If you under-eval, your team misses out on growth opportunities, and your comp and levels won't be fair or meaningful.
Well.. think about the differences between factories and artisans.
You gather stats on # of pairs manufactured per day, how long it takes to make each pair, uniformity of the components, quality of the finished product.
* Did they design the shoes, or are they replicating a design?
* How artistically appealing are they? How faithful is the result?
* Did they have to work around any missing or suboptimal components?
* Is the design optimized for aesthetics, comfort, or endurance?
* Are they working on a compressed deadline?
* Does that person have any medical or orthotic needs? Does the cobbler understand and have the necessary materials to address those needs?
* Is there an apprentice? Does the cobbler effectively use their help? Are they happy?
* Does the cobbler's aesthetic judgment resemble your own?
* How long are the shoes expected to last, and under what conditions?
... and so forth.
Shoes are simple enough that we can stamp them out with factories and they are mostly good enough, but software is not. Software and systems engineers are firmly in the realm of artisan labor.
* degree of difficulty
* prescribed timeframe
* the support from other teams, or lack thereof
* her familiarity with the building blocks
* any changing conditions
* dependencies & design constraints
This is also why managers have a responsibility to educate themselves about their biases.