Ranked for a small or mid-sized agency without a name people already know. If you have twenty years of relationships in one sector, swap one and five.
01
Roles that have been open too long
Fastest to a first meeting, needs constant rebuilding
The signal is not that a company is hiring. Everyone is hiring. The signal is that a company has been trying and failing, and the public evidence of that is a role reposted, open past about four weeks, or pushed out across three boards at once.
A role in week one has an owner who believes they will fill it internally. The same role in week six belongs to someone who is behind on a plan and explaining it in a meeting, and for whom the fee has quietly stopped being the expensive option. Identical vacancy, completely different conversation.
Practically: watch company career pages in your niche, not only the aggregators. The role appears there first, reposts are visible there, and several aggregator feeds strip the original posting date, which is the one field this entire channel depends on.
What kills it is upkeep. This is a list that is wrong within a fortnight, so it has to be rebuilt rather than filtered, and that rebuilding is the work most agencies quietly stop doing by month three.
02
Leading with a candidate, not the agency
Highest conversion, limited by who you actually have
You have someone strong, available now, in a skill set that is scarce. You go to the companies who would want that person and you lead with the person.
This works because it inverts the ask. Everything else in a hiring manager's inbox costs them a meeting to evaluate. This costs them one CV to read, and if the person is genuinely good, reading it is not a favour you are asking for.
It is also the only asset an agency holds that no software and no in-house team can copy: a live view of who is actually looking this week.
The limit is worth stating plainly. It works with a candidate worth leading with and not otherwise. Sent about an average CV it burns the channel, because the reader learns what your recommendation is worth and does not unlearn it.
03
Following the people you placed
Free, slow to accumulate, best margin of the five
Someone you placed three years ago is now a manager somewhere else. They already know you do good work, because they were the good work.
Almost nobody runs this deliberately. It happens by accident, when a former placement happens to call, and the accidental version produces a fraction of what the deliberate one does.
The mechanics are dull: keep the list, watch for the company change, write in the week it happens. A new leader inside their first ninety days is rebuilding a team and spending a budget nobody has scrutinised yet.
It compounds and it cannot be bought, which is the same thing as saying it is worth nothing in your first year. Start it anyway.
04
Funding, expansion and new leadership
Earlier than the ad, lower hit rate
A funding round is a headcount plan with a press release attached. A new VP of Sales in month one is about to hire three people. A second site is an entire team, usually in a market where the company knows nobody.
The advantage is timing. You arrive before the role is posted, which means you are not the fifth agency to email about an ad that four others also saw this morning.
The cost is that plenty of these never turn into a role you can fill. The hit rate is lower and you have to be comfortable writing to people who are not yet in pain, which requires a different email: shorter, less certain, and explicitly early.
05
Referrals and supplier lists
Best clients, worst growth channel
Referred work closes faster, pays better and churns less. Nobody disputes this.
It is still a poor answer to "how do I find clients", because you cannot turn it on. A referral pipeline is responding to work you delivered eighteen months ago, and no amount of wanting it now changes that.
Treat it as the output of the other four rather than a channel of its own, and ask for it explicitly at the moment a placement lands, which is the one week in the whole relationship when the answer is reliably yes.