Beyond SMART goals

SMART goals work fine when the work is countable. Close 40 tickets. Hit $2M. The trouble starts with the work that matters most and measures worst: improving team culture, becoming a better coach, making a codebase less fragile. Managers respond in one of two bad ways. They bolt a fake number onto the work ("increase culture score by 10%"), or they give up and write "do your best" with extra words.

The research under SMART, goal-setting theory, has a better answer that the acronym leaves out.

Match the goal type to the work

Performance goals (hit this outcome) work when the path is known and the outcome is countable. This is SMART's home turf.

Learning goals (acquire this capability) work when the work is new or the path is unknown. "Identify and test three approaches to reducing onboarding time" beats "reduce onboarding time 20%" when nobody knows yet what drives onboarding time. The evidence here is strong: performance goals on unfamiliar work actually hurt results, because people chase the number instead of learning the terrain.

Behavioral goals (do this observable thing at this cadence) work for the fuzzy, important stuff. You can't measure "be a better coach." You can absolutely observe "hold weekly one-on-ones where the report speaks more than you, and log one coaching question used in each." Behavior is measurable even when outcomes aren't.

The rewrite process

• Start with the honest version of what you want. "I want Maya ready for promotion." "I want this team to stop dreading meetings."
• Ask: is the path known and the outcome countable? If yes, write a normal performance goal and stop.
• If the work is new territory, write a learning goal: what capability or knowledge, by when, demonstrated how.
• If the outcome resists measurement, write the behaviors that produce it and put a cadence on them. Review the behaviors monthly and the outcome quarterly.

What this fixes

The fake-number problem disappears, because you stop pretending. The "do your best" problem disappears, because behaviors and learning targets are specific and checkable. And the December problem disappears, because a behavioral goal generates evidence all year. You'll know whether the one-on-ones happened. You'll have the log.