My dashboard said 18 videos never went out. Every one of them had.

I post short videos every day, and almost none of the posting is done by hand. I film, a system cuts the footage, I approve the cuts on one review page, and a scheduling service publishes them on a calendar. That review page is where I check whether the week is on track. One evening in August it told me something that was not true.

I had filmed new footage that weekend. The page showed four of those cuts still waiting to go out, along with 14 others. So I asked why my new videos had not posted.

They had. All 18 had already published. The page was the thing that was wrong.

Here is how that happens. The scheduling service knows what actually went out. My review page keeps its own record of what is waiting. Once a night, a job checks one against the other and marks the published ones as published. That night the check never ran. Another pass was still busy when it was due, so it stood down, which is the correct thing to do. The problem was what came next: nothing. The only job allowed to do the check was the nightly one, so a skipped night was not a late check. It was no check for a full day.

That is the quiet version of a broken system. Nothing crashed. Nothing sent an error. The videos went out on time. The only failure was a screen telling me something false, and I believed it, because that is the screen I built to believe.

The fix was to stop counting on one scheduled moment. The system now notes when the check last finished, not when it last started. If it has been more than 20 hours, whatever job runs next does the check itself. It does not matter why the last one was missed. Busy, crashed, the computer went to sleep. If it did not finish, the next pass picks it up. The most recent one finished last night.

If you look at a dashboard to decide what to do next, ask one question about it. When the thing that keeps it current does not run, what happens? If the answer is “it waits until tomorrow,” you will spend that day acting on yesterday.