Two forecasting models, weighted for how demand actually behaves.
A multi-fleet operator needed one framework that could tell the difference between a scheduled heavy-maintenance requirement and an unscheduled failure — and weight its inputs accordingly. We built both, across every rotable and consumable in the fleet.
Material forecasting is difficult but critical
Strato Technologies has developed several AI-powered tools for material forecasting over the years. These algorithms work at both a statistical and a qualitative level, and the scope spans the full material base — rotables, including hard-time components, and consumables, including expendables.
We incorporate input from domain experts — engineering and production — to calibrate and fine-tune our models. No single data source tells the whole story: historical consumption explains the past, but only expert judgement and live operational signals explain what's changing.
That's why we don't run one model. We run two — one tuned for scheduled, predictable demand, and one tuned for unscheduled, reliability-driven demand — each weighting its inputs differently.
Scheduled and unscheduled demand don't listen to the same signals
Heavy-maintenance requirements are driven by history, engineering judgement and the work-pack plan. Unscheduled requirements are driven by defects, requisitions and field reliability. Each gets its own usage-prediction model — and its own input weighting.
Heavy Maintenance Material Requirements
Downstream, the report moves through Engineering review, Supply Chain requisitioning and Production execution before the short-term planning date.
Accuracy and/or Attainment
We implement a feedback loop to continuously retrain the algorithm, increasing attainment of the predicted part number over time.
Field Experience and Knowledge
An extra training layer incorporates field experience directly into the model, not just historical statistics.
Unscheduled Material Requirements
All five signals converge on a single repair / replacement recommendation, tracked continuously rather than reset each cycle.
Unexpectedness as a Key Metric
Statistical analysis relies on a null model to test the hypothesis. Any deviation from it is tracked as an unexpected-event metric in its own right.
Reliability Tracking, Expanded
Unscheduled events affect reliability directly. We look for synergies across data sources to expand visibility into failure rates.
Publishing the report is the beginning, not the end
Scheduled and unscheduled forecasts feed the same pipeline — and the pipeline doesn't stop at the report. Strato Tech runs the deployment discipline that turns a forecast into action.
-
Heavy Maintenance Requirements
Determines the actual material needs for upcoming heavy-maintenance checks, based on historical data, field input and engineering recommendations.
-
Heavy Maintenance Work Packs
Material requirements are always linked to the development of the work packs, not forecast in isolation.
-
Unscheduled Material Requirements
Analyses and determines additional material-usage patterns for unscheduled events, based on reliability and historical data.
-
Full Report Publishing & Deployment
Producing the report is the beginning, not the end. Strato Tech runs the deployment discipline — regular meetings and reviews with senior management.
A repeatable journey, not a one-off report
Every engineering-forecast engagement follows the same disciplined path — across rotables, hard-time components and consumables alike.
-
We consolidate your history.
Task-card-level execution history, planned work orders, planning records and field input, cross-referenced with ML-powered text matching — across every rotable and consumable in scope.
-
We split scheduled from unscheduled demand.
Heavy-maintenance requirements and unscheduled failures behave differently, so we weight their inputs differently — MPD and work-pack data for one, defects and requisitions for the other.
-
We estimate remaining useful life.
An optimisation algorithm models RUL for every rotable and consumable in scope, whether hard-time or expendable.
-
We publish, and then we deploy.
A report is only the beginning. We run the deployment discipline — reviews with engineering, supply chain, production and senior management — that turns a forecast into action.
-
We track attainment and unexpectedness.
Forecast accuracy and deviation from the expected are both tracked as metrics — fleet by fleet — and fed back into the next cycle.