⌘K
ScopePeriod

OER Root Cause Analysis

Mill A · explainable AI across grading, weighbridge, GPS, SCADA and laboratory dataAll operations
Why did OER decline at Mill A? AI root-cause analysisConfidence86%

OER declined from 22.3% to 21.6% in 7 days (−0.7 pp ≈ 7.6 t CPO/day). Four drivers explain 92% of the decline — three originate upstream of the mill gate.

OER bridge — last week to today (pp)
Contribution to decline
Increased overripe FFB (6.1% → 10.6%)32% · −0.22 pp
FFB grading / CV
Longer harvest-to-mill time (17h → 26h)27% · −0.19 pp
Weighbridge + GPS
Sterilizer ST-04 pressure variation21% · −0.15 pp
SCADA / IoT
Higher fruit moisture12% · −0.08 pp
Laboratory
Other factors8% · −0.06 pp
—
AI conclusion
Crop quality and logistics (59%) outweigh the mill-internal cause, but the sterilizer is the fastest lever: ST-04 bearing wear is already predicted to fail within 11 days.

Causal graph — how plantation, logistics and assets drive OER

Learned from 3 years of grading, weighbridge, GPS, SCADA and lab data · edge width = contribution · click a node to open its evidence

UPSTREAM CONDITIONSCONTRIBUTING DRIVERSMECHANISMOUTCOME32%27%21%12%−0.7 ppRainfall 64 mmEstate B · 23–25 Sep↗Harvest round delayInterval 10 → 13 days↗Truck queue at millPeak 2.4 h · 26 Sep↗Dispatch sequencing7 of 22 trucks late↗ST-04 bearing wearVibration 6.4 mm/s↗Overripe FFB6.1% → 10.6% · 32%↗Fruit moisture+1.6 pp · 12%↗Harvest-to-mill time17 h → 26 h · 27%↗Sterilizer pressure σ0.12 → 0.38 bar · 21%↗Oil losses ↑EFB · fibre · sludgeMill A OER22.3% → 21.6%

Evidence by driver

30-day signal per driver · shaded = anomaly window from 21 Sep

Overripe FFB
6.2 % → 10.6 %
32%

Overripe share at CV grading rose from 6.1% to 10.6% after missed harvesting rounds in Estate B Div 2 and Estate E.

FFB grading / CVOpen
Harvest-to-mill time
17.0 h → 26.0 h
27%

Average harvest-to-mill time 17 h → 26 h; truck queue at weighbridge peaked at 2.4 h on 26 Sep.

Weighbridge + GPSOpen
Sterilizer pressure variation
0.13 bar σ → 0.40 bar σ
21%

ST-04 cycle pressure σ tripled since 21 Sep, coinciding with the door-drive bearing vibration anomaly.

SCADA / IoTOpen
Fruit moisture
21.9 % → 22.8 %
12%

Lab moisture +1.6 pp after 64 mm rainfall in Estate B; wet bunches raise sludge oil loss.

LaboratoryOpen

Source systems per driver

Data lineage behind every contribution

DriverSource systemsFrequencyShare
Overripe FFB
FFB grading (computer vision) at CP & receptionHarvesting round plan
Per load · real-time
98% coverage
32%
Harvest-to-mill time
Weighbridge (arrival stamp)Truck GPS telematicse-Harvest cut time
Per trip
94% matched
27%
Sterilizer pressure variation
SCADA sterilizer PLCIoT vibration & temperature (ST-04)
1 s / 10 s
99.6% uptime
21%
Fruit moisture
Mill laboratory (LIMS)Weather stations (Estate B/E)
4× per shift / hourly
100%
12%
Other factors
NIR oil-loss analyzerCMMS work orders
Hourly / event
91%
8%

Recommended investigation

Generated from the root-cause model · ranked by expected OER recovery

1
Inspect sterilizer pressure control

Check ST-04 door-drive bearing, pressure transmitter and PID tuning on Line 2; compare cycle traces with ST-01..03.

2
Analyze fruit arrival time

Reconcile cut time vs weighbridge arrival per trip for Estate B/E; identify late dispatch waves and queue build-up after 14:00.

3
Inspect FFB quality by estate

Overripe share: Estate B Div 2 12.9% · Estate E 11.8% · Estate A 7.4%. Verify harvesting round compliance and CP grading.

ACT-003 · Recover Mill A OER: sterilizer pressure + crop evacuation
Restore to 22.1% · ≈ Rp 24.59 bn/yr
Pending Approval
Contributions are model-estimated (SHAP on OER prediction model v2.4) from synthetic data. Decision support — requires validation by Mill Manager.
V-TEKI — IT & Business ConsultingDeveloped by V-TEKI · IT & Business ConsultingPrime Agri AI Intelligence Platform — prototype. All operational records are synthetic; model outputs are simulated; financial impacts are scenario estimates; recommendations require human validation. No actual company confidential data is used.