Process Control Parameter Optimisation · Brief 02

Yield fell 2.1 points in six hours on Line 2 · night shift

Humidity, a material lot and a late handover correction — each inside tolerance, all three at once

DataGenie Brief · autonomous
Line 2 · night shift · process historian + MES + inline vision, joined on line · lot · shift · illustrative

The one line you configure

Which lines are tracking below their seasonal yield baseline this shift, and what combination of process, material and handover explains it? Read the parameters, the incoming lot and the inspection count on the same line × shift slice.

In a nutshell

Three ordinary things arrived together. Yield on Lines 1–2 is tracking 2.1 points below seasonal baseline over the last six hours — 94.1% against 96.2% expected — and the inline vision count on the affected batches is running 38% above the line-shift baseline, concentrated at the seal station. Nothing on its own would have raised a flag: ambient humidity is 4.2 points above the seasonal norm for this production week, material lot L-211 sits 0.8 units outside the control-plan viscosity band, and the shift-handover temperature correction was applied 22 minutes later than the control-plan standard. Read together they place the loss precisely: 94% of it falls in the first 90 minutes after the handover.

At a glance

Yield rate · Lines 1–294.1%vs 96.2% seasonal−2.1pp over six hours

Vision defect count+38%vs line-shift baselineat the seal station

Loss inside the handover window94%first 90 minutesnot spread across the shift

Ambient humidity+4.2ppvs seasonal normthis production week

Material lot L-211 viscosity0.8 unitsoutside control-plan bandone lot

Handover correction lag22 minvs control-plan standardone shift

What moved

Material lot L-211viscosity profile outside the established control-plan band0.8 units

Ambient humidityabove the seasonal norm for this production week+4.2pp

Handover temperature correctionapplied later than the control-plan standard22 min

Vision defect count · affected batchesthe outcome, concentrated at the seal station+38%

Where the 2.1 points went · yield loss by window within the shift

2.1ppyield lost

First 90 minutes after handoverthe correction lag window94%

The rest of the shiftat baseline6%

The shift average is a 2.1-point miss. Ninety-four percent of it is inside a single 90-minute window, which is why a shift-level or daily read shows a soft line and an hourly read on the same slice shows a handover.

Three tolerable things, one shift · each against its own reference

Signal	This shift	Its reference	On its own
Ambient humidity	+4.2pp	seasonal norm	A humid week — nothing to escalate
Material lot L-211 · viscosity	0.8 units out	control-plan band	One lot at the edge of acceptance
Handover temperature correction	22 min late	control-plan standard	A routine delay on a busy changeover
Vision defect count · affected batches	+38%	line-shift baseline	The outcome all three produced

Every row above is inside somebody's tolerance and none of them would raise an alert alone. The finding is the coincidence — three edge-of-band conditions landing on the same line, the same lot and the same 90 minutes.

The shift read · four measures, three systems, no join written

Yield rate · Lines 1–294.1%vs 96.2% seasonaloutside

Vision defect count+38%line-shift baselineoutside

Loss in the handover window94%of the 2.1ppconcentrated

Lot L-211 viscosity0.8 unitsoutside the bandwatch

The process historian, MES yield and the inline vision counts stay at their own granularity and join on line, lot and shift. No one wrote a query that puts humidity, an incoming lot and a handover timestamp in the same sentence — that sentence is the brief.

So what

This is a one-lot, one-window problem with two cheap fixes, and it is only visible because three sources were read on the same slice. Tightening the acceptance criteria on lot L-211 and closing the handover correction gap on Lines 1–2 both cost a supervisor conversation. Left alone, the same combination recurs on every humid week that meets a lot at the edge of its viscosity band — and it will keep presenting as a soft shift with no cause.

The points that matter

No single signal was out of tolerance

Humidity, the lot viscosity and the handover lag are each inside somebody's limit. A threshold alarm on any one of them stays silent; the coincidence is the finding.

3edge-of-band conditions

The window names the cause

Ninety-four percent of the yield loss falls in the first 90 minutes after the handover, which is exactly the interval the late temperature correction left uncorrected.

94%in 90 minutes

The inspection count agrees

Inline vision on the affected batches runs 38% above its line-shift baseline and concentrates at the seal station — the same batches, independently measured.

+38%vision defects

Recommendation

Re-examine the acceptance criteria for material lot L-211, and flag the shift-handover correction gap to the Line 1–2 supervisors — 22 minutes against the control-plan standard, in the window that carries 94% of the loss. Neither is a capital item.

Questions it already answers

Is this the lot or the line?

Both, and neither alone. Lot L-211 sits 0.8 units outside the control-plan viscosity band, which the line tolerates on a normal week — this week humidity ran 4.2 points high and the handover correction went in 22 minutes late, and the three together move yield 2.1 points.

Why did nothing alarm?

Because nothing breached. Every contributor is inside its own tolerance, and a threshold alarm only fires on a breach. The band read is per line, per product and per shift, so it can see a combination that no single threshold can.

What would we change first?

The handover. It carries 94% of the loss inside one 90-minute window and it is a procedure, not a purchase. The lot acceptance criteria are the second change and the slower one, because they need the supplier in the conversation.