Safety

Unstable approaches: why crews land anyway, and what research says

October 2, 202613 min readFor pilots

Leer en español · Lire en français · Ler em português

In this article

How often crews continue
What "stabilized" means: the FSF criteria
Why it matters: the runway excursion link
Why crews press on: plan continuation and the 97%
The go-around has its own risks
The pushback: are the criteria right?
Where policy stands in 2026
What it means for pilots

Every airline pilot learned the rule early: if the approach isn't stable at the gate, go around. Yet the data says most crews don’t. The Flight Safety Foundation's go-around project found that crews comply with go-around policy in only about 3% of unstable approaches (FSF, 2017). A U.S. industry committee put the go-around rate even lower in 2024 (FAA ACT ARC, 2024).

On August 4, 2026, the UK Civil Aviation Authority issued Safety Notice SN-2026/008 after its data showed a growing number of unstable approaches continued to landing (UK CAA, 2026). Here is what the research says about why crews press on, why the go-around carries its own risk, and what to do about it.

How often crews continue

The FSF Go-Around Decision-Making and Execution Project published its final report in March 2017. The Presage Group did the psychological research, surveying 2,340 pilots and 128 managers worldwide (FSF, 2017). Its baseline numbers:

Unstable approaches occur on 3.5% to 4.0% of all approaches.
95% to 97% of crews in that state continue to land.
Managers did not know this. Those who gave an estimate of industry compliance guessed about 20% on average, several times the real figure.

The FSF calls failure to go around the number one risk factor in approach and landing accidents (FSF project page).

The ACT ARC, an FAA advisory committee, gave different figures in September 2024: about 10% of Part 121 approaches are unstable, and only about 0.5% of those end in a go-around. By its estimate, that is roughly 2,000 unstable approaches a day in U.S. Part 121 flying, with about 10 go-arounds (FAA ACT ARC, 2024). The two sources define and measure "unstable" differently, so the rates differ. Both point the same way: continuing is the norm.

What "stabilized" means: the FSF criteria

The reference definition comes from the FSF Approach-and-Landing Accident Reduction (ALAR) Tool Kit, Briefing Note 7.1, published in 2000 (FSF ALAR BN 7.1). All flights must be stabilized by 1,000 ft above airport elevation in IMC and 500 ft in VMC. An approach is stabilized when all of the following are met:

#	FSF criterion	Typical value
1	On the correct flight path	—
2	Only small heading and pitch changes needed to hold the path	—
3	Airspeed	Not more than VREF + 20 kt IAS, not less than VREF
4	Configuration	Correct landing configuration
5	Sink rate	No greater than 1,000 fpm (a special briefing if more is needed)
6	Power	Appropriate for the configuration, not below the AOM minimum approach power
7	Briefings and checklists	All completed
8	Approach-specific	ILS: within one dot of glideslope and localizer. Cat II/III: within the expanded localizer band. Circling: wings level on final by 300 ft
9	Unique procedures or abnormal conditions	Need a special briefing

The note's rule is blunt. An approach that becomes unstabilized below the gate calls for an immediate go-around. IATA's 2017 guidance repeats the same core criteria and gates (IATA, 2017). So does the FSF and EUROCONTROL runway excursion action plan (GAPPRE, 2021). Your operator's numbers govern.

Why it matters: the runway excursion link

The original ALAR Task Force found unstabilized approaches were a causal factor in 66% of 76 approach-and-landing accidents and serious incidents from 1984 to 1997 (FSF ALAR BN 7.1). High-energy approaches led to loss of control and runway excursions. Low-energy approaches often ended in controlled flight into terrain.

Runway excursions remain the most common accident type:

From 2005 to mid-2019, they made up 23% (283) of accidents in IATA's global database (GAPPRE, 2021).
IATA's July 2025 risk assessment counted 104 excursion accidents from 2015 to 2024, 21% of all accidents. About 80% happened on landing (IATA, 2025).
An FSF study of 16 years of runway excursions found that 83% could have been avoided by a decision to go around (FSF, 2017).

The UK CAA's 2024–2025 occurrence data puts the risk another way. It says an approach continued while unstable is "80-100 times more likely" to end in an adverse event (UK CAA, 2026). FlightGlobal reported the CAA's concern over hard landings and tail strikes (FlightGlobal, 2026).

Why crews press on: plan continuation and the 97%

Plan continuation bias. NASA researchers named the pattern decades ago. Judith Orasanu and Lynne Martin reviewed an NTSB analysis of 37 crew-caused accidents. Tactical decision errors played a part in 25 of them, about two in three. The common pattern was a crew sticking with the plan when conditions called for a change (Orasanu & Martin, 1998). They named four drivers:

Ambiguous cues. Conditions degrade gradually. It takes a lot of evidence to abandon a plan already in motion.
Underestimated risk. Past success in similar situations makes the next one feel safe.
Goal conflicts. On-time performance and fuel compete with safety margins.
No time to think ahead. Time pressure and stress limit the ability to play out the consequences.

Benjamin Berman and Key Dismukes of NASA Ames studied 19 airline accidents. They found plan continuation bias in at least 9 of them (Berman & Dismukes, 2006). They described snowballing workload that pushes a crew from planning into reacting, and argued that most experienced pilots could make similar choices. Their advice: ask early what happens if your assumptions are wrong, and how you will know in time.

What the FSF survey added. The Presage Group compared pilots recalling an unstable approach they continued with pilots recalling a go-around (FSF, 2017):

Pilots who continued rated the instability as much less risky. FSF's later summary noted they were more confident they could correct it (FSF AeroSafety World, 2019).
Many pilots said current go-around criteria were not realistic for daily operations.
Fewer than half of the pilots who continued believed they would be reprimanded for it.
They were more likely to feel crew pressure to land, and less likely to feel crew support for a go-around.
Pilots who went around were about four times as likely to recall someone prompting the crew to go around: 56%, against 13% of pilots who continued.

The last point matters most. A prompt from someone else, often the other pilot, is what breaks the plan.

Success reinforces the habit. Most unstable approaches end without harm, which teaches the wrong lesson. In August 2025, the Australian Transport Safety Bureau published its report on an Embraer E190 that landed at Darwin from an unstable approach. The crew was not stable at 1,000 ft and judged they could be stable by 500 ft once visual (ATSB, 2025). The landing was uneventful. The ATSB warned that such an approach "can be completed successfully, which may reinforce bad practice."

The go-around has its own risks

Some reluctance is rational. The go-around is rare and high-workload.

The BEA study. France's BEA published its Study on Aeroplane State Awareness during Go-Around (ASAGA) in August 2013 (BEA, 2013). It identified 25 go-around accidents and serious incidents from 1985 to 2010 and analyzed 10 in detail. Its key findings:

Go-arounds are rare. There are 2 to 4 per 1,000 flights, so a medium-haul pilot flies about one a year, and a long-haul pilot may fly one every 5 to 10 years.
Pitch and thrust are hard to manage. Engines at high thrust and low speed give a strong, quick nose-up pitching moment. Pilots surveyed called full thrust excessive when the level-off altitude is close.
Workload is high. Capturing the level-off altitude and managing the automation were the hardest tasks. Flap and gear retraction compete with pitch and thrust.
Monitoring breaks down. The PM was too busy to monitor the PF. Weak monitoring was cited in nine cases. The BEA also suspected somatogravic illusions in several events.

The BEA issued 34 safety recommendations. The FSF survey found a similar risk from the crew side: pilots reported a potentially hazardous outcome, such as a performance exceedance, in about one in 10 go-arounds (FSF, 2017).

Emirates 521, Dubai, 2016. On August 3, 2016, a Boeing 777-300 (A6-EMW) floated in thermal updrafts and touched down about 1,090 m (3,580 ft) past the threshold of Runway 12L at Dubai. The crew then tried to go around. The UAE investigators' final report was published January 20, 2020 (GCAA AAIS, 2020). Because the airplane had touched down, the TO/GA switch logic was inhibited, so thrust stayed at idle. The crew did not notice. The airplane climbed to 85 ft radio altitude at 131 kt, then sank back onto the runway. The thrust levers were moved forward manually only about three seconds before impact. A firefighter died, and the airplane was destroyed by the fire that followed. Flight data showed a 16 kt tailwind from about 200 ft that shifted to a light headwind just before touchdown. The report found the commander had never practiced a go-around after touchdown with the autothrottle active. Training covered go-arounds above 50 ft.

Asiana 214, San Francisco, 2013. On July 6, 2013, a Boeing 777 struck the seawall on a visual approach to Runway 28L, and three people died. The NTSB adopted its report on June 24, 2014 (NTSB AAR-14/01). The PF selected FLCH, then pulled the thrust levers to idle, which put the autothrottle in HOLD. In HOLD, it no longer controlled airspeed. Speed decayed to about 103 kt against a 137 kt target. The NTSB said the approach was unstabilized at 500 ft and the crew should have gone around. The go-around came below 100 ft, too late for the airplane to climb away. The probable cause cited descent mismanagement, the unintended loss of automatic speed control, poor airspeed monitoring and a delayed go-around (NTSB DCA13MA120). The NTSB also noted the airline's policy did not encourage manual flight on the line.

The lessons: a late go-around may lack the energy to work, and an unverified one may not be happening at all.

The pushback: are the criteria right?

If 97% of crews ignore a rule, the rule may be part of the problem.

FSF (2017) recommended separating the definition of "stable" from the go-around decision criteria. It also noted that the decision window extends past the stable approach height, down to thrust reverser deployment (FSF, 2017).
NASA Ames ran three simulator experiments with airline pilots. Touchdown performance changed little when the decision gate moved from 500 ft to 300 ft, but got much worse at 100 ft. The researchers proposed checkpoints at 1,000, 500 and 300 ft, with a mandatory go-around if criteria aren't met at 300 ft (Campbell et al., 2021).
ACT ARC Recommendation 24-2 (September 2024) went further. It asked the FAA to keep stabilized approach policy in SOPs. But it would turn the mandatory go-around into one crews should consider, keeping it mandatory only when a safe landing is in doubt. It also proposed guidance on a latest touchdown point. It noted that four of seven recent approach and landing accidents it reviewed followed a stable approach (FAA ACT ARC, 2024). This is a committee recommendation, not FAA policy. As of October 2026, we found no FAA guidance that adopts it.

Where policy stands in 2026

Regulators outside the U.S. are moving the other way: clearer gates, firmer go-around expectations, and explicit no-fault language.

UK CAA, August 2026. SN-2026/008 asks operators to write unambiguous stability criteria with defined decision gates. It also asks them to use flight data to find out why crews continue, and to train go-arounds from varied configurations and altitudes. It calls the go-around the normal and expected outcome when criteria aren't met. Crews should be assured the decision will be supported and never penalized (UK CAA, 2026).
IATA. Its guidance calls for a policy that treats the go-around as a normal maneuver, with no punitive response from management, while still following up on failures to go around. IOSA standard FLT 3.11.60 requires operators to have a policy that mandates a go-around when not stabilized (IATA, 2017).
GAPPRE. Recommendation OPS16 asks operators to let every flight crew member call a go-around at any time, unless an emergency dictates otherwise (GAPPRE, 2021).
EASA. EASA's Data4Safety program published a flight-data method in 2022 for detecting unstable approaches. It is built on about 1.4 million flights (EASA D4S, 2022). EASA's 2014 bulletin called for unplanned go-arounds that expose crews to startle, at varied weights and altitudes, including balked landings after touchdown (EASA SIB 2014-09). Guidance material adopted in 2019 lists a go-around at low gross weight using maximum go-around thrust as a simulator training item (EASA ED Decision 2019/005/R).
Operators. After the Darwin event, the E190 operator set a single 1,000 ft stabilization height for instrument and straight-in visual approaches. It kept 500 ft only for visual circuits and circling (ATSB, 2025).

What it means for pilots

Brief the go-around as the expected outcome. Don't brief it as a remote option. Cover the missed approach altitude, the first turn, the thrust mode your type will command, and who does what. The BEA found that capturing a close level-off altitude was one of the hardest tasks. Say that altitude out loud on the brief. The NTSB tells pilots to include missed approach procedures in every approach briefing (NTSB SA-077).

Make the gates verbal. At 1,000 ft and 500 ft, or your operator's gates, the PM calls "stable" or "unstable." IATA recommends repeating the call until the deviation is fixed and acknowledging every call (IATA, 2017). A spoken gate is harder to rationalize than a silent one. Watch for the Darwin thought that you will be stable by 500 ft. If your SOP says 1,000 ft for this approach, that is the gate.

Call it early. An early go-around from a high or fast approach is a low-energy event. A late one is not. Asiana's go-around came below 100 ft without the performance to climb. EK521's came after touchdown, when TO/GA no longer worked. Know your type's TO/GA inhibit logic and what you must do by hand if it doesn't respond.

Verify thrust and modes every time. Say the FMA changes and confirm thrust is increasing. EK521's crew pressed TO/GA and assumed the autothrottle had responded. Impact came 18 seconds after the go-around began, and the levers went forward by hand only in the last few seconds (GCAA AAIS, 2020).

Practice the light, all-engines go-around. The surprise on the line is a light airplane with full thrust: a strong pitch-up, a fast climb, and the level-off altitude arriving almost at once. EASA's guidance lists a low-weight, maximum-thrust go-around as a training item. Its 2014 bulletin points to reduced-thrust go-around modes where your type has them (EASA SIB 2014-09). Airbus guidance for its types is to select TOGA first, then go back to the CL detent if full thrust isn't needed (Airbus Safety First, 2014). Use your own type's procedure. Ask for unplanned go-arounds and balked landings in the simulator, not just the scripted one at minimums.

Use the CRM cues. Either pilot may call a go-around, and the NTSB says so plainly (NTSB SA-077). Captains, say it in the brief: "If either of us calls go-around, we go." In the FSF data, pilots who went around were about four times as likely to recall a prompt from someone else (FSF, 2017). FAA guidance tells operators to define when the PM escalates and takes control if callouts go unanswered (FAA AC 120-71B). Learn your company's escalation words before you need them.

Treat your own certainty as a warning sign. The pilots who continued felt more sure they could fix the approach. Feeling that it will be fine is the pattern the NASA researchers described.

Know your no-fault policy, and report. Read what your manual says about go-arounds. File the report when you go around. Those reports are how the next trend gets found.

Sources

This is general information, not operating guidance. Your company's procedures, your aircraft's manuals and the regulations come first.