Purpose

The Throughput Reliability Index is a working method for distinguishing reported performance from performance a planner can depend on.

TRI does not replace Overall Equipment Effectiveness. OEE remains a useful description of availability, performance, and quality. TRI asks a different question:

Given the current level, observed instability, and direction of travel, how much confidence should the operating system place in this line’s output?

Public model

TRI organizes the judgment around three factors:

  • Throughput: current performance relative to a defined target or demonstrated capability;
  • Reliability: the stability of that performance across the relevant operating window;
  • Direction: whether the process is improving, holding, or deteriorating.

The public relationship is:

TRI = throughput factor × reliability factor × direction factor

Detailed calibration choices, thresholds, and client-specific adjustments are not part of this public version. They remain hypotheses requiring commercial and cross-site validation.

Reading the result

TRI is most useful as a state classification rather than an isolated league-table number:

State Operating interpretation Decision implication
Strong and dependable Adequate level, bounded variation, stable or improving Reasonable planning anchor; continue monitoring
Strong but fragile Adequate average, excessive variation or decline Protect commitments; investigate the source of instability
Weak but recovering Inadequate level with credible improvement Support the intervention; verify that gains hold
Weak and deteriorating Inadequate level with instability or decline Escalate; reduce exposure and identify the binding constraint

These states do not assign blame. A fragile line may reflect equipment, material, schedule, staffing, measurement, or product-mix effects.

Required decomposition

A reliability haircut without diagnosis is merely a more pessimistic KPI. The method therefore requires a second step: determine where the variation lives.

  • Within-shift: short stops, process drift, material changes, or equipment behavior.
  • Between-shift or crew: training, standard work, handoff, or supervision differences—after controlling for schedule and product.
  • Schedule-induced: changeover burden, run length, sequence, or product mix.
  • Measurement-induced: reason-code behavior, missing events, rate definitions, or inconsistent planned-downtime rules.
  • Special cause: a discrete event or regime change that should not be averaged into ordinary behavior.

The decomposition routes investigation; it does not prove root cause by itself.

Worked interpretation

Suppose two lines report the same average OEE over four weeks.

  • Line A stays in a narrow range and has no meaningful downward movement.
  • Line B alternates between very high and very low shifts and has declined during the final week.

OEE correctly reports the same average. TRI would classify Line A as the safer planning anchor and Line B as fragile. The important output is not simply that Line B receives a lower score. It is that the planning decision changes: service commitments, buffer, maintenance review, schedule exposure, and follow-up should reflect the instability.

Governance requirements

A TRI output should be withheld or marked provisional when:

  • targets or ideal rates are not stable and documented;
  • the measurement system changed inside the analysis window;
  • required shift or product context is missing;
  • the sample is too small for the chosen reliability estimate;
  • a known special cause makes ordinary comparison misleading;
  • the model cannot distinguish missing data from good performance.

Every alert should retain the input window, calculation version, context tags, confidence state, owner, action, and verification date.

Validation agenda

TRI has not yet established universal validity. The next proof standard is prospective:

  1. Compare TRI against OEE-only, moving-average, and ordinary SPC baselines.
  2. Freeze calibration before testing on a new line.
  3. Measure whether TRI anticipates missed plans, overtime, downtime, or intervention need.
  4. Test sensitivity to windows, targets, product mix, and missing data.
  5. Track whether decisions influenced by TRI produce better verified outcomes.
  6. Publish negative results and cases where the simpler baseline wins.

What would weaken TRI

TRI weakens if its rankings are dominated by arbitrary calibration, if it adds no prospective signal beyond standard process-control methods, if state classifications do not change decisions, or if teams cannot decompose the signal into an actionable investigation.

Selected technical lineage

  • Seiichi Nakajima, Introduction to TPM: Total Productive Maintenance.
  • NIST/SEMATECH, e-Handbook of Statistical Methods.
  • Douglas C. Montgomery, Introduction to Statistical Quality Control.

Revision history

  • v0.1 public — August 7, 2026: Published conceptual factors, state interpretation, decomposition, governance, limitations, and validation agenda; withheld calibration specifics.