Most freelance work comes from three channels: people who already know you, communities where your clients discuss their problems, and direct outreach to specific businesses. Job boards and marketplaces do far less than their reputation suggests, and the effort they absorb is the reason people conclude that freelancing does not work.
Here is each channel, roughly ranked by how much work it produces per hour spent.
1. Referrals and past clients
The highest-yield channel by a distance, and the one people treat as passive.
It is not passive. Referrals happen when somebody remembers what you do at the exact moment a third person mentions a problem. Your job is to make that recall easy and to give it opportunities:
- Finish well and hand over cleanly. The last impression is the one that gets described to someone else.
- Ask, directly, about a month after a project ends. Almost nobody asks, which is why it works.
- Be specific about what you do. "Full-stack developer" is unrememberable. "I build Shopify integrations for wholesalers" gets passed on.
Past clients are the same channel. The cheapest work available is the second project for someone who already paid you once and had a good time.
2. Direct outreach that is not cold
Contacting businesses directly works, and it fails for a predictable reason: most people send a template.
The version that works costs you twenty minutes per message. You use the product, find something specific and real, and lead with that rather than with yourself. The whole message is about their thing, and the offer is almost an afterthought. Ten of those beat a hundred templates, and you can send ten a week alongside real work.
It also has a property the other channels do not: you choose the clients. Everything else is inbound, which means you get whoever arrives.
3. Communities where your clients are
Not developer communities. Your clients are not in the developer room.
If you build for restaurants, be where restaurant owners complain about their systems. If you build Shopify apps, be where store owners are. Answer questions fully and for free, with no pitch attached. Some of the time nothing happens. Some of the time the person messages you, and occasionally someone reading who never posted messages you six months later.
This is slower than outreach and compounds better. Treat it as a background process rather than a campaign.
4. Your own work, findable under your name
Not a channel that generates leads on its own, but the thing that makes every other channel convert.
When someone is told "talk to Ana, she built that scheduling app", the first thing they do is search your name. What they find decides whether the referral turns into a conversation. A page with three finished projects, live links and a clear description of what you do turns a warm introduction into a booked call. Nothing findable turns it into silence.
This is most of what a portfolio is for. It does not go and find people. It closes the ones the other channels send.
5. Job boards and marketplaces
Last, and this is the honest part.
They do produce work. What they mostly produce is competition on price, because the format puts twenty proposals in front of one buyer and the only visible difference is the number. The people doing well on them usually got there by specialising narrowly enough that they are not one of twenty.
The specific trap is effort accounting. Writing proposals feels like working. It is measurable, it fills a morning, and it produces a sense of progress that outreach and community presence do not. Then a month goes by. If you are going to use boards, cap the time, and spend the hours you save on channels one to three.
What this looks like as a week
If you are starting from nothing, the mix that works is roughly:
| Channel | Time per week | What it produces |
|---|---|---|
| Referral asks and past clients | 1 hour | The most work, once you have any history |
| Direct outreach, ten researched messages | 3 to 4 hours | Slow to start, then steady, and you pick the clients |
| One community, used properly | 2 hours | Compounds over months, invisible for the first few |
| Keeping your work findable | Occasional | Converts everything above |
| Job boards | 1 hour, capped | Occasional work, mostly price competition |
The failure mode is not picking the wrong channel. It is doing one of them half-heartedly for two weeks, concluding it does not work, and switching. All of these are slow to start and none of them are slow forever.
The uncomfortable part
None of this is about being a better developer. It is about being findable, being specific, and asking. Plenty of excellent developers are underemployed because they are none of those three, and plenty of average ones are booked out because they are all of them.
If you have nothing to point people at yet, start there: what to put in a portfolio when you have no clients. If you have the work but no first client, how to get your first freelance client with no network is the narrower version of this page.