Most of the Windows machines I get asked to look at aren’t broken. They just feel wrong. Fans run more than they should. Laptop dies overnight in the bag. Boot takes forty seconds for no reason any built-in tool will tell you about. The owners aren’t going to dig through Reliability Monitor and BCDEdit. Frankly neither am I.
SystemHealthReport is a .bat. You double-click it. Sixty seconds later you have a self-contained HTML page with a grade, a ranked recommendation list, and the raw numbers behind both. The data is all stuff Windows already writes out — event log, DRIPS report, boot perf counters, last-known-good state. The work isn’t gathering it, it’s arranging it so the actionable signal is at the top instead of buried in section twelve.
v3.0 shipped a couple weeks ago. Two things moved.
Section count went from sixteen to seven. v2 was thorough in the bad way — it told you everything Windows knew about itself, useless when you’re trying to figure out what to do. v3 collapses related sections and puts the ranked recommendations above the fold, so reading only the top is enough to act.
Short-sleep-cycle filter. DRIPS logs every wake/sleep transition a modern standby laptop makes — a few hundred a day from idle radio wakes, pure noise. v3 drops cycles shorter than ~30 s before analyzing, killing about 80 % of the “your laptop is awake constantly” false alarms. The modern-standby grade went from useless to defensible in one change.
What it doesn’t do is history. The report is a snapshot — run it monthly and you eyeball trends in your head, but the tool remembers nothing. A --history flag that drops a JSON beside the HTML and diffs the last N runs would catch the slow stuff (boot time creeping up two seconds a month) that this tool should catch and doesn’t. Next on the list.
Everything else is fine. One file, sixty seconds, a graded report. The temptation to grow features on a tool like this should be resisted.