Rockets Blow Up... How Often?

On reading this article about a SpaceX rocket engine blowing up during testing, I thought to myself that rockets sure seem to blow up a lot, even with 60 years of experience.   If I had had to guess, I probably would have guessed 10-20% of the launches fail.

But it turns out that this is an exaggeration, probably due to the summer of the shark effect I have discussed before.  One problem with the data is that I would define failure rate from a customer point of view -- did my payload survive and was it inserted into the right orbit where it can do what I want it to do.  A lot of the data on failure rates uses other bases.  From this page, in 2016, the failure rate by my definition would be 3 out of 86 launches, or 3.5%  (that site reports a failure rate of 2 out ob 85, but do not count the pre-launch explosion of a SpaceX rocket that destroyed the payload).    This page has answers to the failure rate question in the 6-8% range for unmanned rockets over the history, with a bit of a trend downwards recently.  Apparently the failure rate on manned launches is much lower.

I guess my reaction is that the failure rate is lower than I would have guessed, but I think my perception that it had not improved a lot over 50 years may be correct.  I don't have the data but my sense is that air travel experienced a much faster rate of improvement in catastrophic failure rates, though the engineering calculus between manned (most air travel) and unmanned (most rockets) travel may be different.  Certainly if it cost $100 million per rocket to reduce the failure rate by a percent or two, it might not make financial sense if there are no people on it.


  1. Steve Burrows:

    For the last 60 years the only goal was to maximize payload delivered, which would be ordinance. We want our missiles to deliver as much punch as possible at a reasonable, or predictable failure rate. Now with the race to develop fully reusable systems, the clock has just started for failure rates to improve substantially as the price per launch diminishes, demand will increase leading to higher launch rates and faster iteration of designs. I think it will be most interesting to see if we are at a "knee of the curve."

  2. Bram:

    Chemical rockets are a terrible technology. Riding a Hydrogen / Oxygen explosion as a form of propulsion is always going to be a risky proposition.

    An intelligent species would have set to work on a space elevator decades ago. Once in space, alternatives like nuclear rockets are a better bet.

  3. Dan Wendlick:

    During the Apollo era, the estimated design probability of a catastrophic loss of vehicle and crew was 10%. For the shuttle program, that dropped to 1%. For the constellation project, that margin is (was) effectively zero. This answers the question of why we could go to the moon in the late 1960s, but can't get to low-Earth orbit with what should be better technology today.

    On the unmanned side of things, transitioning to non-government launchers has changed the design goal of maximum performance for budget to maximum profitability. Lose too many boosters and the insurance rates will cut into that profit margin. That's also where reusability has come to be so important. Even at a 30% economic recovery rate, i.e. you can refurbish a booster for 70% the cost of building new, you can undercut the profitability of throw-away systems to the point of obsoleting them on an economic basis.

  4. Dan Wendlick:

    Aside from the fact that there is as yet no material on the planet with the requisite tensile strength that can be manufactured in the required quantities, building a space elevator would still require a lot of rocket launches to get the materials into space. For a variety of reasons, you can't build a space elevator from the ground up, you need to build it from orbit down.

  5. Aggie -:

    You'll talk about anything in order to avoid updating us on your train set, won't you? :)

  6. Aggie -:

    (Actually I find that the progress being made by the private sector, especially the self-landing rockets on ocean-going barges, to be pretty remarkable achievement when compared to the efforts of NASA and its related mega-industry contractors - and long may it continue. I would like to see the US reinvigorate its space-exploration program).

  7. John_Schilling:

    Mature, well-managed space launch vehicles fail catastrophically about 2% of the time, with another 1% reaching some sort of orbit but not the planned one. We can shave about a percent off that with high standards of mission assurance, involving a few tens of millions of dollars in extra testing and documentation. Maybe a $100 million if we're talking about a mature, "well-managed" government contractor that has learned the fine art of cost-plus contracting and doesn't face any competition; mission assurance activities are an easy way to pad the budget if you really want. But SpaceX is competing for those contracts now, which helps.

    Whether that's worth the cost depends on whether you are launching one more generic telecom satellite or e.g. the only missile warning satellite we are going to have with a good view of North Korea in this decade. No, you aren't allowed to ask for a second missile warning satellite over North Korea, that would be wasteful.

    A mature, well-managed reusable space launch vehicle might be able to beat 1-2%; we haven't reached that point yet so we can't be sure. You're probably not going to do significantly better than 1-2% with any expendable system.

  8. sean2829:

    Having worked in the chemical propulsion industry for a time plus living in the DC area I know a couple of people tied to international arms treat verification. In the 1990's when there seemed to be a rocket failure on a monthly basis in the US, a fried of mine was responsible for verifying that the Russians destroyed their medium range rocket inventory. The US chose to blow theirs up on the ground using the self-destruct charges that line the solid rocket motors. The Russians chose to launch all 350+ missiles. There was not a single failure of the Russian rockets. So they can be pretty reliable but remember, these were solid fuel and not terribly large motors.

  9. Craig:

    Nobody is discussing the fact that most if not all commercial launches are insured. If two percent of the launches of $500 million telecommunications satellite fail, then the expected value of launch failures should be around negative $10 million. Let's double that for insurance company overhead and profits and then double it again to cover the satellite owner for lost revenue. So insurance should cost around 10% of the satellite value. That's pretty reasonable.

    Insurance companies should have pretty good statistics on launch failures.

  10. Mike Powers:

    Yes, if you look at the history of solid-fuel rocket deployment (including the vast number of such rockets used as weapons) then the failure rate of solid-fuel rockets is basically zero.

  11. Mike Powers:

    " mission assurance activities are an easy way to pad the budget if you really want."

    Customer's the one that asks for those, broski. The contractor generally figures that they know perfectly well how to do their damn job and that's why they were the ones asked to do it.

  12. Mr. Generic:

    'Ignition: An Informal History of Liquid Rocket Propellants' is an amazing read, btw. :)

  13. The_Big_W:

    Yes but the most noticeable solid rocket failure is the most remembered launch failure of all time....

  14. The_Big_W:

    Yep, government loves their mission assurances....

  15. jimc5499:

    I just started this. So far it is very interesting. Thanks for posting the link.

  16. kidmugsy:

    I suppose the failure rate depends partly on when you start counting from. This seems a reasonable place.

  17. mx:

    Launch insurance premiums (which also cover the first year of operation generally) are pretty low. Around 5% or so (, a historic low last year. That article mentions problems with the rates for prelaunch insurance (that's a separate insurance policy, because Reasons), which have apparently been set too low to cover the risks of static fire tests.

  18. JTW:

    And was caused by human error causing a launch that should have been scrubbed to go ahead because a bureaucrat considered the cost overruns on his watch had the launch been scrubbed to be worse than the possible loss of the shuttle when it blew up.

  19. JTW:

    One reason launch failures are so low is because most of the potential failures are caused in testing.
    That's why I don't care an engine failing during testing, testing tends to test things beyond their design parameters for a reason and that reason is ensuring that it holds up to AT LEAST the design parameters.

  20. BobSykes:

    The failure rates for crew-certified rockets is much lower. The shuttles were shut down when the failure rate reached two.

  21. marque2:

    Shuttle would be 2/135 or 1.5%

  22. Matthew Slyfield:

    The problem with news and rockets/space stuff:

    Failure is spectacular. Success is boring.

  23. Craig:

    Thanks. It's nice to have facts not intrude too badly on my guestimate.

  24. Dan Wendlick:

    I am referring to the design estimates. In every engineering project there is a balance of requirements. It's like the old joke that if the black boxes always survive plane crashes, why don't they make the whole plane like they make the black box? The answer is that if they did, the plane would be too heavy to get off the ground.

    Ensuring reliability either means more and more testing, or more and more redundancy. Testing adds to program time, and redundancy adds to vehicle mass. Both add to cost.

    And losing 2 orbiters in 135 launches is well within that 1% estimated probability.

  25. Q46:

    Bombs come in two groups: ones that are supposed to explode; ones that aren’t.

    Rockets are in the latter group.

    Their are failures in both groups.

  26. LoneSnark:

    The difference is due to expend-ability. The system for the past 50 years has presumed that the rockets crashed into the sea after every use. The resultant design decisions presume any failure that destroys the vehicle should just destroy the vehicle. Jet Engines go on reusable vehicles, and therefore have shields to contain explosions and turbine failures and fire suppression systems to put out any fires. Rocket engines have none of these things, because explosions and fires only destroy a disposable rocket. For SpaceX's new rocket design, the BFR, they have already said the rocket engines will have shields to contain explosions and fire suppression systems. These safety features will be stock, even on cargo only vehicles, because these vehicles will be fully reusable and therefore worth putting the extra effort into making them failure resilient.

  27. Philip Ngai:

    The failure rate depends to some extent on the mission. For example, the US Navy doesn't like failures.

    So their Trident II D5 missile has completed 165 successful test flights.

  28. marque2:

    I would guess the design goal of 99% of all Rockets is to blow up 99% of the time. Note that devices that carry people and cargo into space are technically missiles, since rockets much like ballistics aren't guided once they are fired. Even so, I would entertain that 99% of all missiles are also designed to blow up or destroy themselves kinetically on some target.

    As for airplanes the primary goal is not to crash and RTCA and FAA have many guidelines to achieve this goal - and it is working, in the 70's there were several crashes a year in the US, today, commercial aircraft crash on the order of once a decade in the US. But in the event of an emergency landing (crash) is not the survivablility of the aircraft, but the survivablility of the passengers that is the goal, so yes, I have heard the joke, put it isn't applicable. A plane made out of a Northrup black box, would probably kill all the passengers in a crash as well.

  29. marque2:

    I don't know, they certainly didn't survive at a 99% rate. They all blew up :P

  30. marque2:

    Rockets are a misnaming by the media, much as Drone is not really an autonomous, or remote controlled aircraft. Drones, and Rockets can't be guided, once they are shot off, so you need to be careful with angle(s) of launch and take things into account like wind speed before actually firing. Missile are guided during flight and have flight controls (and occasionally groups of small charges) that allow for the guidance. The moon "rocket" was actually a missile.

    As for which is allowed to explode. Almost all rockets are designed to explode. Almost all missiles are designed to explode, or at least destroy themselves (note missiles like the Patriot 3 are actually kinetic, they don't work by exploding, but by directly crashing into the opposing vehicle - yes there is a "lethality enhancer" built on the device, but it was mostly because the military didn't believe, at first, kinetic missiles could work.)

  31. FF10:

    I think it may help if you split your analysis by looking at existing engines (over 2-5 launches) vs new (developmental) engines. I believe you may find that due to the complexity of new engines, failure occur a great deal in the development and initial flights, but once it is tested and the problems identified and fixed, the failure rate drops significantly.. As there are a number of new rockets in development, it appears that there is a higher rate of failure then the long term review would indicate.

  32. jimc5499:

    Thank you for an interesting read. I got a kick out of the author saying that a certain fuel mixture should work for a certain purpose, but, the engineers are afraid of it. I'm a Mechanical Engineer that used to work for a company that made chemical analysis equipment. I worked with some Chemical Engineers and they played with things that scared the crap out of me.

  33. Mr. Generic:

    I heard about the book from Derek Lowe, who writes a pharma chemistry blog. His "Thing I won't work with" tagged posts are frightening.