Two weeks ago the press log credited someone with a sixteen-hour breakdown. So I made some updates to the logs. You can check out the earlier development here, here and here. The tablet went to sleep overnight, its software timers stopped with it, and the check-in popup and auto-pause slept right through, while the clock kept counting. °FAI: Screen printing setup times.

The clock

Now Setups prompt at 220 minutes, just above the longest real setup on record. Runs prompt at 8 hours. If nobody answers, the timer stops, flags the entry, and the overnight hours get cut back to reasonable numbers instead of counted as work. Every legitimate long job still passes untouched. Having weeks and weeks of data allows me to set those real limits on time frames for the log.

The target

With runaways out, one number means something: setup minutes divided by screens = setup per screen. Now, the setup per screen target is on the tablet banner and the weekly report.

The target starts where the crew already is, the current weekly average, and the pace a proven operator already holds. An achievable spot. A target nobody’s ever hit isn’t a target.

Then it slowly improves. Hold the number for a full week, and it drops one minute the next. It never tightens until the report confirms the last level held, one heroic week doesn’t get to set the bar.

The banner shows the live weekly average against the target, so the crew can self-correct mid-week.

Where does mins per screen land? Where mistakes start. Faster setups that come back mis-registered aren’t faster. And when the week’s load is heavy, you don’t tighten the screw on a crew that’s already buried. Hold the line, let them work, go back to reducing when it feels right.

What this really shows: an honest clock makes an honest average, and an honest average makes a target the crew can actually reach, and keep reaching. So far, that’s a 24% improvement in setup times, from 12 weeks of collecting clean data, including most recently 3 weeks of set targeted times. (5 months since the project started and we converted handwritten logs to this digital/AI driven method).

Tools used:

• Google Apps Script (doGet endpoint, live progress banner, report and validation logic)
• Google Sheets (dark room logs, weekly/monthly/quarterly reports, TrendReports script)
• Claude (script writing, order-number normalization, form validation rules)
• GitHub (log for the tablets)