What it is

how many standard deviations a value sits above (or below) the mean of its group. It puts different-scale numbers on one common ruler ("standard deviations"), so you can compare or average them fairly.

So z = +1 means "1 std above average", z = 0 means "exactly average", z = −2 means "2 std below".

Tiny example

Value across 3 configs: [60, 62, 64]. mean = 62, std ≈ 1.63.

• 60 → z = (60−62)/1.63 = −1.23
• 62 → z = 0.00
• 64 → z = +1.23

When to use it generally:

• Combining/averaging metrics on different scales (exactly our case).
• Ranking items across mixed units.
• Flagging outliers (|z| > 2-3 = unusual).
• When not to: if the group is tiny (std is noisy) or wildly skewed — z assumes a roughly symmetric spread. With only 9 rows ours is indicative, not statistically rigorous (hence "metrics-only, no judge" caveat).

References