A regional carrier operating 14 aircraft out of a mid-sized hub in the southeastern US grounded one of its ATR 72s for four days in early 2026 over a single MLG actuator seal kit. The part was in the system. It was sitting in a warehouse 800 miles away, allocated to a scheduled C-check that was still 11 weeks out. The planner who bought safety stock six months earlier had no way of knowing that the actuator would start logging removal warnings on a different tail in week three of winter operations.
That story is not unusual. Regional carriers, especially those running narrowbody and turboprop fleets with high daily cycle counts, face a structural mismatch: their removal data is rich, but their planning tools were not built to extract forward signal from it. The result is a cycle of reactive purchasing, expedited freight charges, and AOG events that everyone can explain after the fact but nobody saw coming.
Why Removal History Contains More Signal Than Most Systems Use
Every unscheduled removal generates a record: part number, tail number, position, date, reason code, hours and cycles since last shop visit. Most MRO systems, including widely used platforms like AMOS and CESIUM, store this data faithfully. What they do not do is treat it as a time series with predictive properties.
A demand forecasting model trained on removal history can identify patterns that a planner looking at a static reorder-point report would miss. Actuator seal failures on a specific tail tend to cluster after approximately 3,200 flight cycles in high-humidity operating environments. Brake assembly removals on an older variant of a narrowbody spike in months 18 to 24 after the most recent shop visit. These are not hunches; they are patterns in the removal record that repeat across fleets when you have enough tails and enough history to see them.
The challenge is that a fleet of 14 aircraft does not generate enough removals on its own to train a robust model for every part number. A single tail might see a given LRU removed twice in three years. That is not enough data to distinguish a genuine wear pattern from noise. This is where cross-fleet learning becomes necessary: a model trained across dozens of similar fleets, even anonymized ones, can surface patterns that no single operator's data would reveal alone.
What Aerotrax Actually Looks At
When a carrier loads removal history into Aerotrax, the first thing the system does is not generate a forecast. It segments the removal records by part classification, tail variant, and operational profile. A carrier that does three short-haul turns per day on a coastal route stresses components differently than one doing two long-haul sectors with mountain-airport intermediate stops. Aggregating those into a single demand number produces a forecast that fits neither fleet.
After segmentation, the model builds a removal-rate distribution for each part-variant combination, weighted by cycles and adjusted for age cohort. Parts in their first 5,000 cycles since last overhaul behave differently from the same parts approaching their next scheduled removal threshold. That age-weighting matters enormously for parts with non-linear failure distributions, which includes most hydraulic actuators, pneumatic seals, and brake system components.
The output is not a single point forecast. It is a probability band with a recommended stock position. A part might have a 15% chance of requiring one unit in the next 30 days, a 60% chance of requiring one unit within 90 days, and a meaningful tail probability of needing two units in that window if the fleet is running above average cycles. The planner sees those bands and can decide how much stock to carry against their specific AOG tolerance.
The Gap Between a Good Forecast and a Good Purchasing Decision
We want to be clear about what a forecasting tool does and does not do. A better demand signal does not automatically produce a better inventory position. It still requires a planner to connect the forecast to lead times, current on-hand quantities, repair turnaround commitments for rotables, and the cost of carrying extra stock versus the cost of an AOG event.
What Aerotrax changes is the quality of the inputs to that decision. Instead of looking at a reorder point calculated from a 12-month rolling average that includes a summer maintenance surge and a quiet winter period equally, the planner sees a forward-looking band that accounts for the fleet's current age profile and recent removal patterns. That does not make the decision automatic. It makes it less blind.
In one early-access pilot with a European turboprop operator running eight ATR 42s, the output that produced the most immediate value was not the AOG-risk flags on high-criticality parts. It was the identification of six part numbers that had been maintained at safety stock levels consistent with historical demand that was now six years stale. The fleet had been modified, the routes had changed, and two of those parts were generating zero removals while consuming shelf space and working capital. The planner had suspected this but had no clean way to confirm it against current removal patterns across the fleet.
Where Regional Carriers Face the Hardest Trade-offs
Large mainline carriers have enough procurement leverage and enough warehouse space to absorb some inventory inefficiency. A regional operator with 12 to 20 aircraft and a constrained budget cannot. Every dollar sitting on the shelf in the wrong part number is a dollar that is not available for the next unplanned removal.
The trade-off for a regional carrier is not just "do we carry the part or not." It is: do we carry one unit in our local stores, rely on an AOG pool agreement with a supplier, or accept the expedite risk? That decision should be made with demand data, not just historical purchase records. A supplier agreement for same-day delivery is worth less if the part you genuinely need most is not on the pool list because your historical demand for it looked low when you negotiated the contract two years ago.
Aerotrax surfaces these gaps. The goal is not to replace the planner's judgment about which agreements to hold or how much safety stock budget to allocate. The goal is to make sure the judgment is operating on current demand signal rather than stale averages.
What "AOG Reduction" Actually Means in Practice
Saying that better forecasting reduces AOG risk is technically correct but not very operational. What it means in practice is narrower: it means fewer situations where a planner, running an AOG pool call at 2 a.m., is doing it because their system gave no advance warning that a part was trending toward removal. It means more situations where a planner can look at the next 60-day horizon, see that two specific part numbers are carrying elevated risk based on current fleet cycles and age, and choose to pre-position stock before that risk materializes.
It does not mean zero AOG events. Unscheduled removals will always happen, and some will involve parts that were genuinely unforeseeable at any horizon. But a meaningful share of AOG events in regional operations involve parts with recognizable precursors in the removal record. Those are the ones that better forecasting can move from emergency response to planned procurement.
The four-day grounding in southeastern US that opened this article would not have been preventable with certainty. But with a model that was reading removal patterns on that actuator across the full fleet, the planner would have seen elevated short-horizon demand on the seal kit before the warning arrived. That is the gap Aerotrax is built to close.
See Aerotrax on your own fleet data
30-day pilot on your removal history. CSV upload, no integration project required.
Request a pilot