Machines and humans need to be goaled differently. Depending on your background that will either be super obvious or previously unconsidered.
I love setting the right goals. When you are running s product and releasing a new feature it is exciting to dial in the right goal. What metric should this new feature move? How long will it take to reach significance? How much change should be considered a win? It's an experiment: You press go and at the end of some time-frame, you have a lot to consider.
Setting a goal for a machine is a maximization problem: What is the most precise and independent metric I can find?
Setting a goal for a human is far more artful. It requires empathy and psychology. It requires trade offs.
The core difference between a machine goal and a human goal is motivation: Setting the right goal for a machine will not impact performance The machine will not be compelled to work harder or smarter due to the PM picking the right goal, but it is precisely the motivation of humans that makes the goal setting process so different.
Like in machine goals, when setting a human goal you will want any goal to have a degree of independence from external factors as well as a clear dependence on the subject or variable (feature or employee).
What is different is that humans require clarity and communication of progress. A machine doesn't have to understand the goal you are measuring; A human does. A machine doesn't care if you think about success in terms of numbers or percents; a human does care as they do not intuitively understand the increasing inelasticity of percents as results accumulate.
So with all of that, how should goals be set for humans?
- Independent: A goal should be independent of outside factors
- Dependent: A goal should be strongly dependent on the humans you are measuring and it should be obvious what tactics move the goal.
- Cumulative: A goal should linearly grow over time. This means raw numbers are better than percents.
- Intuitive: A goal should be simple and obvious as to what it is measuring.
No goal is perfect, but like in mathematics, you realize you have a great goal - or even a persistent goal - when it is elegant. Don't overly focus on the negative incentives of a goal - they will always exist. Rather, communicate clearly what is unacceptable and put checks in place to enforce rules. If you can't check all of the above boxes that's fine. This is early days but know that some day the team you are managing will have one or two elegantly crafted goals. It's an iterative process to get there.
And yea, 1 or 2 goals is ideal. If your team needs 4 or more goals, they are probably doing two or more jobs. You'll need to fix that.