The number moved, and both sides could audit it

Every engagement starts with a written baseline captured in shadow mode. These are the outcomes design partners are working toward, with the wedge, the method and the measurement stated plainly.

Baselines agreed before autonomySame immutable log both sides read

Programs running across these segments

PASSENGER RADIALTRUCK & BUSOFF-ROAD & OTRAIRCRAFTSPECIALTYTECHNICAL RUBBERPASSENGER RADIALTRUCK & BUSOFF-ROAD & OTRAIRCRAFTSPECIALTYTECHNICAL RUBBER

Six engagements, six wedges

Each one started on a single cell with one metric.

Splice defects at a passenger radial plant

The most common first wedge, and the clearest illustration of why station-level measurement beats downstream sorting.

Audit sampling and a four-hour feedback loop

Splice quality was verified by periodic audit and confirmed at final inspection. By the time a drift was visible, hours of production carried the same fault, and root cause was reconstructed from memory and paperwork.

  • Sampled verification, not per-component
  • Feedback measured in hours, not seconds
  • Root cause reconstructed manually after the fact
before treadum · shift report
shift B
splice rejects 41
detected at final inspection
mean detection lag 3 h 52 m
root cause "suspected drum 2" (unconfirmed)
corrective action supervisor note, next shift

Per-component measurement with correction in the loop

Treadum measured overlap and gap on every component, scored open-splice risk before the tire left the station, and corrected drum tension and placement offset inside the signed envelope. Faults that remained were held at the drum with evidence attached.

  • Every component measured, none sampled
  • Correction applied within the same build cycle
  • Findings localized to builder, drum, roll and batch
  • Overrides captured as structured training data

How the measurement was done

The method matters more than the number, because the method is what makes the number believable.

Days 1–10

Baseline

Thirty days of production captured across all shifts, constructions and compounds, with no Treadum writes and no process changes.

Days 11–30

Shadow

Predictions compared against engineer judgement and downstream inspection results; accuracy and false-hold gates measured.

Days 31–55

Assist

Engineers committed every proposal. The wedge metric began moving and override reasons were coded.

Days 56–84

Bounded

Low-risk corrections executed automatically inside the signed envelope, with weekly gate review.

Ongoing

Attribution

Effects attributed per agent and per action class from the same immutable log both parties read.

Wedge outcomes at a glance

Wedge outcomes at a glance
EngagementWedge metricBaselineAfter 12 weeksStatus
Passenger radial · line BSplice defect rate0.50%0.31%[ASPIRATIONAL]
Truck & bus · press bank 3Cure energy per tireIndex 100Index 93.9[ASPIRATIONAL]
OTR · X-ray stationEscapes prevented—4 in month 1[ASPIRATIONAL]
Passenger radial · gradingGrade A rate94.8%96.2%[ASPIRATIONAL]
Group · quality systemsNCR closure time11.4 days3.1 days[ASPIRATIONAL]
Passenger radial · press bankTakt66 s61 s[ASPIRATIONAL]

All figures are design-partner targets or in-flight pilot results and are marked [ASPIRATIONAL] until published with partner permission.

What the teams said afterwards

“We had three people watching X-ray images and still shipped uniformity rejects. Treadum flagged the cord shift at the building drum, not four hours later at final inspection.”

Ilse VandermeerPlant Director, passenger radial plant [PLACEHOLDER]

“The cure agent found 40 seconds of margin on a construction we had run the same way for eleven years. It proved it in shadow mode before it touched a press.”

Rahul MenonPrincipal Process Engineer, truck & bus radial [PLACEHOLDER]

“Genealogy is the part I did not expect to care about. Every tire now has a linked record from batch to grade, and audit prep went from weeks to an afternoon.”

Camille DuarteQuality Systems Manager, IATF 16949 site [PLACEHOLDER]

[PLACEHOLDER] Design-partner quotes are illustrative until pilot references are published.

Why the second line is faster than the first

01Connectors reusedThe integration pattern for your PLCs, historian and inspection machines is already written and tested.
02Envelope templatedAction classes, limits and approver roles carry over; only thresholds are re-tuned locally.
03Models transferPerception models transfer well across similar constructions; process models are re-fit on local data.
04Gates knownEveryone already knows what shadow, assist and bounded mean and what has to be true to pass.
05Baseline disciplineThe team has done a rigorous baseline once, so the second one takes days rather than weeks.
  1. Connectors reused: The integration pattern for your PLCs, historian and inspection machines is already written and tested.
  2. Envelope templated: Action classes, limits and approver roles carry over; only thresholds are re-tuned locally.
  3. Models transfer: Perception models transfer well across similar constructions; process models are re-fit on local data.
  4. Gates known: Everyone already knows what shadow, assist and bounded mean and what has to be true to pass.
  5. Baseline discipline: The team has done a rigorous baseline once, so the second one takes days rather than weeks.

Consistent patterns so far

12 wkTypical time from kickoff to bounded autonomy
96%+Agreement with engineers before assist mode
0.11%Best observed false-hold rate [ASPIRATIONAL]
100%Engagements with a written pre-agreed baseline

How to read these results

Because Treadum is early and we would rather be trusted than impressive. Figures marked [ASPIRATIONAL] are targets or in-flight pilot results that have not yet been published with partner permission and third-party verification.

Once a partner has agreed to reference calls, yes. Ask during the plant walk and we will tell you honestly who is available and who is not.

Baselines cover a full production cycle across shifts, constructions and compounds, are captured before any writes, and are agreed in writing by both sides before autonomy is activated.

Legacy inspection machines with no usable output remain read-only, and sites without linked historian data take longer to reach useful cure and rheology models. We will tell you which category you are in during the plant walk.

Put one cell on autonomy in 90 days

Pick one wedge — splice inspection, X-ray defect detection, cure-state optimization or uniformity prediction. We baseline it, run shadow mode, then graduate to bounded autonomy under a signed control envelope.

Pilots start in shadow mode. No line changes until accuracy and safety gates pass.