Fleet Intelligence · How suggestions work

Every screen ends in something to do

The platform checks your records against 23 rules and turns what it finds into one ranked list. Each item says what fired, shows the evidence, lists the steps to take and, where the data allows, puts a figure on what's at stake.

Anatomy of a suggestion

Why it fired, and what to do

This one comes from the demo airline. Open its evidence and you land on the screen that proves it, with every write-up behind it.

SafetyReliabilityLR-MDB

Repetitive defect on LR-MDB: ATA 27-81, 5 times in 12 days

Evidence

"Slat disagree light on approach": five reports between 3 and 15 August. Actions recorded: reset and check twice, check normal once, swap and operational test once, adjustment once. Mostly resets and checks: the fix has not held.

What to do
  • Raise it as a repetitive defect: engineering review before the next release, per your repetitive-defect procedure.
  • Most rectifications were resets, checks or swaps: the fault was not found. Follow the troubleshooting tree to the end or escalate to the manufacturer.
  • Check wiring and connectors in the area, not only the components.
  • If a component was changed, check whether the same serial came back.
Ranking

Safety first, then reliability, then cost

Within each group, items are ordered by the money at stake. The same data always gives the same list.

Safety
MEL items past their limit, engines nearing their EGT limit, repetitive defects in safety-related chapters
Reliability
Chapters in alert, aircraft carrying a problem, rogue units, part types wearing out, dispatch below target
Cost
Part numbers to standardise on, no-fault-found removals, surplus spares, excess fuel, LED retrofit
Data
What's missing to make a figure exact, and costs still on example values
The workflow

From open to done, and kept that way

  1. Decide

    Set the status — open, accepted, in progress, done or dismissed — an owner, a due date and a note such as the work order number.

  2. Recalculated, not reset

    The list is worked out again after every import and settings change. Your status, owners and notes are kept.

  3. Resolved by the data

    When the condition behind an item disappears, it is marked resolved. If it comes back, it reopens, with its history.

  4. Into the report

    Recommendations print in the monthly reliability report with their status, owner and due date.

The 23 rules

What the checks look for, by product

Reliability
  • Chapter above its alert level, or rising three months in a row
  • One aircraft carrying a chapter
  • Repetitive defect, with fixes that were only resets or swaps
  • MEL item past its category limit
  • Item deferred again and again
  • DOI rising
  • Dispatch reliability below target
  • Engine nearing its EGT limit
  • Unscheduled engine removals above their level
Fuel and data
  • Aircraft burning 2% or more above the model
  • Burn rose after maintenance
  • Extra fuel carried at landing
  • Data quality: assumptions in use, costs still on example values
Parts
  • Time between removals down 30% or more
  • High no-fault-found share
  • Rogue serial
  • Repair shop whose units come back sooner
  • Wear-out or early failure (Weibull)
  • Spares short of the fill rate, or in surplus
  • Standardise on the best part number
  • Wheel position removed more often
  • Wheels or brakes wearing faster
  • LED retrofit for frequent light changes
Where AI fits

The rules don't need it

Every alert, suggestion, saving and figure comes from statistics on your own records: alert levels, Poisson and binomial tests, Weibull life analysis and a regression model of fuel burn. The same data always gives the same answer, and every result can be traced to the records behind it.

Generative AI is an optional writing aid, switched on per airline: it can draft an engineer's note for an action, draft the report's executive summary, propose standard defect codes for free-text write-ups, and answer questions about the figures on the page. Every AI text is labelled as a draft to check.