You have somehow re-defined software engineer performance as a mere function of company bottom-line, and ended up measuring performance of individual software developers by proxy of long-term company performance. There are two problems with it.

First, not all software engineers actually do commercially work on commercial software -- as a quick example, Con Kolivas, who is an anaesthesiologist by day and kernel hacker in spare time. Others may work on commerical software by themselves, without any external team or management involved; think Colin Percival. I would even wager a guess that some of the best software was created in a non-commercial way, say, Linux kernel -- at least the early versions of it.

Second, the impact any software engineer has on company bottom line is much affected by other aspects of a company. In a symbolic notation, you could say:

  BOTTOMLINE_IMPACT := RISK_MANAGEMENT(SALES(TEAM_MANAGEMENT(THE_SOFTWARE_DEVELOPER(EXISTING_CODEBASE)))).

Invoking the Amdahl's law -- even if you had a 10x SOFTWARE_DEVELOPER, performance of the whole system will be limited by the other factors. In my limited experience, the RISK_MANAGEMENT (of software development and use) part is the least understood one those days.