prevalence claim (how common the no-event version is relative to the event version). Safe rephrase: drop the comparison — "Loneliness without a triggering event needs no explanation to be legitimate."
• People read them without belief, mostly for the language."Mostly" is a prevalence claim about the motives of a population. Appears in body and in FAQ answer. Safe rephrase: "People read them without belief, and the draw they describe is the language."