Alright, my birthday is on November 26. Back in 2005, I decided to give myself the gift of a lifetime: an engineers' community.
But the story started two days earlier, on November 24. I was thinking, "What if engineers working on interesting projects had a place to talk to each other, solve problems and share ideas?" The idea sounded fantastic.
But I didn't know anything about hosting, pointing a domain to a server or setting up forums. I had just two days, but I was determined. I didn't sleep for the next 48 hours. I registered the domain on November 24, rented a server from Yahoo Hosting and learned to install phpBB, the open source forum software that was super popular back then.
The next step: tell everyone, starting with friends and colleagues. The reaction was unexpected. They laughed. "No one is going to discuss technical stuff."
For the first couple of weeks, they were right. I doubted myself a few times, but the motivation to make the place a reality was bigger than the negativity.
I started posting ideas and questions about the fastest cars and the tallest buildings, and I was the only person reading them.
Twenty years later, CrazyEngineers has 400,000+ members. Along the way I also built a developer community at a B2B SaaS company where I led growth, and today I build Jatra, a community platform. So yes, I'm biased about platforms, and I'll tell you when that matters.
This guide is the whole path, in the order I'd follow if I had to start a community tomorrow. Each part is short. Where you need more, I'll point you to a deeper guide.
What an online community needs to work
Let's get one thing straight. A newsletter is not a community. Neither is your social media following or your support inbox. Those are audiences - people listening to you. A community starts the day one stranger answers another stranger's question, and you didn't have to ask them to.
From what I've seen, the ones that work all have four things: a clear reason to exist for a clear group of people, something useful waiting for the very first visitor, someone who shows up every single day in the beginning, and a way for new people to find it without an invite.
Most communities that die are missing the second or the third. Most that stay small are missing the fourth.
Decide why it exists and who it's for
Before you even look at a platform, get 100% clarity on two questions. Why does this community exist? Who does it exist for?
Those two answers will give you all the clarity you need. Everything else is easy to change later.
Write down your ideal member: their aims, their dreams, their goals, their pain points, what they want to achieve. Half a page is enough. Then write down what success looks like a year from now. Fewer support tickets? Product feedback? Customers learning from each other? Organic traffic? Pick what matters to you.
If you're a business, watch out for one trap. Almost every business owner I meet loves the idea of a community. But they see it as something to build once the business is already successful - or "at best" as a support channel.
Don't build the whole thing around support. AI is great at support, and I expect support-only communities to see less and less use as AI takes over. Build your community around the niche your business operates in, and support becomes one of the things it does. Our own community at Jatra is about the art and science of community building. Support and feedback are a very small part of it.
Pick a niche narrow enough to win
My advice is simple: nail the niche.
You cannot launch a generic community in 2026 or 2027. Those days are gone. Reddit and Quora already cover every broad topic, and AI answers the generic questions in seconds.
I made this mistake myself. CrazyEngineers was for every engineer in every branch, from agricultural to aerospace. Looking back, it would have grown much faster if we had focused on one niche within engineering. Honestly, if I launched CrazyEngineers today just as it was, it would fail.
So what would I do instead? I'd pick a trending topic that people care about and that needs human input. Right now that's AI, which almost every engineer has to learn. Then I'd narrow it right down. A civil engineering community about BIM and AI. An electrical one about AI in circuit design. Or a computer science one about vibe coding - maybe only vibe coding in Python. I'd widen it later, naturally.
Go narrower than feels comfortable. You can always widen a niche later. You can't make a vague community feel specific.
Decide what's public and what's private
The worst advice I keep hearing: hide your community's value behind a login.
Sure, it works at first. It feels exclusive. Then growth gets hard, because Google can't see anything, and you're stuck relying on ads, social media, events and personal invites to bring people in. That's not the way to grow a community.
Almost every community should be a hybrid. Keep the discussions about your niche public, where people and search engines can find them. Keep support and complaints private - when customers complain, they just want to be heard by a real person from the business, and they don't care whether that happens in public. Keep chit-chat and meme sections private too, so people can talk casually without filling your public site with thin pages. And anything members pay for or share in confidence stays private, obviously.
I've gone deeper into this in public vs private communities.
Choose where your community lives
There are three choices to make here. Frankly speaking, they matter more than any feature list.
Chat or discussions
Most new communities start on chat: Slack, Discord, WhatsApp, Telegram. Founders have never run a community before, so they copy their peers.
I firmly believe in structured, asynchronous discussions instead, and I know this starts arguments. In chat, a useful post lives for a few minutes, maybe a few hours. Five or ten new messages later, it's buried. Members need the right keywords to dig it up, nobody outside can find it, and the constant scroll creates a lot of FOMO.
Chat is fine as a side room. It should not be the only or the main way your members talk. If you're already on a chat app and this sounds familiar, read when to move your community off Slack, Discord or WhatsApp.
Rented or owned
You can start a Reddit community, a Facebook group or a LinkedIn group for free. But none of it is yours. A policy change can get you removed, the platform decides who sees your posts, and the member data belongs to them.
A community on your own domain is 100% yours. The content, the members, the data.
Self-hosted or managed
Not everything is rosy about running a community. You have to be constantly learning.
In early 2007, while I was enjoying an outbound with my teammates from Persistent, shit happened. CrazyEngineers got hacked. To my horror, I had no backups. Yahoo Hosting gave their best advice: delete everything and start from scratch. Nearly a year and a half of hard work was about to evaporate into thin air.
I didn't give up. I spent the next four days and nights figuring out everything on my own and fixing it. You have to be prepared to handle all of this when you self-host.
But that's not all. Years later, I had to chase developers and spend a lot of money to fix problems and build features for my self-hosted community. It was a nightmare.
So here's my take: self-host only if you have a technical team that understands running a forum and a server. Running a community and running its server are two different jobs. For almost everyone else, a managed platform makes more sense.
One more thing I learned along the way - the software decides how much of your community Google can see. vBSEO was the magical solution to all the SEO problems vBulletin had. In 2009, I installed the vBSEO plugin, and to my surprise, traffic shot up 40% in a week. Same content, same members, but our content moved higher up in Google's rankings. Most of it was on the front page of Google.
That's when I got fascinated by the power of organic growth.
When you compare platforms, check three things: public pages load without a login, every discussion gets its own clean URL that never changes, and you can export your data if you ever leave. My full checklist is in how to choose a community platform.
Set up a simple structure
Five or six categories are enough for most communities. Ten at the most. More than that just confuses people and spreads your early activity thin.
Let members choose real names or anonymity. Some people will never post under their real name, and that's fine.
And please don't spend a week writing a rule book. Written rules never do any real work. Owners love writing detailed rules - and no one ever reads them. You need two or three lines: no advertising, spamming or trolling; respect each other; speak freely on the topic, and relevant views and constructive criticism won't get you punished. That's it. I never wrote a rule book for CrazyEngineers. If you still want a starting point, the community guidelines generator keeps it short.
One rule for yourself, though: never change a URL once it's published. In 2011 we moved CrazyEngineers from vBulletin to XenForo, and our URLs changed. Traffic dropped about 40% and stayed down for six or seven months. Painful.
Seed it before you invite anyone
There's only one test that matters on launch day: the first person who joins finds something useful. That's what makes them stay and come back.
Nobody likes to post in an empty room. So you fill it first.
Go where your ideal members hang out today: subreddits, Quora, Facebook groups, other forums, your competitors' communities. Read their archives. Note how they describe their problems in their own words. Then answer those exact questions in your community - as discussions, Q&As, articles or quizzes.
Tip: Never promote your community in those places while you're researching. You'll get banned. Your job there is to listen.
I usually recommend 10 to 15 days to create about 50 pieces of content across the niche's main topics. It's not about volume. It's about depth and quality. Some communities need 10 threads and three or four articles. Some need 100 threads.
Is it OK to use a few staff accounts, or to create a few yourself? Yes. Nobody comes in the early days, and people come for the value. That value can come from anyone. Reddit's founders did exactly this; Steve Huffman has talked about the fake accounts openly.
What I'd never do is stuff a community with AI-generated posts. Add your own angles, stories, experiences and opinions. Google's spam policies go after content made at scale to manipulate rankings, whoever or whatever wrote it.
The full method, with a 30-day plan, is in how to seed a community.
Get your first members
Invite the first members personally. People you can talk to face to face or on the phone: customers, colleagues, people you've helped before.
Then watch them for about a week. Their behaviour is the feedback. If they come back, the community is creating value. If they don't, ask them why, and fix it before you invite anyone else.
From there it's a loop. Create content, invite a batch, learn from what they do, create more content, invite the next batch.
Now, what not to do. When CrazyEngineers was new, I found a hack to promote it free of cost: Yahoo chatrooms. Back in those days, Yahoo Messenger had public chatrooms. Anyone could log in and post a message. I'd find the busiest chatrooms and flood them with links to CrazyEngineers and some of our discussions.
It's not something I'd do today. I was stupid and clueless. Dropping links to your community in other forums and groups doesn't help - it actually backfires. There's a better way to launch your community.
Help people instead. After I launched Jatra, I went to Reddit, found people struggling with their communities, and helped them. I wrote about common problems, common misconceptions and what 20 years had taught me. Founders and small business owners started reaching out on their own, and that's how Jatra got its first customers.
There's more on this in getting your first 100 members on Jatra Community.
Make the first 24 hours count
A new member's first day is mostly psychological. They need to feel welcome, and that only comes from other people responding.
Someone posts an introduction? Reply "hey, welcome to the community." Someone posts their first reply? Like it, or ask a follow-up question. Within 24 hours, every time.
In the early days, message every new member personally. Look at their profile, mention something interesting, ask a simple question. At CrazyEngineers I messaged almost every new member myself. About 10–20% replied, and those people became active. I only stopped once we were getting 30 to 40 sign-ups a day and the community no longer depended on it.
Don't automate these messages. People can tell when a message is a template.
Expect the empty room
This is the part nobody prepares you for.
At the start, nobody responds to your posts. It's like speaking into a vacuum, or talking to an empty room, and very few founders are mentally prepared for it.
It took us some time to get people's attention. Google was pretty quick to index and rank our content, and we started noticing organic traffic by the end of the fourth week. Even one stranger on the site would give me a sleepless night. We crossed 100 members at six months.
The good news: you don't need my 2005 timeline. With a narrow niche, proper seeding and personal invites, it takes 10 to 15 days of systematic effort before you start seeing traction. Growth shows by day 30 to 45, and then it gets fun.
Time-wise, one hour a day is more than enough for the first 15 days of seeding. After that, put your time into inviting people, watching how they use the community, and replying fast.
The hours aren't the hard part. The silence is. Burnout happens where there is no visible success, so keep super low expectations in the initial days. If you can sustain the first 15 days to a month, you can be sure you can build a community.
I've seen people underestimate this. Jatra's first customers were solopreneurs and small businesses, and they were kind. They loved the product, but still left after about five or six months. They told me, "The platform is great, but we don't have the resources or energy to build the community."
A community needs someone to run it. Decide who that is before you launch; who should run your community walks through the options.
Turn visitors into regulars
Members follow a founder they can see is active. At CrazyEngineers people saw me sharing things and replying to their comments all the time, and they started taking note. That's what built our early content.
Back then I had a full-time job as a QA engineer. I'd walk home, have dinner, go for a long walk and then spend four or five hours on the forum, every night. Only change: it messed up my sleep schedule. But I was having fun, and when you're having fun, you don't feel the effort.
You don't need four or five hours a night. You do need to show up.
When a regular goes quiet, message them right away. "Hey, what happened? You were active. Is there anything we could do?" I've done this so many times I don't even remember. Members appreciated it, especially because it came from the owner. Proper communication can solve 90% of the problems.
Nudge the lurkers too. Some members visit all the time and never post, because they don't know what to say. Tag them in a thread they'd know something about, and you'll see them respond.
Praise members in public. We announced every new moderator and thanked people inside the community, all the time. We also ran "Know Your CEan" - member interviews over email with fun, personal questions. People love being in the limelight within their community, and readers kept nominating who they wanted to see next.
Skip the leaderboards. We tried them. The same few people always held the top spots, nobody else could ever win, and it created negativity. Let members collect trophies for their own progress instead, so they're competing with themselves.
And keep the conversations human. I've watched AI auto-replies in a very popular developer community. Because the AI answered first, members stopped replying, and a lot of the human connection was lost.
No tactic works everywhere. We once cold-emailed Steve Wozniak for an interview, and he answered every question by email. That series brought us lots of visitors and participation. Something completely different will work in your community. More content is never the answer. Experiment, and keep what works.
Moderate with judgement, not a rule book
Once strangers arrive, so do the advertisers and spammers. You can't watch everything yourself.
At CrazyEngineers I asked the most active contributors whether they'd like to moderate. I'd never met any of them. The first moderator was my sister, Ankita. Then came Ash from Malaysia, Maro from Egypt and Harshad from Surat, all early members who were "just as crazy as I was." Because they lived in different time zones, the forum was moderated around the clock while I sat in my office.
...and guess what - none of them was paid to do the job.
Moderators pick themselves. A few people always stand out through how much they contribute and how much they care. Watch them for a long time, then ask. Some will say yes, some will prefer to stay members and keep helping, and both are fine.
Moderation is personal judgement, every time. You can't have a binary rule for what's allowed. When a thread gets heated, let it run for a while - it often settles by itself. If it doesn't, ask people to calm down. Only when it crosses a line do you lock the thread and talk to each person individually. People are good by themselves. Talk to them like adults, professionally, and the vast majority will comply. If someone still won't, they don't belong in the community. It's super simple.
Spam has changed, though. In the early days spammers signed up and dropped a link. Today a lot of spam comes from real people hired in large numbers, who make a few random AI-written posts to build trust and then start adding links. Very difficult to catch. Watch new members' posts closely for their first few weeks, and let people promote what they've built only after they've created real value.
Let Google and AI find it
The first organic visitors showed up within a month. The real traffic came after about a year and a half, when everything our members had written was indexed and Google started sending us tens of thousands of visitors. New people arrived while searching for their own problems - people who were just like me, who thought like me. I never spent a single rupee or dollar on marketing. When organic growth kicks in, you get users and new content on autopilot.
It matters even more now. Since 2022, Google has shown a Discussions and forums section for searches that benefit from people's experiences, and it has said it wants to surface first-hand "hidden gems" from forums and blogs. And AI tools like ChatGPT, Perplexity and Claude cite real discussions by real people all the time.
I saw this myself on November 11, 2026. A thread I'd started on Jatra Community was the top organic result for "discourse vs xenforo," above Reddit, and it was also a source in Google's AI Overview. Results change every day, but the direction is clear: Google wants what real people say.
So keep your best discussions public and keep your URLs permanent. Move thin threads and chit-chat into private sections. Around 2019 we went lenient and let lots of thin discussions stay public on CrazyEngineers, and Google started sending us less traffic. Lesson learned. And rewrite bad titles on important threads. Members will write "help, please." Change it to the actual question. In my experience nobody minds, and they write better titles next time.
Two communities on Jatra show what this looks like. ArtofBlockchain grew from zero to 1,000+ members in about a year and a half, organically. DMTWorld spent a long time on Discord, with members but hardly any discussions and no organic growth - the community felt dead. After moving to Jatra it was opened to Google, and new people now find it through Google, ChatGPT, Perplexity and Claude before they even join.
If your community already exists and Google is ignoring it, start with why your community isn't ranking on Google. In my 30-60-90 day launch plan, I called SEO our number one growth hack. I still stand by that.
Measure the numbers that matter
The biggest vanity metric is the member count. It feels good and it looks good to management, but it holds no value at all. It's just the number of records in the database.
What you track depends on the stage. In the first months, watch whether people come back, the posts-per-member ratio and, if you're public, visitors. Once you're growing, add retention (logins per week or month), time spent, which content people engage with, likes per member. And track traffic, search impressions and clicks all the time. When those grow, you can be sure people find it useful.
The number I trust above all is returning members. Get about 10% of your registered members coming back and you're building a very useful resource. Getting them to post and like is the next problem - and it's a solvable one.
Watch for the first signs of decline too. The clearest one: threads stop getting a reply the same day. Tagging members doesn't bring them back, and when you talk to a few regulars, there's little enthusiasm, or maybe fake enthusiasm. Find out why before it gets worse.
Jatra puts returning members and engagement together into one community health score, so you don't have to work it out by hand.
Making money from a community
A lot of people want to know how to build a community that people will pay for. Build the value first. Money follows value.
CrazyEngineers made nothing for its first couple of years. Then in 2008, out of the blue, an Australian university got in touch to advertise. Around August, Google sent our first AdSense cheque - about $110. That was the trigger: this site can make money! Every advertiser after that, from Siemens to Texas Instruments, came to us. We never did outbound sales.
But here's the lesson that cost me the most. By 2017, one advertising network brought in 75–80% of our revenue. Then it stopped counting impressions from India, and our revenue fell almost 70%. Never let one partner bring in most of your money.
Your members will tell you what they'd pay for, if you listen. Around 2014, lots of CrazyEngineers members were uploading their resumes and asking for reviews. That's a paid service, waiting to be built.
Building a community for your business
Everything above applies to a business community. A few things are different. For the full SaaS version, read how to build a B2B SaaS customer community. Building for developers? See how to build a developer community. Running a Web3 project? Read how to build a Web3 community. Maintaining an open source project? Read how to build an open source community.
The buyer isn't the one posting. In the SaaS communities I've run, the paying customer was usually a CXO-level person, and the person actually in the community was someone from their team.
Someone has to own it. Inside a company that's usually marketing, or product. In a small company the founders and top executives should show up too, even if it's just 5–10 minutes a day: reacting to posts, replying, sharing their views.
Employees will help once there's traction. Nobody really cares when there is no traction. Once it's growing, tag them in questions they can answer, and they won't hesitate. You shouldn't have to push them.
Weave the community into the product. Gumlet's developer community launched on Circle in June 2022 and, in the company's own words, got crickets for months. It turned around after the move to Discourse, once the community came first across the website, the product dashboard and support emails, and employees committed to answering questions promptly. It grew to about 2,200 visitors and about 60 sign-ups a month. The post's closing lesson stuck with me: it's much easier to rev up a slow engine than to start a stopped one.
For the business side in detail, start with what a customer community is, check whether your business is ready, and then follow the 90-day B2B SaaS launch plan.
Frequently asked questions
How long does it take to build an online community?
With a narrow niche and proper seeding, you'll see the first traction after 10 to 15 days of steady work, and growth by day 30 to 45. Big search traffic takes longer. At CrazyEngineers it took about a year and a half to two years before Google became our biggest source of new members.
How much time does a community need each week?
About an hour a day for the first 15 days of seeding. After that it depends on the niche: some communities do fine on an hour a week, others need 5–10 hours. As it grows, step back, until your own posts are only 1–2% of the total.
How many members do you need to start?
A handful. Start with people you can talk to personally. Ten members who come back every week are worth more than a thousand who signed up once.
Should I start on Discord, Slack or a Facebook group?
Only if you're fine with your best conversations getting buried in the scroll and staying invisible to search. They're free and familiar, but you don't own them. If you want a community that keeps growing on its own, start on an open platform you control.
Is it OK to seed a community with your own posts?
Yes. Nobody joins an empty room. Seed with your own account, staff accounts or a few accounts you create yourself, and make sure the answers are genuinely useful. Members care about the answer, not who typed it.
Do I need community guidelines before launch?
Two or three lines. No spam or advertising, respect each other, stay on topic. A long rule book won't change anyone's behaviour. You can't define a culture with a rule book - people learn it by being there.
Where I'd start tomorrow
If I were starting a community tomorrow, this is the order:
- Write down why it exists, who it's for, and what success looks like in a year.
- Pick one narrow niche.
- Choose an open platform with public discussions and permanent URLs.
- Set up five or six categories and two lines of rules.
- Spend 10 to 15 days creating about 50 useful pieces of content.
- Invite a small batch of people personally, and message every one of them.
- Watch who comes back, learn, and invite the next batch.
- Show up every day, and keep your expectations super low for the first month.
That's it. No magic trick. The communities that make it are the ones whose owners don't give up in the quiet weeks.
If you'd like help with those first months, Jatra's managed plans include setup, content research and strategy, and on the higher plan we seed the community with you. Book a call and tell us what you're building.
