Leading Indicators vs Lagging Indicators in Aviation Safety
01The distinction
Strip away the jargon and the difference is about time. A lagging indicator measures an outcome that has already happened. A leading indicator measures something that happens before the outcome and tends to move with it.
Lagging indicators look backward. They count results, events, losses, failures. They are usually easy to define, hard to argue with, and directly meaningful: an occurrence rate is unambiguous. Their limitation is built into their nature: they can only tell you about a problem once it has already produced the outcome you were trying to avoid. By the time a lagging indicator moves, the lesson has been paid for.
Leading indicators look forward. They measure the conditions, activities and precursors that precede outcomes, the state of a barrier, the rate of a precursor event, the currency of a critical competency. Their strength is timing: a leading indicator can deteriorate while every lagging indicator is still green, giving the operation a window to act. Their difficulty is that they require a real hypothesis about cause, a belief, supported by evidence, that this condition tends to precede that outcome. A leading indicator with no causal link to anything is just a number.
One subtlety matters throughout: an indicator is not leading or lagging in the abstract. It is leading or lagging relative to a specific outcome. The same measure can be a lagging indicator for the event it counts and a leading indicator for a more serious event that event can contribute to. Naming the outcome an indicator is meant to anticipate is half the discipline.
02What each looks like in aviation
The distinction becomes concrete in an airline's actual measures.
Typically lagging: the rate of reportable occurrences; the accident and serious-incident rate; the number of audit findings raised; the count of high-consequence events by category. These describe what the operation has already produced. They are essential for validation and trend confirmation, and useless for early warning.
Typically leading: the stabilised-approach rate; the health of specific barriers, tracked as they are verified; the currency of safety- critical training and competency; the rate of voluntary reporting as a measure of the reporting culture itself; the trend in a flight-data exceedance that is still inside limits. Each watches a condition that tends to move before the outcomes above.
The trap in the middle: activity metrics. Many measures that feel leading are merely activity counts, “number of audits completed,” “number of safety meetings held,” “percentage of actions closed on time.” They are easy to collect and they predict almost nothing about safety outcomes, because doing an activity is not the same as the activity changing risk. An activity metric tells you the machine is running; a true leading indicator tells you which way the operation is heading. Confusing the two is the single most common way an indicator set looks forward-looking while seeing nothing.
03Why you need both
It is tempting to conclude that leading indicators are simply better and to chase them exclusively. That is a mistake. The two types do different jobs and check each other.
Lagging indicators validate. They are the ground truth. A leading indicator is a hypothesis about what precedes harm; the lagging indicators are how you find out whether the hypothesis is right. If a leading indicator improves for two years while the outcomes it claims to predict do not, the indicator is measuring the wrong thing, and only the lagging data can reveal that.
Leading indicators steer. Lagging data confirms direction too late to change it. Leading indicators are what let an operation adjust before the outcome lands. Used together, lagging indicators keep the leading ones honest, and leading indicators give the lagging ones somewhere useful to point.
Under ICAO Annex 19, safety performance monitoring is meant to be exactly this balanced, target-bearing picture, indicators with safety performance targets and alert levels, read together rather than in isolation. A dashboard that is all lagging is a history book; one that is all leading is an untested theory. The serious program carries both, and knows which is which.
04The mistakes that weaken an indicator set
Most indicator sets do not fail loudly. They quietly stop meaning anything. The recurring ways:
- Activity dressed as foresight. Counting work done and calling it a leading indicator. If the measure would not change when risk changes, it is not leading anything.
- Threshold-only thinking. Watching whether a line is crossed and ignoring the slope toward it. The approach to the threshold usually carries more warning than the breach, a point that recurs in weak signals.
- Too many to watch. A long list of indicators nobody owns or reviews. Coverage is not the goal; attention is. A handful of well-chosen, owned indicators beats a wall of charts.
- No causal story. Tracking a number because it is easy to get, with no articulated reason it should precede harm. The reason is what makes it an indicator rather than a statistic.
- Gaming and goal displacement. Once an indicator becomes a target people are judged on, the temptation is to improve the number rather than the underlying condition. Pairing every leading indicator with the lagging outcome it claims to predict is the main defence.
05From the concept to a working library
Understanding the distinction is the foundation. Turning it into a maintained set of indicators, choosing which hazards to instrument, defining each measure precisely, setting baselines, targets and alert levels, assigning ownership, and keeping the dashboard honest over time, is a separate, practical body of work. That is deliberately not this article's job.
For the build, see how to build a practical SPI library, which covers the mechanics end to end. For why leading indicators matter so much to seeing risk early, see why mature SMS programs miss emerging risk. And for where a balanced indicator set sits in the larger picture, see what aviation safety intelligence is, living indicators are one of its load-bearing foundations.
Questions
Lagging indicators measure outcomes that have already occurred, occurrence rates, accident rates, the number of findings raised. Leading indicators measure the conditions and activities that precede those outcomes, stabilised-approach rates, barrier health, training currency, the rate of voluntary reporting. Lagging tells you where you have been; leading tells you where you are heading. A safety performance picture needs both, but only leading indicators give you the chance to act before the outcome happens.
Safety performance indicators can be either, and a healthy SPI set contains both. ICAO frames SPIs as the measures an operator uses to monitor and assess safety performance, with associated safety performance targets and alert levels. Some SPIs are lagging (a high-consequence event rate); others are leading (a precursor or activity rate that tends to move before the outcome does). The useful question is not "is this an SPI?" but "does this indicator look backward at outcomes or forward at conditions?"
Because they are the only indicators that give you time. A lagging indicator can only confirm a problem after it has produced an outcome; by then the cost of the lesson has already been paid. A leading indicator watches the conditions that precede the outcome, so a deteriorating trend becomes visible while intervention is still cheap. An operation that steers only by lagging indicators is, in effect, navigating by the wake behind the aircraft.
It depends on what it precedes. An indicator is leading or lagging relative to a specific outcome, not in the abstract. A runway-excursion rate is a lagging indicator for excursions, but it can act as a leading indicator for the more serious outcomes that excursions can contribute to. The discipline is to be explicit about the outcome an indicator is meant to anticipate, rather than labelling it once and forgetting what it is for.
Few enough that each is owned, watched and acted on, a focused set tied to the operation's real hazards beats a long list nobody reviews. The right number depends on the size and complexity of the operation, the hazards that matter most, and the data available to support each indicator. The practical work of choosing, defining, setting targets for and maintaining a balanced set is the subject of the companion article on building a practical SPI library.