Product update

One role is rarely one search

A job description usually describes several genuinely different candidates. Each becomes its own persona, sourced as its own search, with its own filters and its own message.

All posts
EL
The EmployLabs team
· 4 min read
In short

A single job description usually describes several genuinely different kinds of candidate, who do not live in the same place or answer the same message. EmployLabs carves the role into personas and sources each as its own search with its own filters, instead of one averaged search that sits between two real groups and reaches neither well.

One job description, several different people

Read almost any job description carefully and you will find that it describes more than one kind of candidate. Not variations on a theme, but genuinely different people, who have had different careers, who describe their work in different words, and who would each be a perfectly good hire for reasons that have very little in common.

This is not sloppy writing. It is what happens when a team writes down the outcome they want rather than the biography they expect, which is the right way round. The difficulty appears one step later, at the point where the description has to become a search, because those different people do not live in the same place and do not answer the same message.

What a single averaged search does

The usual response is to write one search that tries to cover everything the description mentions. The filters end up as a compromise: a title list broad enough to catch both groups, a seniority range wide enough to include both, keywords chosen because they are the ones the two halves share.

The result sits in the gap between two real groups and reaches neither of them well. Each group is under-represented, because the filters were loosened to admit the other. The message is pitched at nobody in particular, because it had to work for both. And the response rate looks like weak demand for the role, when what it actually shows is a search that was aimed halfway between two targets.

A search aimed halfway between two real groups reaches neither. The averaging happens in the filters and shows up in the replies.

So the role is carved into personas

On EmployLabs a role is broken into personas first, and each persona is then sourced as its own search, with its own filters. Not one search with a wider net, but several searches, each one narrow enough to be about somebody in particular.

Because each persona is a separate search, each one can be specific in the way that actually finds people. The titles can differ. The seniority signals can differ. The companies worth looking at can differ entirely. Nothing has to be loosened to accommodate a group it was never about.

A worked example: one backend role

Take a single backend engineering role. One persona is a platform engineer already working at a product company: they have owned a service in production, the vocabulary on their profile is the product world’s vocabulary, and the move you are offering them is a sideways one between similar environments.

The second persona is an engineer at a services firm who is ready to make the switch. Strong, often very experienced, and describing that experience in terms of clients and projects rather than products and ownership. The search terms are different. The seniority signals are different, because a title means something different on each side of that line.

And the message that lands for one does not land for the other. The platform engineer is being asked to consider a comparable job at a different company, and wants to know what is interesting about the system. The services engineer is being asked to make a change of direction, and wants to know that the change is real and that the move is credible. Send either message to the other person and it reads as though it was meant for somebody else, because it was.

You see the personas before anything is spent

The personas are shown to you before any sourcing spend happens, which is the point at which they are still cheap to change. They are a plan you approve rather than a decision taken on your behalf and discovered afterwards.

  • Drop one. If a persona is not a hire your team would actually make, it should not be consuming any part of the search.
  • Add one. Hiring managers usually recognise a missing group instantly, even when they could not have described it in the brief.
  • Rebalance them. How much of the search goes to each persona is yours to set, so a long-shot group can be explored without taking the search over.

How to use it

  1. Read the personas as a description of who you are about to contact. If one of them does not sound like a person you would interview, that is the moment to say so.
  2. Show them to the hiring manager. A list of two or three recognisable kinds of candidate is a far better prompt than a job description, and they will often correct it in a sentence.
  3. Revisit the balance when replies start arriving. If one persona is answering and another is not, that is information about the market, and the weighting can move to follow it.

Common questions

Why split one role into several searches?
Because one job description usually describes several different kinds of candidate, and they do not sit in the same places or respond to the same message. A single averaged search compromises its filters to cover both groups and ends up reaching neither of them well.
What does a persona actually change about the search?
Each persona is sourced as its own search with its own filters, so the titles, the seniority signals and the companies worth looking at can all differ, and the outreach can be written for that group rather than pitched between two of them.
Can I change the personas before any money is spent?
Yes. You see the personas before anything is spent on sourcing, and you can drop one, add one, or rebalance how much of the search goes to each. Hiring managers tend to recognise a missing group immediately.

Related reading

All posts →

See it on one of your own roles

Upload a job description and watch the pipeline run before you commit to anything.

Start for free