What is the failure rate of launchers if you exclude all flight prior to the first success?

[ringsider]

Just wondering if the overall failure rate is skewed by including the initial test flights before a launcher achieves reliability.

[MattMason]

Your answer would require a lot of study to gather a succinct answer.
Not impossible, but there are several factors to consider, in my opinion.

1) Vehicle origin. The first modern (liquid fueled) rockets built since 1943 were missiles, suborbital weapons. To compensate for reliability, both the US and former USSR built many of them. When some of these missiles were selected to become launch vehicles to push payloads into orbit or into direct ascent trajectories to the Moon or beyond, it took a lot to tame some of them or develop technologies to keep them stable. The original Atlas SM-65 is a notable example since it was used for the first Mercury orbital missions. Korolev's R-7 had similar woes as it moved from missile to Vostok, Voshkod and Soyuz launch vehicle. When rockets built specifically as orbital launch vehicles were designed, like the Saturn and Centaur upper stage, reliability was easier to attain as development allowed more integrated testing from the start.

2) Vehicle lifetime. Obviously, the longer a rocket remains in service, the higher the launch rate and success statistics that follow.  This is where I won't do the math because I'm bad at math and answers can't always be free. There are many orbital-class launch vehicles since 1957, using the start of the classic Space Age as an arbitrary starting point. You could use Wikipedia and other sources to gather the reliability record of the original Atlas (the most commonly used rocket for uncrewed spacecraft, especially with the Centaur upper stage), and its successors to the name, right up to the 100% reliable Atlas V (which shares only the name, and little of the tech save an advanced Centaur). The Delta/Thor family, Titan II, Titan III are also things to gather. You can skip the Saturn family; it had a 100% reliability (from launch, at the least) on all flights. The R-7/Soyuz, Ariane I through V also have histories which have improved over time.

Commercial launch companies that were not part of a government/armed service program at first, such as SpaceX, Astra, Rocket Lab, Firefly, Armadillo and others have a more traditional level of failure to success similar to early missile to launch vehicle development. SpaceX had 4 consecutive failures before success. Many have slower rates as they haven't launched as much.

[ringsider]

Your answer would require a lot of study to gather a succinct answer.
Not impossible, but there are several factors to consider, in my opinion.

Agreed, it would take some study. What raised the query was, for example, sources like Gunther's Space Page listing early suborbital flights by Astra as "F" for failures. That has to skew a study of launch success rates unfavorably.

[MattMason]

Your answer would require a lot of study to gather a succinct answer.
Not impossible, but there are several factors to consider, in my opinion.

Agreed, it would take some study. What raised the query was, for example, sources like Gunther's Space Page listing early suborbital flights by Astra as "F" for failures. That has to skew a study of launch success rates unfavorably.

Gunter's Space Page is a great resource.

Data is data, so listing all the launches is just the facts. If there are 4 failures before success, that's the way it was, and provides the answer. In fact, great idea to use Gunter's as a good place to start, since it tends to group launch history by family already as tables there and notes success or failure. You might just be able to copy and paste the tables into a spreadsheet to quickly get the  answers.

If the orbital rocket was just a test not meant to reach orbit, that's fine. You'd just have to set a variable that notes the objective and whether it succeeded in doing it.

[tleski]

SpaceX had 3 consecutive failures, not four.
https://en.wikipedia.org/wiki/Falcon_1#Fourth_flight

[edkyle99]

SpaceX had 3 consecutive failures, not four.
https://en.wikipedia.org/wiki/Falcon_1#Fourth_flight

Three launch failures, yes, but they also destroyed a launch vehicle, or at least a stage, on the ground during propellant detanking.

 - Ed Kyle

[trimeta]

SpaceX had 3 consecutive failures, not four.
https://en.wikipedia.org/wiki/Falcon_1#Fourth_flight

Three launch failures, yes, but they also destroyed a launch vehicle, or at least a stage, on the ground during propellant detanking.

 - Ed Kyle

Including failures during ground handling would significantly increase the difficulty in collecting statistics for the past 70 years of rocketry.

Which doesn't make it the wrong thing to do, it just highlights the importance of understanding exactly what question we're trying to answer.

[edkyle99]

Just wondering if the overall failure rate is skewed by including the initial test flights before a launcher achieves reliability.

Skewed only slightly, I think.  It really is true that about half of the new launch vehicle variants succeed on their inaugural flight.  That leaves the other half, which would then see their total failure (and launch) numbers cut by one (or two, or three).  Thus Delta 4 Heavy would be 12/12 instead of 12/13, Electron 18/20 instead of 18/21, the Ariane 5G series would be 22/23 instead of 22/25, PSLV 50/52 instead of 50/53, and so on.

 - Ed Kyle 

[trimeta]

Skewed only slightly, I think.  It really is true that about half of the new launch vehicle variants succeed on their inaugural flight.  That leaves the other half, which would then see their total failure (and launch) numbers cut by one (or two, or three).  Thus Delta 4 Heavy would be 12/12 instead of 12/13, Electron 18/20 instead of 18/21, the Ariane 5G series would be 22/23 instead of 22/25, PSLV 50/52 instead of 50/53, and so on.

 - Ed Kyle

The real question is "how many rockets failed on their first 1-3 launches and then didn't go on to have a string of successful missions"? Because for those, the initial failed launches wouldn't be a drop in the bucket -- they'd be the whole bucket. And maybe knowing what fraction of rockets (not individual launches, but rocket designs) fall into this "failed and never became successful" category would give a sense of the probability that future vehicles will as well.

[edkyle99]

The real question is "how many rockets failed on their first 1-3 launches and then didn't go on to have a string of successful missions"? Because for those, the initial failed launches wouldn't be a drop in the bucket -- they'd be the whole bucket. And maybe knowing what fraction of rockets (not individual launches, but rocket designs) fall into this "failed and never became successful" category would give a sense of the probability that future vehicles will as well.

My retired launch vehicle reliability list shows eight out of 59 vehicles listed ended up more or less "still born", a rate of 13.6% of the total.  That would be 27.1% of those that suffered an inaugural failure.  Just a snapshot, but it  suggests that there is a decent chance that Firefly or Astra succeeds - and a slightly better than 50-50 chance that *both* succeed.   
https://www.spacelaunchreport.com/log2021.html#rate

 - Ed Kyle

[Steven Pietrobon]

You can skip the Saturn family; it had a 100% reliability (from launch, at the least) on all flights.

Note that the Saturn V on Apollo 6 had major problems with the second and third stage, such that they could not relight the S-IVB engine once it got to LEO. I would class this flight as a partial failure.

https://en.wikipedia.org/wiki/Apollo_6

[Yggdrasill]

Another complication here is that *commercial* launch vehicles is a relatively new thing. Government-funded launch vehicles arent under as much pressure to succeed right of the bat, as funding isn't an immediate and ever-present concern.

Though maybe it doesn't make much difference, with how easy it is to get a market valuation of billion of dollars, these days.

[TrevorMonty]

Another complication here is that *commercial* launch vehicles is a relatively new thing. Government-funded launch vehicles arent under as much pressure to succeed right of the bat, as funding isn't an immediate and ever-present concern.

Though maybe it doesn't make much difference, with how easy it is to get a market valuation of billion of dollars, these days.

Lot new government LV borrowed flight proven systems from existing LVs which help reduce failure rate. The new commercial small LVs tend to be built from scratch, plus teams aren't as experienced so higher probability of a failure. Their next generation LVs should have lower probability of failure during maiden launch. SpaceX has proven this theory correct but they are only ones to have built 2nd LV.



Sent from my SM-T810 using Tapatalk

[RoadWithoutEnd]

It's worth reflecting on exactly what "prior flight" means.  It's almost a taxonomic issue, because of extended rocket families.  Like in human evolution, if you asked what the extinction rate of hominid lines is, the answer is "approaches 100%", and yet the world is currently crawling with one that has significant DNA contributions from several. 

Given the need for some arbitrary limits, looking at local peaks could be more useful.  E.g., weighting against heritage, so that real technical differences stand out more and are less hidden in the noise of management, policy, distant economic factors, and so on.

It would be real work to figure this out in any case.  Intriguing topic though.

[edkyle99]

Another complication here is that *commercial* launch vehicles is a relatively new thing. Government-funded launch vehicles arent under as much pressure to succeed right of the bat, as funding isn't an immediate and ever-present concern.

That may have been true once upon a time (see early Scout or Atlas-Centaur), but I'm not so sure about today.  How many early failed SLS launches will NASA tolerate?  How many exploded Vulcans or Falcon Heavies will Space Force accept?  Space Shuttle was one of the most reliable launch vehicles of its time, but two catastrophic failures out of 135 launches was too many in the end.  Compare that with five failures during the first seven Atlas-Centaur attempts back in the day.

Ultimate success is about fortitude and funding.  NASA poured money into Atlas Centaur during the 1960s, shifted project management, etc., because the Agency wanted the rocket that ultimately gave us Surveyor, Lunar Orbiter, Mars orbiting Mariners, and the first Jupiter and Saturn Pioneers.  Elon Musk rode out the early Falcon 1 failures because he was committed to the goal and we see what has resulted.  Astra is enduring similar circumstances now.  We'll see.

 - Ed Kyle

[libra]

Ariane 5 between 1996 and 2003 had pretty abysmal beginnings - the first sixteen flights reliability rate was pretty appalling, as can be seen on Gunter space page (attached screenshot).

Two failures and two partial failures, that four "bad days" over sixteen flights: a 75% reliability rate or even less. Coming from Ariane 1 - 2 - 3 - 4 rather impeccable record, that was a VERY brutal awakening for Arianespace, ESA, CNES, and contractors altogether.

And then they reacted strongly, with stellar results: not a single failure since 2003, only a partial one in 2018.

Ariane 1 - 2 - 3 - 4 failures are a bizarre, mixed bag. It is quite straightforward: HM-7 failed in 1982, 1985, 1986, and twice in 1994 at both end of the year (talk about a bad year: January 24 and December 1).

So HM-7 can be blamed for 5 out of 7 failures of "the Arianes before Ariane 5".

Which mean in passing that Ariane 1-2-3-4 lower stages only had two failures out of 144 flights.
Sounds familiar ? somewhat ironically, it looks like a carbon copy of the Shuttle own flight reliability: 2 failures over 135 flights.
https://space.skyrocket.de/doc_lau/ariane.htm

(And yet, Ariane 1-2-3-4 never got a manned capsule when Ariane 5 was to have Hermès. The irony is pretty strong, even more when considering Ariane 5 catastrophic beginnings in 1996-2003. Hermès was to have flown right there - not very reassuring with perfect hindsight ! )

As for the other two, they are worth mentionning.

Flight 2 in May 1980 was lost to POGO.

And finally, the other one: the "cloth of doom", one of the most absurd and "head scratching" failures in rocketry history.

Narrated right here, by yours truly (blatant, shameless self promotion - entirely assumed).
https://www.thespacereview.com/article/4085/1

In the end, "Ariane 5 beginnings" and "that goddam HM-7" all by themselves are responsible for nearly every single Ariane failures in history.
Plus POGO and a cloth.

On top of that, as if the HM-7 reputation wasn't already bad enough, the very stage used to launch Spot 1 in 1986 and abandonned thereafters, exploded in (polar) orbit - and a decade later one of its debris shot down CERISE, a French satellite. The Murphy Law as its best: launched by an Ariane, shot down by another. Frack.  https://en.wikipedia.org/wiki/Cerise_(satellite)

[Steven Pietrobon]

Another complication here is that *commercial* launch vehicles is a relatively new thing.

Commercial launch vehicles have been around since 1990.

https://www.northropgrumman.com/space/pegasus-rocket/

"World’s first privately developed space launch vehicle."

[Yggdrasill]

Another complication here is that *commercial* launch vehicles is a relatively new thing.

Commercial launch vehicles have been around since 1990.

https://www.northropgrumman.com/space/pegasus-rocket/

"World’s first privately developed space launch vehicle."

Yes, and the first rockets were launched into space in the 1940s. So that's five decades without any commercial launch vehicles.

[Hog]

Another complication here is that *commercial* launch vehicles is a relatively new thing.

Commercial launch vehicles have been around since 1990.

https://www.northropgrumman.com/space/pegasus-rocket/

"World’s first privately developed space launch vehicle."

It was difficult to have a commercial space industry(in the USA) when space launches were mandated to use STS(shuttle) and such launches were subsidized by the gov.   1986 set about major changes in US space launch efforts.

[Jim]

Another complication here is that *commercial* launch vehicles is a relatively new thing. Government-funded launch vehicles arent under as much pressure to succeed right of the bat, as funding isn't an immediate and ever-present concern.

Not true.  Government launches are self insured and commercial ones use insurance.

[Yggdrasill]

Not true.  Government launches are self insured and commercial ones use insurance.

I doubt most early test flights with limited or no payload are insured. If they are, the insurance would at least have to be extremely expensive. Like 50+% of the cost of the launch.

[Jim]

Not true.  Government launches are self insured and commercial ones use insurance.

I doubt most early test flights with limited or no payload are insured. If they are, the insurance would at least have to be extremely expensive. Like 50+% of the cost of the launch.

Wriong, see the payloads for the 1st Delta IV or Atlas V.  Just discounted launch service price.

[Yggdrasill]

Wriong, see the payloads for the 1st Delta IV or Atlas V.  Just discounted launch service price.

I'm sure some first/early launches have insurance, but I doubt *most* are insured. Do you think Astra has gotten $10-20 million for their three failed Rocket 3 launches? How much do you think they paid for that insurance?

Do you think Firefly has a check in the mail for $15 million? How much did they pay for that insurance?

How much did SpaceX get for their first three Falcon 1 launches?

[Kryten]

 Launch insurance is for payloads.

[Yggdrasill]

Launch insurance is for payloads.

That was my understanding as well, but I am a little fuzzy on the details. My understanding is that the launch costs can be covered by insurance as well, but if it pays out to the owner of the payload, how does that affect the launch provider? I would guess the launch provider would have to give the owner of the payload their money back, or a new launch. But in that case the insurance shouldn't pay the owner of the payload for their losses on the launch, as they have none. This would mean the insurance doesn't protect the launch provider at all.

Generally though, it's always possible to buy insurance. It's just a question of paying enough. Like if there's a test launch with an estimated 50/50 chance of failure and a cost of $10 million, the insurance company should require a fee of something like $6 million or more to insure the launch. The expected average payout is $5 million, which means they would make $1 million, on average. For the launch provider, they have spent $6 million if the flight is a failure, but $16 million if the flight is a success.

This *could* make sense for a company, even though it's expensive. Say they have $16 million to spend, but if they have a successful launch they could take the company public and raise hundreds of millions. Without insurance, they only have enough money for one launch. If it fails, they're bankrupt. So they're betting the company on a 50/50 shot. While with insurance, if the launch is a success, they are out some money, but the company survives. And if it is a failure, they can actually afford a second launch attempt with the insurance money.

But to back to my point, government self insurance vs commercial insurance doesn't provide commercial launch providers with an economic benefit, like Jim suggested. In the best case, where one assumes they actually insure their test launches, they are paying huge sums of money to the insurance companies. In the event of launch failures, they will still be bleeding money, and their their reliability records will also gradually get worse, which isn't great for attracting well-paying customers.

If a self-insured government launch fails, there's usually still money in the budget for next year. The budgets might even increase, to try to avoid embarrassing launch failures in the future. The taxpayers are an endless source of money, as long as the politicians don't pull the plug.

[joek]

Launch insurance is for payloads.

That was my understanding as well, but I am a little fuzzy on the details. My understanding is that the launch costs can be covered by insurance as well, but if it pays out to the owner of the payload, how does that affect the launch provider?
...

Insurance may cover anything from payload only, to replacement launch cost, to payload on-orbit life or achieving specific objectives... or whatever else the owner wants to throw in the bucket. Some are outside the control of the launch provider and would not be considered part of "launch insurance" (e.g., on-orbit life). Some may be negotiated; e.g., does launch provider provide replacement launch at their expense, or is that part of the insurance?

If we narrow this down to "did the launch provider get the payload to intended destination-orbit?" then the equation becomes somewhat simpler. From an insurance perspective, the primary factor is going to be: how much have the insurer(s) had to pay out for a particular launch provider or launch vehicle? That directly impacts the price of launching on that provider.

However, that simple equation elides factors such as the cost of delays (and loss of revenue) due to a launch failure. Again, insuring those would be part of how insurer's price their offering, and thus impacts the total cost of using a given launch provider. Hope that helps clarify a bit.

[Jim]

1.  how does that affect the launch provider? I would guess the launch provider would have to give the owner of the payload their money back, or a new launch. But in that case the insurance shouldn't pay the owner of the payload for their losses on the launch, as they have none. This would mean the insurance doesn't protect the launch provider at all.

2.  But to back to my point, government self insurance vs commercial insurance doesn't provide commercial launch providers with an economic benefit, like Jim suggested. In the best case, where one assumes they actually insure their test launches, they are paying huge sums of money to the insurance companies. In the event of launch failures, they will still be bleeding money, and their their reliability records will also gradually get worse, which isn't great for attracting well-paying customers.

3.  If a self-insured government launch fails, there's usually still money in the budget for next year. The budgets might even increase, to try to avoid embarrassing launch failures in the future. The taxpayers are an endless source of money, as long as the politicians don't pull the plug.

1.  The launch provider gets nothing and pays nothing from the payload insurance.  He doesn't need protection, he got paid for his services.   Unless the launch provider got its own insurance, but why bother.

2.  Test launches aren't insured unless there is a paying payload.

3.  No there isn't.  It has to go through budget cycles to get a replacement added back in.