Task Manager's Startup tab shows an impact column that looks like an opinion. It isn't: Windows measured the process during an actual boot and bucketed it.
The measuring window
During boot and sign-in, Windows Diagnostics-Performance instrumentation tracks each startup process's CPU time and disk I/O — continuing briefly after the desktop appears (the busy tail of boot). Call the thresholds:
High impact
> 1 second of CPU time or > 3 MB of disk I/O in the boot window
Medium
lower-band positives: ~0.3–1 s CPU or ~1–3 MB I/O
Low
below those bands; Not measured = no completed observation yet
The thresholds are deliberately unglamorous. 1 CPU-second at sign-in is a user-perceptible delay to a responsive desktop — exactly why it's the cut-off.
What to do with it
- High + you use the app daily: keep it enabled, accept the tax; disabling saves boot time you pay back launching it manually.
- High + rarely used: disable via the approval blob, not by deleting anything.
- “Not measured” forever: the entry never actually ran during a monitored boot — check that the command path exists at all.
The number answers “does this cost me boot seconds,” not “is this app good.” A backup agent with High impact may be exactly what you want. The discipline is deciding per entry — not bulk-disabling everything High.
ProofTune ProjectEngineering notes — every claim here names the bytes a real tool touches. Verify first, install second.