A journalist from Göteborgs-Posten came by the office in April with a draft article about us. The piece was friendly — agency growing fast, Gothenburg story, the usual frame — and she asked me to mark anything that didn’t sound right. I marked one paragraph and inserted, in my own handwriting in the margin, the line I have been circling around ever since: we are looking for self-taught nerds rather than well-educated academics. It made it into the printed piece. Four months later I keep coming back to it, because the longer I sit with it, the more I notice it is the closest thing we have to a hiring philosophy — and the more I notice it costs us something that an outsider does not see.
I want to be precise about what I mean, because the line reads worse than it is. I do not mean we reject candidates with degrees. Roughly half of the technical people on the team have one. I mean that when I am sitting across from a 23-year-old marketing graduate from a respected programme and a 19-year-old who has been running their own affiliate sites since they were fourteen, my prior is on the 19-year-old. That prior turns out to be load-bearing.
Why this is the prior
Our work — SEO, paid search, analytics, conversion optimization — has a property that I do not think most hiring managers in our field have internalised. The body of knowledge resets roughly every eighteen months. Google ships an algorithm update; the optimisation playbook from two years ago becomes counterproductive. AdWords reorganises its account structure; the agency that knew the old layout has to relearn the new one. Universal Analytics gives way to a new model and the dashboards we built six months ago need to be reasoned through from scratch.
In a field that reinvents itself this often, what I am actually hiring for is not what someone knows on the day they walk in. It is the learning loop they bring with them. Can they sit with a half-released Google feature, read the documentation that does not exist yet, run their own experiments, and figure out what is going on before our clients ask us about it? That loop is hard to teach in a classroom. It is also hard to fake. Self-taught practitioners have done it for years before we meet them — that is the only way they got good at the thing in the first place.
The signal in the room
The reason this filter works is that it has a strong signal in the interview that has nothing to do with credentials. When I ask a candidate what they have built that nobody asked them to build, the self-taught ones light up. They tell me about the WordPress plugin they wrote because their cousin’s restaurant needed a booking system. The Python script that scrapes a classifieds site for boat listings. The blog they have been running about a strange niche hobby since they were sixteen, which now ranks for half a dozen surprisingly competitive terms and pays for their phone bill. When I ask the same question of candidates whose CVs are stronger on paper, I get a polite description of a group project from their final semester.
There is a version of this question that any hiring manager can use: tell me about something you made for an audience of one — yourself — and what you learned from it. The candidate who can answer that question concretely has been running the learning loop on their own time. The candidate who cannot has been running it on someone else’s schedule. Both can be excellent in the right role; only one survives a field that resets every eighteen months without supervision.
Where the philosophy stops working
I owe a caveat here because the line in the GP article reads more absolute than the practice. The hiring philosophy works inside the technical and creative core of an agency like ours. It does not generalise to every role we have ever hired. The two functions where I have learned to override the prior are finance and legal — anywhere the cost of being wrong is regulatory, and the body of knowledge does not reset every eighteen months but instead accumulates over decades. Our bookkeeper has a degree. Our accountant has more than one. The person who reads our contracts before we sign them has been through formal training that I have not, and I am grateful for it every time.
The other place the philosophy quietly stops is when we hire for client-facing roles where credibility into the room comes partly from the credential itself. A senior strategist sitting across from a CMO at a publicly listed company is sometimes hired in part because of where they went to school. I do not love it, but I have stopped pretending it is not a factor in some conversations.
What it costs
This is the part I did not say to the journalist, because it would have made the line less quotable, and the part I have been thinking about all summer.
The self-taught pool is narrower than the credentialed pool — narrower in a few specific ways. Self-teaching as a path requires unsupervised access to a computer, electricity, an internet connection, time outside of work or school, and — most underrated — the social permission to spend hundreds of unsupervised hours on something with no obvious payoff. Those conditions are not evenly distributed. A 17-year-old building a Drupal site in their bedroom is often, statistically, from a family that already had a computer in the house and parents who did not need them earning at the supermarket after school. The candidate who learned the same skills inside a structured programme — sometimes a public one — often came from a household where the bedroom-server route was not available.
If we hire only off the self-taught signal, we end up with a team that is narrower than the talent we could be reaching. We have noticed this in our own composition. The fix is not to lower the bar on the loop signal — the loop signal is the thing that actually predicts who succeeds here. The fix is to widen the funnel before the filter applies, by going to places where the loop has been running but where we do not naturally meet it. Internship programmes that pay properly. Reaching into vocational tracks rather than only academic ones. Recruiting through people in the team whose own paths look unlike mine. We are not great at this yet. I want to be honest that the line in the GP article is what we practice, not what we will be proud of in five years if we do not also do the wider work.
What I am keeping
The line stuck because it is true about what works inside our field, and it is honest about how we hire. I would write it again. But I would write it knowing that it is a description of a filter, not a description of who deserves to do this work, and that the company has a separate job to do upstream of the filter.
In the meantime: if you are running an agency in a field where the playbook resets every eighteen months, hire for the learning loop. Ask the question about the thing they built for an audience of one. Watch what happens to their face. That is the signal. The rest is sorting.
Has anyone else been wrestling with this? I am especially curious about hires you have made where the credential pointed the other way and you were right anyway — or the other direction, where you ignored a credential and shouldn’t have. Hit me.