AI tools helped gather research and write the first draft of this article. I edited it and checked every platform rule, figure and quotation against the platform's own page, the study's official record or a news report that quotes the platform, listed at the end. The cover illustration is AI-generated. Details at the end.
The easiest number to collect about an author is how much they published. It is exact, it is available on the first day, and nobody argues about it. It is also the number that says least about what the author gave readers.
Twenty short posts in a month and four long ones are both "output". One set may have been skimmed and forgotten; the other may have been read to the end, quoted by other writers and still found through search a year later. A count of posts cannot tell these apart. A count of views cannot either, because a view records that a page was opened, not that anything on it was used.
This article asks a narrow question: if output is not enough, what else can be measured about an author's contribution, and what does each measure miss? It looks at what several large platforms decided to count, at a classic observation about who takes part online, and at one research idea from Wikipedia that carries over to articles.
This is the sixth part of Influence & Incentives, a series on how other people, rules and rewards shape what we decide, write and publish. It opens a short run inside the series about authors and creators. This part is only about what to count. How rewards attached to those numbers change behaviour, and how the numbers themselves get gamed, are questions for the following parts.
Layer 1: output is a gate, not the measure
Platforms that pay creators rarely pay for output alone. They use output as an entry condition and pair it with a signal from the audience.
When Twitch launched its Affiliate programme in April 2017, the criteria for an invitation were: at least 500 total minutes broadcast in the last 30 days, at least 7 unique broadcast days in that period, an average of 3 or more concurrent viewers over the same 30 days, and at least 50 followers [1]. The first two conditions measure work done. The last two measure whether anyone came to watch it. Twitch added that the criteria were likely to change as the programme developed [1].
YouTube's Partner Program today asks for 1,000 subscribers together with either 4,000 qualified watch hours in the last 12 months or 10 million qualified Shorts views in the last 90 days [2]. There is no count of uploads in that rule at all. The channel has to have produced something, of course, but what is measured is the time and attention viewers spent on it.
The pattern is the useful part. Output shows that an author is active. Something else has to show that the activity reached anyone.
Layer 2: attention actually spent
The most direct "something else" is time. In October 2019 Medium said it would pay writers in its Partner Program based "primarily" on reading time rather than on claps, its version of a like button, TechCrunch reported [3]. Medium counted only active reading time: if a reader left the window open and stepped away, that time should not count [3].
Medium's explanation is worth reading closely. A Medium representative, Emma Smith, called reading time "a closer measure of quality and resonance with readers", and said it reflects all readers, not only those who clap [3]. That second point is the important one. Claps come from the readers who choose to react. Reading time comes from everyone who read.
Attention has limits as a measure. A long article earns more minutes than a short one that answers the same question better. A confusing text can hold a reader longer than a clear one. Time says that a reader stayed, not that the reader was helped. It is a better signal than a click, not a final verdict.
Layer 3: responses, and the readers who never respond
Comments, replies, shares and reactions look like the clearest sign of contribution: someone cared enough to answer. They are worth counting. But they describe a small and unusual part of the audience.
In 2006 Jakob Nielsen described what he called participation inequality: "In most online communities, 90% of users are lurkers who never contribute, 9% of users contribute a little, and 1% of users account for almost all the action" [4]. The exact proportions vary by community, but the shape has held up well enough that it is still a standard reference.
Two consequences follow for anyone measuring an author:
- Responses are a sample, and not a random one. They over-represent the most active, most opinionated readers. An article that quietly helps many people can collect fewer comments than one that annoys a few.
- The number of responses matters less than what they contain. A correction, a follow-up question or a reader adding a source is a different contribution from a one-word reaction. Counting them together hides that.
This is also why Medium's argument for reading time rested partly on the silent majority [3]: a measure built only on visible reactions measures the 10%.
Layer 4: what survives
A different way to measure contribution is to ask what happened to the work afterwards. Did it stay in use?
In 2007 B. Thomas Adler and Luca de Alfaro proposed a reputation system for Wikipedia authors built only on how content changes over time, not on ratings or comments [5]. In their system an author gains reputation when later editors keep the author's edits, and loses reputation when the edits are quickly undone [5]. They applied it to the full Italian and French Wikipedias, more than five million revisions in total, and found that edits by low-reputation authors were significantly more likely than average to be judged poor quality by people and to be undone later [5].
The idea carries beyond encyclopedias. Other people's later decisions are evidence about earlier work. For an article, the equivalents are concrete:
- Is it still being read months after publication, through search or links, rather than only in the first two days?
- Did other writers cite it, link to it or build on it?
- Did it need corrections, and were the corrections small or fundamental?
- Was it updated when facts changed, or left to go stale?
Persistence is slow. It cannot judge an author in the first week. But it is the layer that most directly measures whether the work was useful, because it records what other people chose to keep.
Layer 5: who else benefits
The last layer looks outward: does the author's work help other authors and readers find each other?
In September 2022 Substack wrote that "the Substack network is driving more than 40 percent of all subscriptions across the platform", and 12 percent of paid subscriptions [6]. It said the effect came from network features, including Recommendations, which it described as a human-powered system that lets writers suggest other writers to their readers when they subscribe [6]. The 40 percent is for the network as a whole, not for recommendations alone [6].
The point for measurement is that a writer's contribution can show up in someone else's numbers. An author who recommends, cites and links to good work by others is adding to the whole publication, even when their own counts do not move. A dashboard that looks at one author at a time will never see this.
A recent count beats an old total
One more lesson concerns time windows. In September 2025 Reddit removed the public member count from community pages and replaced it with weekly activity: visitors over the past seven days, and contributions, meaning posts and comments in the past week that were not removed, according to reports in Social Media Today and Tubefilter [7][8]. Reddit's reason, as quoted there, was that member totals "have never fully reflected true engagement", because people "don't need to be a member of a community to post or comment" [7].
A follower count is a total that only grows; it says what happened over years. Recent activity says what is happening now. For authors the same applies: lifetime followers or lifetime views reward having started early. A rolling window, compared with the author's own previous windows, shows whether the work is still reaching people.
Five layers, and what each one misses
Putting the examples together gives a simple map. None of the layers is a score on its own.
- Output. What was published, and how regularly. Misses: whether anyone used it.
- Attention. Time actually spent reading, active rather than open in a tab. Misses: whether the time was well spent; favours length.
- Responses. Replies, questions, corrections, sources added by readers. Misses: the large majority who read and never respond.
- Persistence. Reading months later, citations, links, corrections needed. Misses: everything in the first weeks; slow to collect.
- Network. Readers and writers the author sends onward, and work by others the author helps surface. Misses: the author's own reach; hard to attribute.
Reading the numbers honestly
A few rules make these layers more useful and less misleading, whatever tool collects them:
- Use at least two layers. Output with attention, or attention with persistence. A single number invites a single kind of effort.
- With few authors, use counts, not percentages. When a group has twenty or thirty people, one person changes a percentage dramatically. "Four of twenty-six" is more honest than "15%".
- Compare an author with their own past, not only with others. Topics, formats and audiences differ. The question "is this author's work reaching more people than three months ago?" is fairer than a league table.
- Separate the staff from the audience. Replies from editors or the author's colleagues are not the same signal as replies from readers. Count them apart.
- Keep a window. Report the last 30 or 90 days next to the lifetime figure, so early starters and newcomers can be compared.
- Write down what each number means before using it. If a team cannot say in one sentence what "an engaged reader" is, the number will mean whatever is convenient later.
A worked example (fictional)
The two authors and all figures below are invented to show the layers in use.
In one month, Author A publishes twenty short posts. Author B publishes four longer guides. On output alone, A looks five times more productive.
Look at the other layers. A's posts are opened often, but the median active reading time is under a minute, and three months later almost no one reads them. No one linked to them or sent a correction. B's guides average several minutes of active reading. Two of them are still found through search after three months. One reader sent a correction to a figure, which B fixed and noted. Another writer on the same site linked to one guide as background for their own article.
Neither author is "better" on this evidence. A may be doing a different job: fast news, which is useful and short-lived. The point is that the two contributions are different in kind, and output alone would have ranked them in the wrong order for any question about lasting value. A fair review would describe both with at least two layers each, in plain counts, over the same window.
Five questions to ask of any author dashboard
- Does it show anything besides output? If the only numbers are posts and views, it measures activity, not contribution.
- Is time counted as active reading, or as a page left open?
- Whose responses are counted? Readers only, or also staff and colleagues?
- Is there a window? Lifetime totals next to the last 30 or 90 days.
- Does it see what happens later and elsewhere? Reading months later, citations and links, and readers sent on to other authors.
What this article cannot tell you
The platform rules described here say what Twitch, YouTube, Medium, Substack and Reddit chose to count at particular dates; several of them have changed since and may change again. They are evidence of what large platforms considered meaningful, not proof that any measure works well for a small publication. Nielsen's 90-9-1 is a widely cited rule of thumb from 2006, not a law; the proportions differ between communities. The Wikipedia reputation study measured edit survival on an encyclopedia, which is not the same as an article's lasting value. Medium's and Reddit's statements are cited through news reports that quote them, because their own pages could not be read automatically. Nothing here measures any author on Dapp.Expert, and this part does not cover how paying for these numbers changes what authors do, or how the numbers get gamed; later parts of the series take those up.
How this article was made
| Task | What the AI did | What a person did | How it was checked |
|---|---|---|---|
| Research | Collected platform rules and studies on creator metrics into research notes | Chose the question and which sources to rely on | Rules, figures and quotations checked against platform pages, the study's official record and news reports quoting the platforms [1]–[8] |
| Draft | Wrote the first draft from the notes | Edited the text, the structure and the worked example | A separate review compared figures and claims with the sources |
| Images | Generated the cover photograph-style illustration | Chose the illustration; designed the explanatory diagrams | Labelled as AI-generated in the disclosure line and alt text |
| Not done by AI | — | The decision to publish and responsibility for the text | — |
Conflicts of interest: none. The author has no commercial relationship with the platforms, researchers or publications named; the authors in the example are fictional.
Responsible for this article: Dmitriy Lastovka, author. Corrections: contact form.
Sources
Labels: official (the platform's own announcement or help page), paper (peer-reviewed study, read via its official record), research note (an expert's published article), report (a news report that quotes the platform).
- Twitch, "Twitch Affiliate Program Launches Today, First Invites Going Out", Twitch Blog, 24 April 2017. https://blog.twitch.tv/en/2017/04/24/twitch-affiliate-program-launches-today-first-invites-going-out-9c020994a042/ Official.
- YouTube Help, "YouTube Partner Program overview & eligibility" (eligibility thresholds), read 15 November 2026. https://support.google.com/youtube/answer/72851 Official.
- Anthony Ha, "Medium says it will compensate writers based on reading time, not claps", TechCrunch, 22 October 2019. https://techcrunch.com/2019/10/22/medium-reading-time/ Report.
- Jakob Nielsen, "The 90-9-1 Rule for Participation Inequality in Social Media and Online Communities", Nielsen Norman Group, 8 October 2006. https://www.nngroup.com/articles/participation-inequality/ Research note.
- B. Thomas Adler, Luca de Alfaro, "A content-driven reputation system for the Wikipedia", Proceedings of the 16th International Conference on World Wide Web (WWW 2007). https://doi.org/10.1145/1242572.1242608 Paper, official record and abstract.
- Substack, "Substack's most powerful feature is growth", On Substack, 29 September 2022. https://on.substack.com/p/growth Official.
- Andrew Hutchinson, "Reddit Updates Activity Indicators on Subreddit Communities", Social Media Today, 9 September 2025. https://www.socialmediatoday.com/news/reddit-updates-subreddit-activity-metrics/759689/ Report.
- James Hale, "Reddit is revamping its metrics to emphasize content creation (with a hat tip for lurkers)", Tubefilter, 10 September 2025. https://www.tubefilter.com/2025/09/10/reddit-new-performance-metrics-member-count-visitors-contributions/ Report.










