We build a tool that maps a CV onto the European skills taxonomy and looks for roles you have the skills for but have never thought about. Last week we tested it on fifteen occupations most people can name — nurse, accountant, teacher, chef, electrician — and got the same answer over and over.
Accountant → accounting manager. Librarian → library manager. Secondary school teacher → deputy head teacher. Chef → restaurant manager.
Forty-eight per cent of the results were managerial roles, presented under a heading that said careers you might not have considered. Four of the fifteen occupations had no other option at all.
That is not a discovery. That is a promotion, and everybody has already considered it.
It was not a ranking problem
The obvious explanation is that generic skills dominate the overlap. Every manager needs to adhere to organisational guidelines and accept own accountability, and those appear in hundreds of occupations — so managers look similar to everyone.
We tried the obvious fix: weight each shared skill by how rare it is, so that ubiquitous skills stop deciding the ranking. It is a correction we had already applied twice elsewhere in the product, and it works well.
Here it made things worse.
What an accountant shares with an accounting manager is not accept own accountability. It is accounting entries, accounting department processes, accounting techniques — skills so specific that almost no other occupation requires them. Weighting by rarity gave those enormous weight and pushed the accounting manager further up, not down.
The hypothesis was wrong, and measuring it took ten minutes. Worth writing down, because it is the kind of fix that sounds right in a meeting.
It was arithmetic
ESCO, the European Commission's occupation taxonomy, classifies its 3,039 occupations using ISCO-08. ISCO's major group — the first digit — is usually described as what field you are in: 2 is professionals, 3 is technicians, 7 is crafts and trades.
But ISCO's first digit encodes something else at the same time: skill level, meaning roughly how much qualification the role demands. And that is where it breaks.
| ISCO major group | skill level | occupations |
|---|---|---|
| 1 Managers | 4 | 350 |
| 2 Professionals | 4 | 868 |
| 3 Technicians and associate professionals | 3 | 646 |
| 4–8 Clerical, service, trades, plant operators | 2 | 1,078 |
| 9 Elementary occupations | 1 | 76 |
Only two major groups sit at skill level 4: managers and professionals.
Our rule said a suggestion must be in a different field and never below your level of qualification — because a career you have not considered should be a door that opens, not one that lowers. Sound rule. But apply it to any of the 868 professionals — every teacher, nurse, accountant, engineer, journalist, pharmacist, architect — and “different major group, level 4 or above” leaves exactly one possibility in the entire taxonomy: group 1, managers.
Not a preference. Not a weighting. The only remaining box.
The 350 managers are in the mirror image of the same trap, which can only offer them professionals. Between them, 1,218 of 3,039 occupations — 40% of the taxonomy — could not structurally receive any other answer.
The fix is one digit
ISCO does encode field, properly, at the second digit:
21 science and engineering 22 health 23 teaching 24 business and administration 25 information and communications technology 26 legal, social and cultural
These are real fields, all at the same level, all different from one another. Measuring distance there instead of at the first digit costs one character in a left() call, and the answers change completely:
| occupation | before | after |
|---|---|---|
| nurse | child day care centre manager | clinical psychologist |
| secondary school teacher | deputy head teacher | corporate trainer |
| electrician | electrical supervisor | heating and ventilation service installer |
| bartender | — | travel agent |
| photographer | — | video artist |
A nurse and a clinical psychologist share twenty-seven essential skills and sit in different fields at the same level of qualification. That relationship was always in the data. It was unreachable by construction.
We also had to change how we recognise a promotion in disguise. We were comparing job titles word by word, which catches secondary school teacher → secondary school head teacher but sails straight past accountant → accounting manager: those two share no word at all. Comparing the first four characters of each word catches both, and librarian → library manager with them.
Managerial results dropped from 48% to 32% on the well-known occupations, and from 27% to 22% across a random sample of 150. Coverage went the right way too: from 11 of 15 occupations with a real alternative, to 15 of 15.
What we take from this
Thirty-two per cent is not zero. Some of what remains is legitimate — a chef becoming a restaurant manager is a genuine change of function, not a rank. Some of it we have not solved yet.
The wider point is about classifications. ISCO-08 is the backbone of ESCO, which is the backbone of a great deal of European labour-market tooling. Its first digit blends what you do with how qualified you are, and that blend is invisible until you build something on top of it and watch the output come back wrong. It is not a flaw in ISCO — the standard is explicit about the two dimensions. It is a flaw in reading the standard too quickly, and we read it too quickly.
If you use a tool that suggests career moves, it is worth asking it about a nurse, a teacher and an accountant, and seeing how many times it says manager.