Jakob Nielsen Published the 90-9-1 Rule in 2006, and the Evidence Since Is Thin

What is the 90-9-1 rule?

Jakob Nielsen's 90-9-1 rule states that in an online community 90% of members only read, 9% contribute occasionally, and 1% create almost all the content. Nielsen called the underlying phenomenon participation inequality. The split describes a distribution of effort across accounts, and it came from observing large user-contributed sites in 2006 rather than from a controlled measurement of any single platform.

The numbers get quoted as though they were measured yesterday. They were not, and the man who published them said so at the time.

Who came up with the 90-9-1 rule?

Jakob Nielsen published it at the Nielsen Norman Group on October 8, 2006. The article now carries the title "The 90-9-1 Rule for Participation Inequality in Social Media and Online Communities", and his wording is worth reading in full rather than in paraphrase: "90% of users are lurkers (i.e., read or observe, but don't contribute). 9% of users contribute from time to time, but other priorities dominate their time. 1% of users participate a lot and account for most contributions."

The opening sentence of that article carries the claim that survives everything else: "All large-scale, multi-user communities and online social networks that rely on users to contribute content or build services share one property: most users don't participate very much."

Did Nielsen present the 90-9-1 rule as a law?

No. Nielsen wrote that "user participation often more or less follows a 90-9-1 rule," and that phrasing is his own, not a later softening by someone quoting him. He then broke his own split twice inside the same article. Blogs, he wrote, "have even worse participation inequality than is evident in the 90-9-1 rule," closer to 95-5-0.1. Wikipedia he put at 99.8-0.2-0.003.

An author who publishes three different ratios in one piece is describing a shape, not a constant. Treat the 90-9-1 rule the way he wrote it: a rough sketch of a heavily skewed curve, with the real numbers varying by platform.

Does the 90-9-1 rule hold on modern platforms?

The direction holds and the exact figures do not. The strongest published test is van Mierlo (2014), "The 1% Rule in Four Digital Health Social Networks: An Observational Study", in the Journal of Medical Internet Research. Across four long-running support communities, 63,990 accounts created 578,349 posts. Fewer than 25% of accounts made even one post, and the heaviest contributors, whom the paper calls Superusers, accounted for 74.7% of content by weighted average.

That result cuts both ways. Roughly three quarters of accounts never posted, which is short of 90%. A tiny fraction still wrote most of what everyone else read, which is the part of Nielsen's claim that matters.

Source	Population measured	Reported split
Nielsen, Nielsen Norman Group, 2006	online communities in general	90% lurk, 9% occasional, 1% heavy
Nielsen, Nielsen Norman Group, 2006	blogs	95, 5, 0.1
Nielsen, Nielsen Norman Group, 2006	Wikipedia	99.8, 0.2, 0.003
van Mierlo, Journal of Medical Internet Research, 2014	63,990 accounts across four health support networks	under 25% posted at all, top contributors wrote 74.7% of content

Past that, platform-specific published data is scarce. Nobody has put a peer-reviewed 90-9-1 breakdown on Reddit, X, Bluesky, Hacker News or Facebook groups that you can cite with a straight face. The figures repeated in marketing decks trace back to one 2006 observation, and anyone who hands you a fresh-sounding split for a named platform this decade should be asked which dataset it came from.

What does participation inequality do to brand monitoring?

It means the conversation you can read is not a sample of what your customers think. If a small share of accounts writes most of the posts in a community, then a keyword search returns those accounts' opinions at a rate far above their share of the population. You are reading the loudest quarter, weighted toward the loudest one percent inside it.

That does not make the data useless. It makes it evidence of a different kind. A social listening feed tells you what is being said in public and who is saying it, which is exactly the input you need for reputation work and for finding people asking to buy. It does not tell you what the silent majority believes, and no amount of volume fixes that. Pair it with voice of the customer research that reaches people who never post.

The practical trap is share of voice read as market opinion. A competitor with three enthusiastic superusers can outrank you in mention count while losing on every commercial measure you care about.

How do you measure the contributor share of a community you monitor?

Count the accounts, not the posts. The formula is one line:

contributor share = (accounts that posted at least once / total accounts observed) x 100

Say you track a subreddit for 30 days and your tooling logs 12,000 distinct accounts that viewed or interacted, of which 890 posted or commented at least once. That is 890 / 12,000 x 100, or 7.4% contributors. Then run the same count on the top of the distribution: if 41 of those 890 accounts wrote 600 of the month's 1,050 comments, 4.6% of contributors produced 57% of the comments. Those two numbers tell you more about the community than any single mention ever will, and they are yours rather than borrowed from 2006.

Do the count before you decide a subreddit is worth working. A place where nine accounts write everything is a place where nine relationships decide your reception.

One subreddit counted

Accounts observed12,000

Accounts that posted7.4%

Comments from top 41 accounts57%

What should a brand change because of the 90-9-1 rule?

Stop treating mention volume as a vote and start treating individual contributors as people. The 1% is small enough to know by name in most communities, which is the entire premise of community marketing and the reason a heavy contributor's opinion of you propagates further than a customer survey ever will. It is also why a parasocial relationship with a prolific poster is worth more, and costs more when it breaks, than its raw follower count suggests.

The monitoring consequence is concrete. Rank what you find by intent rather than by frequency, because frequency measures the writer's habit and intent measures the reader's need. RedReplier watches Reddit, X, Bluesky, Hacker News and Facebook in one place, grades each mention by buying intent, and explains with AI why it was flagged, which is how you separate a superuser's tenth post this week from one person asking for a recommendation. That separation is the whole job of a social listening tool.

One more consequence: a skewed contributor base is easy to fake. When 1% of accounts produce most of the visible opinion, buying or fabricating twenty accounts buys a lot of apparent consensus, which is the mechanic behind astroturfing and a reason to check account histories before you believe a wave.

Frequently Asked Questions

Who invented the 90-9-1 rule?

Jakob Nielsen published it at the Nielsen Norman Group on October 8, 2006, under the banner of participation inequality. He was describing a pattern he had observed across large user-contributed sites, and he presented it as an approximation. The article is still online at nngroup.com and states the three percentages in his own words.

Is the 90-9-1 rule the same as the 1% rule?

They describe the same phenomenon at different resolutions. The 1% rule states only that 1% of a community creates most of its content, while the 90-9-1 rule adds the middle tier of occasional contributors. Academic work, including van Mierlo's 2014 study in the Journal of Medical Internet Research, mostly uses the 1% framing because the top of the distribution is the part that reproduces across sites.

Has anyone proved the 90-9-1 rule wrong?

Nobody has overturned the shape, and several counts have moved the numbers. van Mierlo (2014) found fewer than 25% of accounts posting at all across four health support networks, well short of the 90% lurker figure, while still finding that a small group wrote 74.7% of the content. Nielsen himself reported 95-5-0.1 for blogs and 99.8-0.2-0.003 for Wikipedia in the original 2006 article.

Does the 90-9-1 rule apply to Reddit?

No published study gives a reliable 90-9-1 breakdown for Reddit specifically. The honest position is that the skew is visible on any large subreddit, where a handful of accounts dominate the comment sections, but the exact ratio has not been measured in a source you can cite. Count it yourself for the subreddits you care about instead of importing a number.

Why does the 90-9-1 rule matter for social listening?

Because it tells you what your data is and is not. Public mentions are a record of what the contributing minority chose to write, which is the right input for finding buyers and for reputation work, and the wrong input for estimating what your whole customer base believes. Reading a mention feed as market research overstates the views of the most prolific accounts.

How do you get more of the 90% to contribute?

Nielsen's own advice was to lower the barrier: make contribution a side effect of something the user is already doing, such as a rating rather than a written review. He also argued for editing over creation, letting people modify existing templates instead of building complete entities from scratch. His answer to whether the inequality can be eliminated was two words: "You can't."

What is participation inequality?

Nielsen called the underlying phenomenon participation inequality, the pattern where a small share of users create almost all the content in a large user-contributed community. He described it as a general shape he observed across several sites in 2006, not a measurement of one platform. The 90-9-1 rule is one way of writing that shape into three rough tiers.

Can participation inequality be used to fake consensus?

A skewed contributor base is easy to fake, because when 1% of accounts already produce most of a community's visible opinion, buying or fabricating a handful of extra accounts creates a lot of apparent consensus. That mechanic is the same one behind astroturfing. Checking account histories before treating a sudden wave of mentions as real sentiment is the guard against it.

Why can a competitor's mention count be misleading?

Share of voice counts mentions, not commercial outcomes, so a competitor with three enthusiastic superusers can outrank you in mention volume while losing on every commercial measure that matters. The count reflects a few prolific writers' habits, not the market's opinion. That is the trap the post points to directly.

Why pair social listening with voice of customer research?

A social listening feed shows what the contributing minority chose to write in public, useful for reputation work and for finding people ready to buy, but it says nothing about the silent majority who never post. Voice of customer research reaches those people directly instead of inferring their opinion from the loudest accounts. Combining the two gives a fuller picture than either alone.