Measuring skill progression at an SMB

Closing Out Our Story

Two quarters in, Tom asked Maya whether any of this was working. She had a list of activities: problem ownership rotations, a decision log, a cohort of ten learning AI tools. She did not have an answer, because course completions and attendance would not show whether anyone could do something new. She needed a way to tell whether the work had changed.

‍

The measures below are our recommendations. They follow the purpose each skill type serves: essential skills help people move across roles, enabling skills help them move up, and empowering skills help them keep adapting.

‍

Maya started with a four-level scale that managers and employees could both use.

1.       Aware. Can explain the skill and has seen it done.

2.       Practicing. Has done it with help from a manager or peer.

3.       Independent. Does it without help, on real work, and the result holds up.

4.       Teaching. Others learn the skill from watching or working with them.

‍

She scored each skill from evidence in the work, not from impressions. Here is what she collected for each type.

‍

Essential skills

•  Evidence: problem write-ups, debrief notes, and peer feedback on a short list of observable behaviors.

•  Independent looks like: owns a problem from diagnosis to fix without escalating, and writes up the cause and the change.

•  Company-level signal: fewer repeat issues, shorter time to resolve, and more people who can cover another function.

‍

Enabling skills

•  Evidence: the decision log, with the call, the reasoning, and the outcome at review.

•  Independent looks like: decides within the agreed boundary, states what would change the call, and the outcomes hold up at review.

•  Company-level signal: the share of decisions made without escalation, experiments completed, and internal moves into coordination or project roles.

‍

Empowering skills

•  Evidence: tasks changed with AI tools, hours redirected, demo participation, learning hours used, and the monthly workload score.

•  Independent looks like: has changed several recurring tasks, checks AI output before using it, and shares the method with the team.

•  Company-level signal: hours redirected to higher-value work, steady participation, and workload that is not rising.

‍

The check ran quarterly. Maya took a baseline in the first month. At the end of each quarter, the employee scored themselves, the manager scored from the evidence, and the two talked through any gap of more than one level. The conversation mattered more than the score.

‍

Four cautions came out of the first round. Keep the scores separate from pay and performance ratings for the first two quarters, or people stop being candid. With a cohort of ten, one or two people can swing an average, so look at each person’s movement rather than group percentages. A low score on empowering skills may reflect a heavy workload rather than a lack of curiosity, so read the workload score beside it. And if the evidence is thin, fix the practice before blaming the person.

‍

By the end of the second quarter, Priya had moved into the customer success coordinator role, and a second agent had joined the QA automation work. Luis’s team had fewer repeat issues reaching engineering. Maya could show Tom the decision log, the changed tasks, and two internal moves, and she told him which parts she could not yet prove. These details are illustrative, and your results will differ.

‍

The measure that ties it together is internal movement. Count lateral moves, promotions, and people who took on new tasks when their roles changed. If skills are rising and nobody is moving, the career paths are not connected to the training, and that is the first thing to fix.

‍

To start at your company, pick one practice from each of the three skill types, name a manager for each participant, and put the first quarterly check on the calendar before the work begins.