Victor Ueda· Job Architecture Lead, Independent
Posted in Job Architecture & Careers ·
There is a lot of enthusiasm for skills-based everything, so here is what happened when we tried it. We are a professional services firm in Singapore and Malaysia, about 2,800 people. In early 2025 we introduced a taxonomy of 410 skills alongside our 19 job families, intending skills to drive development and internal moves while families and levels stayed the basis for pay.
What worked: internal mobility. Around 140 people moved across families using skills matches, roughly double the previous rate. What did not: keeping skill profiles current. Self-assessed profiles went stale within months and managers validated under 20% of them.
We have since cut the taxonomy to 120 skills. I no longer believe skills can replace job families for pay, but I'd welcome a counter-view.
3 helpful · 1 insightful · 4 replies · 211 views · 1 following
Samir DelaneyCBS, CEP
Senior Compensation Analyst, Silverline Media
No counter-view from me. We priced a handful of 'hot' skills as premiums on top of the range at a US technology company and it lasted two years. By then 60% of one family held the premium, so it had become base pay with extra administration. Families and levels for pay, skills for development, is where we ended up as well.
Compensation Consultant, Independent
I'll offer a partial one. For a narrow set of scarce, certifiable skills, premiums work, because someone other than the employee verifies them. We pay for 11 such skills in a Dutch logistics business. Anything self-assessed is hopeless for pay, I agree.
Samir DelaneyCBS, CEP
Senior Compensation Analyst, Silverline Media
On stale profiles: we stopped asking people to maintain them. Skills are inferred from project assignments and completed learning, and the employee corrects rather than creates. Accuracy is not perfect but profile coverage went from 35% to 88%.
How did you choose which 290 skills to drop?
HR Business Partner, Beacon Retail Group
Mostly by usage. Any skill that had not appeared in a vacancy, a project request or a learning plan in twelve months went, which removed about 240. The rest were near-duplicates that our family leads merged in two workshops. Inferring from assignments is next on my list.
