If you’ve spent any time reviewing ILS deviation events in an FDM program, you already know the frustration. Your analysts open the queue, and a significant portion of what’s waiting for them isn’t a meaningful deviation at all — it’s noise. Aircraft that weren’t even on an ILS approach. Signals that were present, technically exceeded threshold, and triggered an event that should never have fired.
In traditional FDM systems, this is treated as an inevitability. You build workarounds: derivative filters, signal stability logic, tuned inhibit windows. You get clever with mathematics to try to infer whether a glideslope or localizer signal is meaningful. And you accept that a certain percentage of your analyst’s time is going to be spent dismissing events rather than acting on them.
We didn’t accept that.
The Root Cause Is Simple — and So Is the Fix
On most aircraft, the glideslope and localizer signals are present on the bus regardless of whether a valid ILS frequency is actually selected. The aircraft doesn’t care. It’s just reporting what the receiver sees. And if your event detection logic is watching those signals without knowing what’s tuned in the nav radios, you’re going to generate events on approaches where ILS wasn’t even the intended navigation source.
Traditional detection logic tries to solve this indirectly — working backward from signal behaviour to infer validity. It’s an approximation, and approximations have error rates.
Sky Analyst’s new Event Detection Engine, EDEN, takes a more direct approach: when navigation frequency is recorded in the flight data, we use it
With Sky Analyst’s EDEN, we can now pull the ILS Frequency from the airport database and compare it to what’s recorded in the airplane.
The impact on event quality was immediate and significant.
Prior to updating our glideslope and localizer events to use this logic, our false event rate for these event types was running at approximately 30% — meaning roughly one in three events flagged for analyst review had no operational significance.
Since deploying the updated detection through EDEN, that rate has dropped to under 5%, and it continues to improve as we refine the implementation across different aircraft types and nav system configurations.
That’s not a marginal improvement. That’s a fundamental change in what analysts are looking at when they open their queue.
Why This Matters Beyond the Numbers
A 30% false event rate doesn’t just waste analyst time — it erodes confidence in the program. When analysts learn through experience that a large fraction of a given event type is noise, they start triaging differently. The events get reviewed faster, with less scrutiny. In a worst case, genuine deviations start to blend into the background.
Reducing false events isn’t just an efficiency gain. It’s a credibility gain. When every event in the queue has a higher prior probability of being real, analysts engage more carefully with each one — and safety-relevant findings are less likely to be dismissed on pattern recognition alone.
This is the philosophy behind EDEN: event definitions should reflect operational reality. Using nav frequency data to validate ILS events isn’t a novel idea — the information has been sitting in flight data recorders for years on aircraft that capture it. The question was whether the detection engine was capable of acting on it.
Is Your System Leaving Signal in the Data?
If your current FDM platform is generating high false event rates for glideslope and localizer deviations (or other events), the fix may already be in your data. The question is whether your detection engine can reach it.
Let’s keep in touch
Sign up to get notified of new blog posts, videos or other news and information related to flight data.
