Five rules for building a successful online community

[Editor’s note: Robin ‘Roblimo’ Miller is Editor in Chief of OSTG. He has also written three books about computing and the Internet and wrote hundreds of freelance articles for assorted newspapers and magazines before he learned how to make a living on the Internet. Miller also is a member of OJR’s new editorial advisory board.]

I often shudder at the poor quality of online forums run by newspapers and other local media outlets. Come on, people! This reader interaction thing may be new to you, but some of us have been doing it for 10 or 15 years, and have a pretty good idea of what works and what doesn’t.

This article outlines five basic rules for building sustainable online communities that are based on my 15+ years of experience with various online services, discussion groups, usenet forums, and — for the last seven years — as part of the management team behind the famous Slashdot discussion site. I’m not saying that you should follow slavishly in my footsteps, but I assure you that a forum you build (or rebuild) in accordance with my rules will be more popular, easier to manage, and more profitable than one that doesn’t follow them. These rules — and the software that helps enforce them — are the driving force behind hundreds of popular and profitable discussion-based Web sites.

Rule One:

Your discussions must be threaded or nested, not just “flat.”

A flat discussion tags the newest comments onto either the top or the bottom of ones already listed. A threaded discussion shows “discussion threads” but doesn’t display the entire content of posts replying to “parent posts,” just their subject headers.

The Sarasota (Florida) Herald-Tribune has “flat” message boards. In this example, a poster has replied to three other posts, but the new posts are not associated directly with the ones to which they are replying. This discussion had 62 posts at the time this screenshot was taken, and it was almost impossible to follow any of the sub-conversations within it because of the way it was displayed.

Screen grab

Kuro5hin is a “geek interest” news and discussion site that uses software based on Slashdot’s. This screenshot shows part of a 62-message discussion displayed in “threaded” mode.

Screen grab

Groklaw discusses legal issues related to free software. This example of “nested” discussion display is part of a string of over 300 reader comments attached to one article. Slashdot and Groklaw routinely run articles that draw 1000+ reader responses. Threading, nesting or some other sorting mechanism is necessary to keep discussions this large from becoming unintelligible.

Screen grab

Even in small discussions (20 or fewer posts), conversations are easier to follow if new comments are linked directly to comments they are responding to than if they are displayed in the order in which they were submitted.

Without reader-to-reader conversations, an online forum is nothing but a giant “letters to the editor” page. While posting responses to your published stories gives your readers more voice than they’d have without this ability, your forums or bulletin boards (or whatever you want to call them) will only achieve their full potential when readers start using them to talk directly to each other instead of merely reacting to content you have posted.

Rule Two:

You have readers who know more than you do about any given topic — and plenty of readers who don’t know nearly as much as they think they do.

I’ve been writing online long enough to realize that I should be be thankful for readers’ corrections and accept them graciously instead of letting them upset me. It takes a while to accept the constant barrage of criticism and nitpicking you get if you have reader comments attached directly to all articles on your site, but in the end you and your fellow writers will become more careful reporters.

One thing many online writers have noticed over the years is that compliments are more likely to be sent to you by private email, while critical comments are more likely to be posted on public boards. I’ve also noticed — speaking strictly from my own experience — that unfair public attacks from uninformed or mindlessly vituperative readers almost always draw rebuttals from other, more knowledgeable readers. I have learned not to get into arguments with readers who attack my online work in public, but to trust other readers to come to my defense if I have been wronged.

Of course, if I make a factual error or grammatical mistake and a reader posts a comment about it, the right thing to do is post something along the lines of, “Corrected. Thanks for noticing.”

This makes it clear to the readers that I pay attention to posts attached to articles published under my byline, and makes it even more clear that I respect my readers and happily give them credit if they give me informed criticism that helps improve my work.

The only problem with this philosophy is that it can be hard to separate experts from yow-yowers, especially if you’ve written about a topic area in which you are not an expert. But that’s why we have…

Rule Three:

Let your readers judge each other so you don’t have to judge them yourself.

Slashdot, Groklaw, Kuro5hin and many other geek-oriented discussion sites have moderation features built into the software that drives them. Slashdot’s moderation scheme, from which the others were derived, works like this:

  • You cannot post in and moderate the same discussion; you cannot, therefore, moderate your own posts.
  • Moderation powers are distributed semi-randomly, and only to readers who have had login identities for at least a few weeks. And no individual reader gets more than a few moderation points at a time, so it’s hard for one knucklehead to mess up the whole scheme.
  • Obscenities, personal attacks, and other unwelcome speech will almost inevitably be moderated down into oblivion. “Community standards” have been used as a legal test of what constitutes obscenity. Give your readers the power to moderate other readers’ posts, and you will soon find what they consider obscene.

Slashdot has tried all sorts of additions and tweaks to its moderation system over the years, so many that a pretty good percentage of the Slashdot FAQ (Frequently Asked Questions) page is dedicated to comments and moderation.

The point of moderation is to separate dreck from diamonds. Readers who aren’t logged in view Slashdot comments that are rated +1 or above (on a -1 to +5 scale) but do not see comments rated 0 or -1 without special effort. Logged-in users’ comments post automatically at the +1 level, while comments from readers who are not logged in start at 0. A post from someone who is not a logged-in user, therefore, needs at least one logged-in reader to consider it worthy of a positive moderation point before most readers can see it at all, while a post from logged-in user that a few users who have moderation powers that day find offensive can easily drop from public view.

On the positive side, comments that add something useful to the discussion will be moderated upwards, so readers who only want to see the most cogent comments can set their preferences so that they see only comments moderated to +2, +3 or even +5.

On the negative side, you may want to give readers a little help. Most Slashdot-type posting systems allow employees or other selected forum monitors extra moderation privileges so that they can save readers from the task of removing strings of especially vituperative comments.

You may also want to only allow comments from registered, logged-in users. Slashdot allows anonymous comments because of the “whistle blower” factor; some of the site’s best posts have always come from people who might lose their jobs if they posted inside information about their employers’ actions under a traceable name. In return for occasional anonymous gems, Slashdot suffers from plenty of anonymous garbage down at the 0 and -1 moderation levels. You may decide this tradeoff isn’t worthwhile, and I won’t blame you if you take the easier course. I often wish we’d taken it ourselves.

In any case, you need to realize that your forums will need some watching and nurturing if they are ever going to become a valuable part of your online offerings.

Rule Four:

All good things must come to an end.

You can’t leave online conversations “open” forever. Sooner or later you need to close them off, if only to keep comment spam from taking over posting threads on older stories. You may chose to allow comments on stories for as long as 30 days, although you’re probably better off closing comments on most stories after a week or two if you publish weekly, daily or constantly.

Archiving older discussions as static pages instead of serving every completed conversation on your site as a dynamic page can also save dramatically on server usage, which will help keep costs down.

Rule Five:

Why buy a cow when the software is free?

By now you’re probably saying, “Whoa, man… where can I buy the cool software that runs Slashdot?”

I’m sorry. We don’t sell Slash. We give it away. For free. Right here. The only caveat is that if you figure out a way to make Slash run better or more efficiently, we ask you to share your improvements with us and other people who use Slash under the terms of the GNU General Public License.

Scoop, the code that runs Kuro5hin, is also freely available, as is the Geeklog software behind Groklaw. And these are just a few of the best-known free content management systems out there that have Slashdot-like comment and moderation systems. There are many others. Slash is far from the easiest one to install, customize, and maintain, but it is also the most proven one for sites that may deal with millions of pageviews and tens of thousands of comments every day. Your IT people or hosting people may already have a favorite piece of free forum software; if so, that might be your best choice as long as that software is already being used successfully to power sites at least as large as yours is likely to become in the foreseeable future.

But the main thing isn’t the software. It’s your (and your management’s) attitude. It is not easy to give readers near-total control over some of your vital Web real estate. There is an endless temptation to do things like create topics you think will interest readers instead of letting your readers choose what to discuss on their own.

Communities aren’t created by management fiat. They grow on their own. You can provide a fertile environment for yours, and nurture it with light-handed moderation and by having staff members participate in its early conversations.

Note that I haven’t mentioned blogs as a factor in any of this. A reader-driven forum that allows users to start new topic threads gives readers the option of posting entries that are similar enough to blogs that calling some threads “blogs” becomes redundant.

And when it comes to staff members blogging… perhaps I’m showing my age here, but I remember when the people we now call “bloggers” were called “columnists.” But that’s another discussion for another time.

'Comment is Free,' but designing communities is hard

In a recent lecture to the Royal Society of Arts (RSA) in Londoni Guardian editor Alan Rusbridger candidly admitted that established newspapers had an equivocal relationship with the concept of reader-based discussion. “Occasionally the little people would write a letter… and we would print a few, very graciously” he noted wryly. With the growth of the Internet email-based submission of letters was added, but this changed little except the volume and speed of response. With the development of online newspapers a variety of models have been tried with limited success, including Web-based discussion forums, separated from newspaper content; published the email addresses of journalists at then end of stories; and hosting live debates.

The growth of Weblogging has forced newspapers to address this relationship (for a recent development see Syndicate this! Linking old media to new, Stephen Bryant, OJR, 2006-03-27), and not just because of the rise of online discussion. The development of easy to use publishing tools has allowed newspaper columnists and opinion writers to develop their own spaces onlineii. And the syndication concept that underlies the Weblogging model allows commentary to be found and aggregated away from the hosting news site.

The Guardian situation

In the UK the Guardian newspaper, particularly notable for its columnists and the quality of its letters page, had long employed Web-based discussion forums such as Guardian Talk as part of its Guardian Unlimited network.

Having visited the US to research new developments, editor Alan Rusbridger observed the disagreggation of advertising and editorial, and the changing nature of editorial. He also observed the growth of non-traditional opinion aggregators such as the Huffington Post, which was competing with everything from The Nation and MotherJones to the Wall Street Journal and the Philadelphia Inquirer. On the home front he noted that the Guardian‘s readers had “started talking to one another, and going behind our backs to our sources and reports.”

Characterizing the papers of record as being stuck in ‘journalism as revelation’ mold, he questioned whether saying “We have got all these distinguished newspaper columnists” was an adequate response. “They’re not laughing at Ariana Huffington now,” he observed. “They are saying, ‘That is really interesting.'”

In response to these developments, in March this year, the Guardian launched Comment is Free, a “collective group blog, bringing together regular columnists from the Guardian and Observer newspapers with other writers and commentators representing a wide range of experience and interests”iii. Comment is Free is edited by Guardian veteran Georgina Henry and politics page editor Tom Happold, and inspired by celebrated Guardian editor C. P. Scott’s famous aphorism ‘Comment is free but facts are sacred’iv. Contributors, of who there are now more than 200, are asked to write without remuneration (other than those who are published in the paper), with a £75 fee offered if their contribution is promoted as a pick of the day – a model not disimilar to the BlogBurst model discussed in Stephen Bryant’s piece. The payback for authors would appear to be access to customised publishing tools, and a larger audience than they might have for their own Weblog.

Comment is Free overview

The Comment is Free home page includes a Blog section (listing recent posts), Editors’ Picks, and an overview of all the comment from the Guardian and the Observer. The most active posts are listed, next to a ‘Best of the web’ selection. Each author has an associated RSS Webfeed, and each post includes links to ‘Digg this’, ‘Add it to’, and ‘See who is linking here’ (using Technorati).

In order to post responses to pieces, commenters have to have registered and signed in for Guardian Unlimited blogs. Comments are limited to 5,000 characters, and flagged with a country (inferred from the user’s IP address). No HTML formatting is allowed, and there is no comment preview function. Each comment has a unique URL, allowing it to be directly ‘pointed to’. Post-moderation is employed, and each comment has a link allowing it to be flagged as offensive or unsuitablev.

The overall interface design of Comment is Free was developed in house by the Guardian Unlimited design team with direction from Guardian creative director Mark Porter. It is a good and appropriate improvement on the 1998 Guardian Unlimited design, though it presents a similarly confused information architecture, eliding links to comments from ‘This week’, ‘Subjects A-Z’, the ‘Editors’ blog’, and the ‘Steve Bell’ cartoon.

As with most online publications, the Comment is Free interface is text-driven and linearvi, and fails to exploit people’s visual powers, even to the extent, perhaps, of making more commented on posts more attention grabbing – a model pioneered as early as 1997 in the discussion areas at the BBC’s commercial site Beeb.comvii.

In May, two months after its launch, Henry announced the Big Blogger contest, to elicit nominations for a commenter to become a regular Comment is Free blogger. This initiative was in response to the feedback that “not only is the debate on most of the threads not as bad as I sometimes make out, but… in many cases is of higher quality than the posts by some ‘professionals'” (Big Blogger: let battle commence, May 23, 2006)viii.


The Guardian appears to consider Comment is Free to be a success, though Henry is frank about the challenges with which they are dealing. To its credit Comment is Free was nominated for the Innovation category of New Statesman New Media Awards 2006, though it didn’t win.

In his RSA lecture Rusbridger was honest in identifying the Huffington Post as the inspiration for Comment is Free, and Henry notes that the Post “has outstripped its liberal old media competitors in the 10 months since it launched” (Welcome to Comment is free, March 14, 2006). Despite being a ‘me too’ product the Guardian deserves credit for the ambition and scale of its engagement with the developments in online debate.

Flaws in the model

However, Comment is Free has a number of flaws. Some possible flaws (around the defensive selection of authors and tiny fraction of readers who comment) were addressed by Bob Cauthorn in Can newspapers do blogs right? (Robert Niles, OJR, 2006-04-23), and were responded to by Guardian Unlimited assistant editor Neil McIntosh in his comments posted on the piece.

Although Comment is Free is presented as a Weblog-based service, the posts often don’t follow a basic tenet of Weblogging: linking. This doesn’t significantly undermine the system, though it reinforces the feeling that this is an established newspaper space. This limitation is partly a product of the existing Guardian Unlimited infrastructure. One must assume that any columnist researches their piece online, and saves links and references with their manuscript (not least for fact checking). I suspect there is no way for these references to make it through the Guardian‘s print and online publishing system and onto Comment is Free – and the product suffers somewhat for it.

Comment is Free also appears to have ignored another tenet of Weblogging: instant publication. According to one author “you do not actually have access to ‘your’ page. You send copy to the editors, who then vet it, edit it and put it online at their own pace. There is no immediacy, and no direct control of your copy. The entire feeling of speaking directly to readers is lost, as it is so heavily mediated, not just by the editing process but the technology itself”.

The dynamics of posting a comment in Web-based fora have rarely been well addressed in publishing or any other sector. This may be because forum creators don’t properly think through any scenarios of use. For instance: someone reads a piece online (or in print), and thinks they might want to comment on it. Assuming they are registered (and, if they are starting in print, can find the piece online) they judge the quality of the discussion and decide to post their comments. But if comments are pre-moderated, how do they know when their comment has been posted? And when they have been posted, how might they know when further comments have been posted, particularly comments referring to their contribution? If we consider Web-based discussions to be akin to conversations or public debates we can see the dynamics are all wrong, with people speaking (often only with someone’s permission) only to find their conversants responding minutes later, or when they are out of earshot.

Not least for this reason, quality of discussion has been a significant issue for Comment is Free – though it is as much of an issue for other online publications. Discussions may start well, but they tend to lose focus and collapse, with commentators engaging in personal attacks on each other or the author. In one thread to which I contributed it only took one mild post for someone to refer to me in their response as having “attention span of a Christmas tree”, while the thread degenerated into McCarthyite exposés of the author’s background – rather than rational engagement with his arguments.

Lee Bryant of the London based social-software-oriented consultancy Headshift has been following Comment is Free. He says “the main issue which is that their ‘1%ers’ are so quick, opinionated and polarised in their many views that the other 99% are largely discouraged from joining in.” Henry acknowledges that “Too many comments have nothing to do with the original post, or degenerate into back-and-forth slanging matches with others which just get in the way of reasoned argument and put off people who want to engage with the original piece” (Less is more, July 18, 2006). In response to this, a comment frequency cap was introduced, preventing individuals commenting more than once every half an hour.

It would help people reading more active posts if the authors reviewed the comments on their pieces, and linked to and responded to the key points made therein. This activity was partly hampered as result of Comment is Free breaking another Weblogging tenet: ease of use. According to the aforementioned author “to respond to comments, you have to check in all over again quite separately, as a ‘punter’ – again, you go through several technological hurdles just to respond to the comments on your own piece. So the sense of immediacy and ‘out there’-ness is non-existent “.

Clear discussion would be aided if commenters could more easily link to other comments, and if threads were closed at a preset time, with the interesting points of the discussion summarised to create a sense of closure. Henry has considered the issue of authors revisiting discussions, and notes that “more contributors are finding their way back onto their posts to answer points”, though she adds that “looking at some of comments [sic] certain individuals attract, I’m not that surprised they feel discouraged from joining in” (Open threads – what do you think?, June 30, 2006). Somewhat against the spirit of Henry’s comment, in his RSA lecture Rusbridger mused that if leading Guardian writer Polly Toynbee asking him if she should reply to the many emails she had received in response to a previous piece or write her next piece he would want her to do the latter.

Finding a way into a discussion is also hampered by the lack of visibility and profile of commenters. When people register, they are not required to use their real names, and their ‘handles’ vary from the moderately personable ‘PaulMac’ and ‘AndyV’ to ‘Rashers101’ and ‘GrunTuMolani’ix. There are no mugshots associated with commenters, and no links to profiles, personal sites or Weblogs. As a result, it is difficult to get any sense of who one is debating. Toynbee reflected on this in one post which asked “Who are you all? Why don’t you stop hiding behind your pseudonyms and tell us about yourselves?” (Civil discourse? A vain hope, May 19, 2006).

Neither do these handles link to other comments their owners have made – comments that might at least help one understand more about that particular commenter. And in the event that someone you know and trust has made a comment in a thread you are reading, there is no way you could work out if the comment from AndyV was from the AndyV you know. As well as linking commenters to their other comments, it might help potential posters judge the quality of the discussion if commenters were listed prominently on each piece.

Part of the problem is that the ‘person’ is not considered to be an ‘object’ in the Comment is Free system in the same way a story is. If people were top-level objects, and tied into a social networking tool such as LinkedIn or MySpace, many of these problems could be properly addressedx. But like many established online operations, the Guardian is tied into an existing login system that is used for its Media and other online services, and was built with other scenarios in mind.

The Comment is Free team is aware of the importance of profiles in the context of politeness. Writing on his Weblog, lead site developer Ben Hammersley noted that “[o]ne of the techniques we’re using is to display the town and country the commenting user’s ISP is considering them to be from. This, we think, calms things down a little, which is good” (Social bugs and localities, viewed 08/03/2006), though based on third party research he adds that this thesis “might well be wrong”. Again, Henry acknowledged that identity is important when she wrote “your identities are safe for now. I accept that for some there are sound reasons for not using your own names, but I doubt it’s true for most people” (Open threads – what do you think?, June 30, 2006).

The most considered (and accessible) discussion in the Guardian remains the letters pages, but oddly, if not unsurprisingly, there is no link made between the letters page and the stories to which they respond – or the comments on those stories on Comment is Free.

Learning from experience

There has been much study of the dynamics of online community, and there is considerable empirical evidence about how to foster good dynamics in this area. One of the earliest examples was the Sausalito, California-based bulletin board The Well. It flourished partly because of its limited, and professionally and geographically proximate, membership and by promoting maxims such as ‘You own your own words’xi.

Reflecting on the problem of ‘scale’ and ‘common purpose’ in online communities, Bryant says “there is simply no way of creating a single community at this scale with only semi-authenticated users”.

The equivalent of The Well today is the Weblog model, in which individual- or group-owned Weblogs link to other Weblogs and posts and use trackback to create discussion threads from distributed contributions. This encourages a higher quality of debate, as people tend not to post offensive or ill-thought-out comments on their own Weblog, where they would be prominent for days or weeks, potentially damaging their online profile and reputation.

A number of Comment is Free contributors have suggested the use of trackback to facilitate this model (see Welcome to Comment is free, March 14, 2006). However, Hammersley noted that when the Guardian‘s Newsblog still supported trackbacks “we were getting 1000+ trackback spams an hour. It kills the server, and fills the blog with porn” (comment in Pick of the week, March 24, 2006). One solution to this danger might be to only allow trackback ‘pings’ from sites registered in the profiles of Comment is Free members.

If any model of distributed contribution is to work the Guardian will have to address another issue, which is that its stories often appear at more than one URL. For instance leading Guardian writer Simon Jenkins’s article on the BBC charter renewal ‘The BBC will never cut its cloth to suit any cloak but big’ is published on the main site at,,1731048,00.html and on Comment is Free at,,1731054,00.html. This makes it difficult for Technorati, or any other tool, to aggregate references to the piece.

While these criticisms are made in the context of Comment is Free, the Guardian is in good company – which includes the Huffington Post – in the struggle to create a complementary public sphere online. What they also have in common is an apparent reluctance to build on existing research about these spaces, or to properly use design thinking to best address the scenarios of use around them.

The bigger issues

With Comment is Free the Guardian has moved smartly to address a significant development in its domain and its initial stumblings are at least understandable. At the RSA lecture Guardian Unlimited editor-in-chief Emily Bell joined the debate to argue that they had “just invented the Spinning Jenny” and were “up against people making piles of Levi’s in their bedrooms”.

However, to the extent that it sells itself on its commentariat it is not clear how Comment is Free can sidestep the continuing disaggregation of the newspaper industry that Rusbridger described. What is to stop an enterprising startup or a Yahoo! creating an opinion portal aggregating RSS feeds from the Guardian and its free-to-access competitors – as well as the key English language current affairs publications – and creating a better editorial, discussion and business model around them? Yes, readers would have to go to the free-to-access hosting site to read the original piece. But other than brand loyalty to the Guardian, and investment in one’s reputation as a Comment is Free contributor, it is hard to see what would stop commenters from moving. If the smart commenters were to move, the authors would be likely to engage with them on their new ground – and if a more author-friendly business model were devised they may move wholesale.

It is not clear that Comment is Free has yet got the commenters it wants. The reality is that the people the Guardian would really want leading the commentary on its pieces aren’t doing it at all yet. They are still writing letters, or posting on their own Weblogs – which are rendered almost invisible in the Comment is Free space.

There is also a wider issue to address. In his RSA lecture Rusbridger noted, “There is a huge explosion… of people who want to have a form of self-expression”. As other outlets for political discussion and self-expression have waned (partly a product of a more cynical attitude in the mainstream media), people have been turning to the online media as a source of identity and association. Unfortunately the quality of the relationship is low level and rather forced – hence the disappointing nature of much debate on Comment is Free and in other publications. With Comment is Free the Guardian’s ambitions appear to have been embraced by the wrong audience.

i Newspapers in the age of blogs, March 16, 2006

ii See for instance the regular Guardian contributor George Monbiot, who also includes footnotes and links to his stories

iii Guardian Unlimited editor-in-chief Emily Bell anticipated the launch in an Editor’s week piece, Confessions of a launch addict, March 11, 2006. For more on the site rationale see the About page.

iv From the CP Scott 1921 article ‘A Hundred Years

v However, my flagging of an unsuitable contribution by ‘TheMaster’ on a comment piece on the main site, Resilience, James Harkin, July 8, 2006, has not lead to it being removed.

viMy comments on broader design failure can be reviewed in ‘Publishing by Design: Time to Make Human Factors a Concern‘ Nico Macdonald, Online Journalism Review, 20 May 2004

vii There has been some experimentation in this area, and Ben Hammersley has presented a ‘tag cloud’ model on his Weblog (Weighted Tags Clouds Are So 2005) but this concept hasn’t been implemented.

viii The winner was announced in Big Blogger: We have a winner, June 23, 2006

ix See for instance comments on The flight test, Brendan O’Neill, August 3, 2006

x LinkedIn’s concept of degrees of separation could be useful in flagging comments from ‘people who are trusted by people you trust’

xiFor a considered history of The Well see ‘The Epic Saga of The Well‘ Katie Hafner, Wired magazine, Issue 5.05, May 1997

Comparing blog publishing tools: an update