I also have a definition of what I think level should ideally represent: the marginal contribution of a person’s influence on the outcome of a company, relative to the counterfactual situation where that person never worked there, adjusted for a specific threshold of risk tolerance.
In other words, you can consider two hypothetical futures for a company: one with a specific person and one without that person. You then have a probability distribution defined over the difference in outcomes. For someone at a very high level, the absolute area under this curve is large—they have a big impact on the company (whether positive or negative). For someone at a lower level, their impact is small.
Someone who is good at their job should hopefully lead to positive impact, but you can certainly adjust this for risk. Perhaps you bring in a CEO who has a 90% chance of tripling the company’s revenue/growth and a 10% chance of leading the company to failure. That might be acceptable for a hypergrowth startup. For a larger and more mature company, you might have a different risk profile.