Semiconductor test floor
Lots come in, run through the testers, and ship. Yield by site is current — the test data is pulled every night — and it says site 3 has slipped five points. The reason is somewhere the summaries do not reach. Channel calibration was last checked two weeks ago, so it is stale. Chuck temperature is logged on the prober but nobody pulls it, and nobody counts touchdowns on the probe card, so both are dark. The Frame draws the lines that could explain site 3 and marks which are actually measured.
Sample — illustrative mock-up
Sixty-person semiconductor test floor (sample)
The company and the numbers are invented. The shape is the one we find.
as of March 6, 2026 at 17:00 UTC
3 current 1 stale 3 dark
Two weeks back, one week ahead
Each row is one record. The band pinches when the record changes and widens as it ages. Past today it is the Frame looking ahead, and it stops where it can no longer say.
- the record changed
- top row: how far ahead the whole picture could see, day by day
- your own move, marked so the Frame doesn't read it as a trend
- already on the calendar, so it doesn't fade
- Past this line the Frame can't say: Probe is dark.
Incoming lots
1 sensors · 1 current · 0 stale · 0 dark
- Lots waiting for test current11 lots
age 1h
lot tracker
Test cells
2 sensors · 1 current · 1 stale · 0 dark
- Yield by test site currentsite 3: 89%
age 30m
STDF summaries, nightly
site 3 is slipping
- Tester channel calibration stalechecked 14 days ago
age 14d
calibration log, by hand
old enough to question
Probe & handlers
2 sensors · 0 current · 0 stale · 2 dark
- Chuck temperature by shift darkDARK
age —
logged on the prober, never pulled
the likely tie to site 3
- Probe-card wear (touchdowns) darkDARK
age —
nobody counts touchdowns
Engineering
1 sensors · 0 current · 0 stale · 1 dark
- Retest hours darkDARK
age —
buried in timecards
where the shift goes
Shipping & billing
1 sensors · 1 current · 0 stale · 0 dark
- Lots shipping late current2 lots
age 1h
shipping log
- Chuck temperature by shift feeds → Yield by test site
- Tester channel calibration feeds → Yield by test site
- Probe-card wear (touchdowns) feeds → Yield by test site
- Yield by test site feeds → Retest hours
- Retest hours feeds → Lots shipping late
a value nobody measured is null and renders dark, never zero
Start with a $400 consultation: a generous half day.
In person when we can. You leave with a Frame sketch on one page and a lot of good questions; within a week, a written proposal with deliverables and a price.
Sample — illustrative mock-up
Sixty-person semiconductor test floor (sample)
Two weeks back, one week ahead
Each row is one record. The band pinches when the record changes and widens as it ages. Past today it is the Frame looking ahead, and it stops where it can no longer say.
- the record changed
- top row: how far ahead the whole picture could see, day by day
- your own move, marked so the Frame doesn't read it as a trend
- already on the calendar, so it doesn't fade
- Past this line the Frame can't say: Probe is dark.
This sample, three ways
Sixty-person semiconductor test floor
Sample — illustrative mock-up
Sixty-person semiconductor test floor
Sixty-person semiconductor test floor
Sample — illustrative mock-up