When it breaks, find out why

By the end of this track you will have

1. Step 1One symptom sheet, saved before your first click: what is not happening, counted, since when, what changed, and what fixed will look like.
2. Step 2The fault pinned to one link in the chain, a before-number and an after-number measured the same way, and a note of every test that got you there.
3. Step 3Your diagnosis written as a claim, the one check most likely to break it, what that check found, and a write-up carrying exactly one caveat.
4. Step 4A failure log in a shared spreadsheet, four rows in it, and two rules written under it: search it before any theory, and add a row whenever a failure costs more than fifteen minutes.
5. Step 5A block note for the last thing that stopped you: the exact message, what kind of block it is, the routes you already hold, the one permission to ask for, and the shortcuts you will never take.

The steps

Each one ends with a check you run yourself. The box is yours to tick when it passes.

1. 1
   Step 1.Write the symptom down before you touch anything

   30 minutesOpen to everyone

   Done means

   Put the time the sheet was saved beside the time of the first change anyone made while fixing it. Take that time from the broken system's change history, or ask whoever made the change. The sheet comes first. The symptom line carries a count and names the screen it came from. The success line names a number and the figure it must reach. If any of those three is missing, you have notes, not a sheet.

   The check passed
2. 2
   Step 2.Narrow it by halves, with a number before and after

   45 minutes, plus one real run to measure again

   Done means

   You can read out a before-number and an after-number. Both came from the same screen, by the same method. They differ in the direction the fix promised. No test line in your note names more than one thing that changed. If the after-number came from anywhere else, it is a different number, and the fix is not proved.

   The check passed
3. 3
   Step 3.Try once to prove your own answer wrong

   25 minutes

   Done means

   Read your Expected line against your Found line. Had your answer been wrong, the check would have come back differently, and you can say how. Your write-up carries exactly one caveat. Your Leans on it line names at least one thing, or names where you searched and found nothing. If the Expected line would read the same whether you were right or wrong, you ran a confirmation. Pick again.

   The check passed
4. 4
   Step 4.Keep a failure log of symptom, cause and move

   30 minutes

   Done means

   Type a word or two a colleague would use about your newest row into the spreadsheet's Find box. Words like declined, missing or not sending. The row comes back. Then hand the log to someone who was not involved, with that symptom said their way. They find the row and read you the move without asking a question. If they need you to explain any column, rewrite that row until they do not.

   The check passed
5. 5
   Step 5.Blocked is not broken

   30 minutes

   Done means

   The job is done. A fresh screenshot of the security settings and the user list matches the one from before, apart from the single permission you asked for, if any. Two-step sign-in is on everywhere it was on. Nobody's password was used by anyone but its owner. If the screenshots differ in any other way, you went through the wall.

   The check passed