Research

Open research for material-centric production intelligence

Selected white papers, methodology notes, open reference models, synthetic datasets, and benchmarks. We publish reference intelligence; customer plant data and private deployments stay confidential.

White papers

The thinking behind material-centric production intelligence

White paperPublished14 min

Material-centric production intelligence: improving outcomes before buying assets

The category thesis. Why the constraint is often losing material value still unrealized rather than truly saturated, and how detecting and valuing Material Value Leakage improves material outcomes from existing assets.

Full details

Problem addressed: High-volume plants add capex or shifts while material value still unrealized is lost inside existing assets.
Industry relevance: Metals, chemicals, FMCG/CPG, paper & packaging, glass, cement, ceramics, battery materials.
Data assumptions: Assumes access to PLC/SCADA/historian rate data, MES run/shift records, and downtime logs.
Method: Defines material-centric production intelligence against planning, scheduling, OEE, and predictive maintenance, and frames Material Value Leakage as sub-downtime time loss.
Evaluation: Illustrated with worked material-value economics; not a benchmark of a specific model.
Limitations: Conceptual framing. Any plant conclusion requires validation on that plant's data.
License: CC-BY-4.0
Discussion: 18 comments

Updated 2026-06-28Download

White paperPublished9 min

Material-value economics: turning seconds into realizable material value

The economic model behind the calculator. How to translate material value lost in transformation into realizable material value and a defensible annual value range without false precision.

Full details

Problem addressed: Small, repeated time losses are individually ignorable but compound into large annual output loss.
Industry relevance: Any 24/7 or multi-shift high-volume operation with a monetizable constraint.
Data assumptions: Operating hours, demonstrated rate, contribution margin, and estimated material value lost in transformation per hour.
Method: Presents the material-value formula (material value lost in transformation × frequency × rate × contribution value) and a conservative recovery-range approach.
Evaluation: Range-based worked examples; explicitly avoids single-point savings claims.
Limitations: Estimates depend on user assumptions; ranges are illustrative, not guaranteed savings.
License: CC-BY-4.0
Discussion: 22 comments

Updated 2026-06-22Download

Material Value Leakage benchmarks

Published Material Value Leakage benchmarks

Published definitions behind the Material Value Diagnostic: what each benchmark answers, how it is computed, where it applies, and where it stops.

Material Value Index

What share of planned production time may be lost through Material Value Leakage?

annual operating hours × Material Value Leakage percentage range

Intended use

Opportunity screening: sizing how much production time may be lost before any plant-data work.

Limitations

Uses configurable band assumptions, not measured losses. Requires plant data validation.

Related diagnostic output

Shown as Potential Material Value Leakage on the Preliminary Material Value Map.

Apply this benchmark to your plant

Material Minute Value

What is one lost production minute worth in this plant?

production-hour value ÷ 60

Intended use

Making small time losses commercially tangible for prioritization discussions.

Limitations

Depends on the contribution value supplied; treat as indicative until finance confirms.

Related diagnostic output

Shown per minute on the Preliminary Material Value Map.

Apply this benchmark to your plant

Preliminary Material Value Leakage Signal Score

How strongly do observed behaviors indicate hidden material-value loss patterns?

weighted signal answers, normalized to 100

Intended use

Selecting the Material Value Leakage assumption band and identifying likely patterns to validate first.

Limitations

A self-reported signal, not a diagnosis. Weights and bands are published and configurable.

Related diagnostic output

Sets the signal band on the Preliminary Material Value Map.

Apply this benchmark to your plant

Realizable Material Value Potential

How much production time and output may be recoverable from existing assets?

Material Value Leakage hours × recoverability percentage range

Intended use

Framing the size of the recovery opportunity before capex or extra shifts are considered.

Limitations

Recoverability ranges are conservative assumptions; actual recovery depends on validated causes.

Related diagnostic output

Shown as recoverable hours and output on the Preliminary Material Value Map.

Apply this benchmark to your plant

Bottleneck Recovery Index

How much of the constraint asset's variation gap can be closed toward its demonstrated best?

(demonstrated best rate − typical rate) × constraint hours, from plant data

Intended use

Ranking recovery opportunities on the constraint during a plant-data Material Value Map.

Limitations

Requires actual bottleneck rate history; not computable from the web diagnostic.

Related diagnostic output

Produced during a plant-data Material Value Map engagement.

Apply this benchmark to your plant

Material Value Leakage methodology notes

How the method works, in the open

Methodology notePublished7 min

The Material Value Leakage taxonomy

The eight-class Material Value Leakage taxonomy used across the platform, from cycle-time drift to masked routine inefficiencies, with the data source that tends to expose each.

Full details

Problem addressed: Teams lack a shared vocabulary for sub-downtime losses, so they go unnamed and unmanaged.
Industry relevance: Cross-industry; examples span discrete and process manufacturing.
Data assumptions: Maps each Material Value Leakage class to the plant data most likely to reveal it.
Method: Defines eight Material Value Leakage classes with plain-English definitions, examples, affected outcomes, and data sources.
Evaluation: Reviewed with operations practitioners; open for structured industry validation.
Limitations: A classification aid, not a measurement. Presence of a class must be confirmed in data.
License: CC-BY-4.0
Discussion: 31 comments

Updated 2026-06-30Download

Methodology notePublished6 min

Material Value Diagnostic: transparent scoring and assumptions

How the Material Value Diagnostic maps weighted answers to signal bands, how each band selects conservative Material Value Leakage and recoverability assumptions, and why it avoids any claim of diagnostic certainty.

Full details

Problem addressed: Self-assessment tools often hide their scoring and overclaim certainty.
Industry relevance: Applicable to any plant self-assessing Material Value Leakage exposure.
Data assumptions: None; the diagnostic is a structured self-report, not a data measurement.
Method: Documents the plant-profile inputs, the 10 weighted signal questions, the four signal bands, and the configurable Material Value Leakage and recoverability ranges behind each band.
Evaluation: Face-validity review only; explicitly not a statistical instrument.
Limitations: A preliminary signal with illustrative ranges. It indicates where to look, not what will be recovered.
License: CC-BY-4.0
Discussion: 14 comments

Updated 2026-07-12Download

Methodology noteWorking draft8 min

Data readiness for Material Value Map validation

What data a plant needs before Material Value Leakage can be validated and valued, mapped class-by-class, so a pilot starts from a realistic data picture.

Full details

Problem addressed: Plants are unsure what data is needed before a Material Value Map can be produced.
Industry relevance: PLC/SCADA/DCS/MES/historian environments with run, batch, or shift records.
Data assumptions: Discusses tag coverage, timestamp fidelity, and product/grade context required.
Method: A readiness checklist linking each Material Value Leakage class to minimum viable data for validation.
Evaluation: Draft; open for practitioner review in Collaborate.
Limitations: Readiness varies by site; the checklist is a starting point, not a guarantee.
License: CC-BY-4.0
Discussion: 9 comments

Updated 2026-07-02Download

Open models & model cards

Reference models from the community

Selected open models surfaced from the Matereal Community, each with a serious model card (intended use, inputs, outputs, training-data class, evaluation, limitations).

All models

Synthetic-data modelOpen · synthetic

Rare Earth Recovery Predictor

Predicts element recovery across a solvent-extraction separation train from stage conditions and feed assay. Reference implementation trained on synthetic separation data.

Rare earthsRecovery prediction

rare-earth-separation

8.4k 214 46Apache-2.0

Reference modelOpen · reference

Rare Earth Purity Classifier

Classifies product purity band from stage chemistry and impurity indicators. Reference implementation for benchmarking and education.

Rare earthsPurity classification

rare-earth-separation

5.1k 148 22Apache-2.0

Public-data modelOpen · public data

Aluminium Melt Chemistry Predictor

Predicts resulting melt chemistry from charge composition, alloying additions, and furnace conditions. Public-data reference implementation.

AluminiumChemistry prediction

aluminium-melt

9.3k 267 51CC-BY-4.0

Reference implementation. Industrial deployment requires validation on plant-specific data. Reference implementation. Industrial use requires validation on plant-specific data.

Synthetic datasets

Open, non-confidential data

All datasets

SyntheticCC-BY-4.0

Synthetic Line-Speed & Micro-Stop Dataset

High-frequency line-speed and machine-state traces for a packaging line, labelled with short stops that never reach the downtime threshold.

FMCG / CPGMicro-stoppage mining

Hidden Small Stoppages

540k

rows

features

5.1k

downloads

SyntheticCC-BY-4.0

Cycle-Time vs Standard Dataset

Per-cycle timestamps paired with the validated standard time, product, and shift, with injected gradual drift for detection tasks.

MetalsCycle-time drift detection

Cycle-Time Drift

210k

rows

features

4.4k

downloads

SyntheticCC-BY-4.0

Changeover & Transition Events Dataset

Changeover start/end events with product sequence and crew, labelled against the best repeatable time per transition family.

FMCG / CPGChangeover creep estimation

Changeover Creep

38k

rows

features

downloads

Benchmarks

Reproducible leaderboard tasks

All benchmarks

MetalsFlag cycles inside a drift episode

28 subs

Cycle-Time Drift Detection

Detect gradual cycle-time drift against validated standards from per-cycle timestamps. Higher F1 on drift episodes is better.

MetricF1 (drift episodes)

BaselineCycle-Time Drift Detector

#1 h.novak0.905

View Benchmark

FMCG / CPGDetect sub-downtime micro-stop events

21 subs

Micro-Stoppage Mining

Recover short stops that never reach the downtime threshold from PLC state changes and line-speed signals. Higher event recall is better.

MetricRecall (micro-stops)

BaselineIdle-Wait Opportunity Model

#1 p.nakamura0.871

View Benchmark

Paper · Glass · CementQuantify slow-running loss versus best rate

19 subs

Throughput Drag Estimation

Estimate sustained below-best-rate operation without a classified stoppage. Lower rate-gap MAE is better.

MetricMAE (rate gap %)

BaselineThroughput Drag Estimator

#1 r.okafor3.8

View Benchmark

Research discussions

Every open asset links to a structured discussion in the Matereal Community, with validation notes, method questions, and dataset requests from practitioners.

Material value first

Identify Your Material Value Leakage

Start with one high-value production decision. Prove value in one plant. Then scale across the installed base.

Identify Your Material Value Leakage Explore Technology Partnerships