Who will compete with SpaceX? The last two and next two years.

[Robotbeat]

Agreed. Probably only 8 or 10 more launches this year (assuming no failures). I expect they'll end with either Falcon Heavy launching or right before a FH launch attempt.

[garcianc]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

[envy887]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

There is no entirely correct way to calculate reliability, especially with this small a sample. But there are plenty of wrong ways to do it. In any case, the F9 rockets before, during, and after AMOS were all different, and it's different again now, and will change again in a few months. We have no way of knowing if Block 5 will be more or less reliable (at least for its first 10-20 launches) than the current vehicle, because we don't know what the changes are or whether they will introduce new unknown failure modes. SpaceX/customers/insurers know the changes, but not the effects, so they are only slightly better off than we are.

[AncientU]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

There is no entirely correct way to calculate reliability, especially with this small a sample. But there are plenty of wrong ways to do it. In any case, the F9 rockets before, during, and after AMOS were all different, and it's different again now, and will change again in a few months. We have no way of knowing if Block 5 will be more or less reliable (at least for its first 10-20 launches) than the current vehicle, because we don't know what the changes are or whether they will introduce new unknown failure modes. SpaceX/customers/insurers know the changes, but not the effects, so they are only slightly better off than we are.

This is so wrong I don't know where to begin...

[AncientU]

Historically SpaceX was limited by various issues, production rate, failures and are now limited by pad capacity. But what if they don't encounter another failure soon? Once LC40 and LC39A are both up they will be capable of launching almost weekly, and they actually have an *inventory of boosters* to support this. SpaceX launch rate will go up like crazy. So far this year they are already #1 by launches, responsible for fully a quarter of all successful orbital missions.

So far existing providers have felt a pressure to reduce costs but did not yet encounter issues finding enough payloads to launch. If SpaceX can keep increasing their launch rate then this will change.

This is what I've been referring to as taking these payloads off the table.  If SpaceX launches them, then someone else is out a more expensive launch.  To date, most everyone has had enough to keep busy -- aided by SpaceX bad days -- but this may be ending.  When launching their own constellation, they'll be working from a base of 50+ launches a year, and marginal launches could be offered free (but obviously won't be).

Building just 20 Block 5s next year could be sufficient to fly the World's payloads for the next decade.

[JamesH65]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

There is no entirely correct way to calculate reliability, especially with this small a sample. But there are plenty of wrong ways to do it. In any case, the F9 rockets before, during, and after AMOS were all different, and it's different again now, and will change again in a few months. We have no way of knowing if Block 5 will be more or less reliable (at least for its first 10-20 launches) than the current vehicle, because we don't know what the changes are or whether they will introduce new unknown failure modes. SpaceX/customers/insurers know the changes, but not the effects, so they are only slightly better off than we are.

This is so wrong I don't know where to begin...

I'll start. Initial assumption is that SpaceX change stuff for every flight. This is almost certainly true. BUT, its also true of all the other launch providers - there are changes to Ariane, or Delta for every flight. I know this because Jim said it a year or so back. So incorporating that to SpaceX probability without also taking it in to account for other providers is wrong.

Of course, we know that SpaceX are making changes for block 5 to improve reliability and reusability. So to say we don't know whether it will improve reliability is to basically saying the engineers don't know what they are doing, and Block 5 is a waste of time. We CAN assume that changes made specifically to improve reliability are going to actually improve reliability. Its not proven until they fly of course, but we can make that assumption.

[woods170]

Will we ever get to a point where a higher failure rate - on unmanned flights - becomes an acceptable price to pay for dramatically lower launch costs?

Say 5 failures in every 100 launches, but in exchange for an order of magnitude drop in launch costs?

5 percent is the overall current failure rate for orbital launch vehicles world-wide, expendable or not.  Falcon 9 has an 8 percent failure rate so far if AMOS 6 is included.  SpaceX will have to improve that rate no matter what it charges for a launch if it wants to compete with Arianespace and ULA, which have demonstrate 1-3% failure rates.

 - Ed Kyle

The better reliability of Ariane 5 compared to Falcon 9 comes courtesy of the former rocket having flown more often than the latter.

Now, let's do apples to apples:

- Ariane 5 suffered two (2) complete failures and two (2) partial failures in it's first 38 launches.
- Falcon 9 suffered two (2) complete failures and one (1) partial failures in it's first 38 launches.

So, comparing vehicle reliablity, based on a similar number of launches, the Falcon 9 is actually more reliable than Ariane 5 was back then. And that little fact becomes more remarkable considering that it took SpaceX just seven (7) years to do those 38 launches whereas it took 12 years for the first 38 launches of Ariane 5.

[AncientU]

Will we ever get to a point where a higher failure rate - on unmanned flights - becomes an acceptable price to pay for dramatically lower launch costs?

Say 5 failures in every 100 launches, but in exchange for an order of magnitude drop in launch costs?

5 percent is the overall current failure rate for orbital launch vehicles world-wide, expendable or not.  Falcon 9 has an 8 percent failure rate so far if AMOS 6 is included.  SpaceX will have to improve that rate no matter what it charges for a launch if it wants to compete with Arianespace and ULA, which have demonstrate 1-3% failure rates.

 - Ed Kyle

The better reliability of Ariane 5 compared to Falcon 9 comes courtesy of the former rocket having flown more often than the latter.

Now, let's do apples to apples:

- Ariane 5 suffered two (2) complete failures and two (2) partial failures in it's first 38 launches.
- Falcon 9 suffered two (2) complete failures and one (1) partial failures in it's first 38 launches.

So, comparing vehicle reliablity, based on a similar number of launches, the Falcon 9 is actually more reliable than Ariane 5 was back then. And that little fact becomes more remarkable considering that it took SpaceX just seven (7) years to do those 38 launches whereas it took 12 years for the first 38 launches of Ariane 5.

And Ariane 5 had the legacy of Ariane 1, 2, 3, and 4 to build expertitise, dating back to the 1970s.
  Spacex had Falcon 1...

[woods170]

ULA is retiring Atlas and Delta not due to competition, but because it is being forced to stop using RD-180.

There are multiple errors in your statement.

- Delta IV doesn't use RD-180. The reason to phase-out Delta IV is because it is too d*mn expensive. Atlas V is in fact it's most fierce competition.
- Atlas V will eventually be on it's way out, due to the use of RD-180, because of competition. Had SpaceX not existed, there would not have been a second certified NSS launch services provider. As such, US Congress never would have come up with the RD-180 ban given the need for redundant access to space for NSS and military purposes.

[woods170]

Delta is 100% being retired because it is not cost competitive -- nothing to do with RD-180.  The supply of RD-180s extends to the horizon due to political boosterism.  Atlas V as-is is not cost-competitive... and its renowned reliability is insufficient to overcome that fact.  As the new Block 4/5 come into play, the Atlas V 531/541/551 versions will be even less competitive because they are all going up against the same Falcon as the 401.

If Atlas 5 is not competitive, why does it keep winning new launches?  If Delta 4 is not competitive, why is it slated to most-likely out-last Atlas 5?  Did SpaceX even make a profit launching Bulgariasat for $60 million, or whatever it was (given the company's $1 billion cost to develop first stage recovery)? 

My guess is that ULA made money launching WGS 9, etc..

 - Ed Kyle

NSS and USAF launches (which are the bread-and-butter for ULA) are not about cost-effectiveness, but about "best value". Reliability is everything. Capabilities come second. The cost aspect comes as third (at best). That's why Atlas V (and Delta IV to a lesser degree) continue winning launches: the competition is yet to prove itself on reliability and capabilities.
I suggest you stop focusing on the cost aspect. "Competitive" in the NSS/USAF arena has a very different meaning than "competitive" in the comsat arena.

[AncientU]

Agree that 'competitive' in the NSS arena is not simply based on cost, as it also is in the commercial arena -- ask Proton.  Schedule performance plays a big role as does reliability (a.k.a., launch record), the dissimilar redundancy needed for launch vehicles, the yet-to-be-proven capability of Block 4/5 of F9, FH delays due to launch failures, and a backed-up manifest.  All of these combined are the basis of recent Atlas wins. 

One final factor is the bureaucratic inertia (read: Good Old Boys Club) that Jim continually polls, that, while betting on SpaceX failure, are propping up the established provider so they don't go toes up... 

At the highest levels, this is called "Managed Competition."  It came to fore with the Block Buy which gave a certain provider five years of breathing room against competition, was repeated with the 'warm pipeline' award of several Delta heavies stretching out to mid-2020s, and will continue into the 2020s with RD-180 waivers and awards based on 'best value' -- wink, wink...

[envy887]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

There is no entirely correct way to calculate reliability, especially with this small a sample. But there are plenty of wrong ways to do it. In any case, the F9 rockets before, during, and after AMOS were all different, and it's different again now, and will change again in a few months. We have no way of knowing if Block 5 will be more or less reliable (at least for its first 10-20 launches) than the current vehicle, because we don't know what the changes are or whether they will introduce new unknown failure modes. SpaceX/customers/insurers know the changes, but not the effects, so they are only slightly better off than we are.

This is so wrong I don't know where to begin...

I was only referring to their ability to predict expected failures based on that knowledge. Even with all the data available they can't predict unknown failure modes any better than a very simple historical analysis.

I'll start. Initial assumption is that SpaceX change stuff for every flight. This is almost certainly true. BUT, its also true of all the other launch providers - there are changes to Ariane, or Delta for every flight. I know this because Jim said it a year or so back. So incorporating that to SpaceX probability without also taking it in to account for other providers is wrong.

I wasn't making a specific comparison to other providers, but sure. It will be even worse for them when Ariane 6 and Vulcan start flying, as those are very big changes.

Of course, we know that SpaceX are making changes for block 5 to improve reliability and reusability. So to say we don't know whether it will improve reliability is to basically saying the engineers don't know what they are doing, and Block 5 is a waste of time. We CAN assume that changes made specifically to improve reliability are going to actually improve reliability. Its not proven until they fly of course, but we can make that assumption.

Block 4/5 is not just fixing reliability issues. We know there is a thrust upgrade coming from running Merlin harder. You can assume that's going to make it more reliable, but customers and insurers will assume its a potential failure mode.

[AncientU]

Another article:

Elon Musk’s SpaceX is out-launching its biggest competitors for the first time ever

This is what it looks like when a start-up disrupts an existing industry.

With its successful launch of a communications satellite last night, SpaceX has flown ten rockets this year. Elon Musk’s rocket company has now beat its own annual record, and the firm is on pace to out-launch its key rivals in the commercial launch market, Europe’s Arianespace and United Launch Alliance (ULA), a joint venture of Boeing and Lockheed Martin.

It has taken just seven years since the company’s first Falcon 9 rocket flew for SpaceX to dominate the launch market by offering cheaper and, increasingly, faster service to orbit.

https://qz.com/1022622/spacex-launch-elon-musks-rocket-company-is-out-launching-its-competitors-for-the-first-time-ever/

[ZachF]

Will we ever get to a point where a higher failure rate - on unmanned flights - becomes an acceptable price to pay for dramatically lower launch costs?

Say 5 failures in every 100 launches, but in exchange for an order of magnitude drop in launch costs?

5 percent is the overall current failure rate for orbital launch vehicles world-wide, expendable or not.  Falcon 9 has an 8 percent failure rate so far if AMOS 6 is included.  SpaceX will have to improve that rate no matter what it charges for a launch if it wants to compete with Arianespace and ULA, which have demonstrate 1-3% failure rates.

 - Ed Kyle

The better reliability of Ariane 5 compared to Falcon 9 comes courtesy of the former rocket having flown more often than the latter.

Now, let's do apples to apples:

- Ariane 5 suffered two (2) complete failures and two (2) partial failures in it's first 38 launches.
- Falcon 9 suffered two (2) complete failures and one (1) partial failures in it's first 38 launches.

So, comparing vehicle reliablity, based on a similar number of launches, the Falcon 9 is actually more reliable than Ariane 5 was back then. And that little fact becomes more remarkable considering that it took SpaceX just seven (7) years to do those 38 launches whereas it took 12 years for the first 38 launches of Ariane 5.

Also, soon SpaceX's competitors' (ULA, Arianespace, etc.) will be transitioning to lower-cost, unproven designs.. So the reliability card won't be one they can count on for much longer.

[JamesH65]

Of course, we know that SpaceX are making changes for block 5 to improve reliability and reusability. So to say we don't know whether it will improve reliability is to basically saying the engineers don't know what they are doing, and Block 5 is a waste of time. We CAN assume that changes made specifically to improve reliability are going to actually improve reliability. Its not proven until they fly of course, but we can make that assumption.

Block 4/5 is not just fixing reliability issues. We know there is a thrust upgrade coming from running Merlin harder. You can assume that's going to make it more reliable, but customers and insurers will assume its a potential failure mode.

Does that potential failure mode have more of an impact on their assessment than the reliability upgrades? I suggest, no it doesn't. So its still a win for SpaceX, and their customers.

AFAICA, the thrust upgrade isn't a change to the engine, but a change to how hard they push it, although my memory could be wrong. So there is always the option of not pushing it harder, and retaining the same level of reliability. Although I would also suggest that SpaceX are not going to make any changes that deliberately reduce reliability.

[AncientU]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

Wow, great article!  Thanks for making it available.

One gem on Outcome Bias (emphasis mine):

Near-misses offer evidence of both a system’s resilience (failure averted) and its vulnerability (failure
narrowly averted) (March et al. 1991). Therefore, a near-miss resembles a success in terms of ultimate
outcome, but resembles a failure in other details, such as problems arising before the outcome. However,
consistent with outcome bias (Baron and Hershey 1988), we believe that decision makers will evaluate
a near-miss like a success based on the ultimate outcome.  Thus, we expect near-misses to be judged in a
similar manner as successes and in a significantly different manner from failures despite the fact that such
near-misses would have been failures except for an element of chance.

Sounds familiar...

[ZachF]

It seems more and more like the strategy for SpaceX's competitors is "hope they get RUDs and we don't"

However, using this 'strategy' is going to have diminishing returns. ULA or Arianespace could have a RUD, and the business case for their higher prices disappears.

Falcon 9 could go on like Ariane 5 did at this point and not have another RUD for a decade, throwing up, and landing rockets like clockwork. Then in the 2020s Arianespace, ILS, ULA, etc are going to have to sell their more expensive launch services and unproven designs against a cheaper workhorse rocket with a high <weekly launch cadence.

Honestly Vulcan, Angara, Ariane 6, and Soyuz 5 are all probably obsolete before they even fly. They need clean-sheet multi-engine methalox launchers to field a design that's going to be competitive well into the 2020s. Like BO did. Anything else, if it's not obsolete before it flies, will have a life cycle too short to pay off the development costs.

The Russians are already in deep crap. They're losing the commercial market, They'll be losing ISS soon, and oil prices don't seem likely to get out of the doldrums. They've rested on their laurels and used the world-beating Soviet launch infrastructure it inherited and put little into the future. Rocket engineers who were young during the end of the USSR are now getting well into the range of Russia's poor male life expectancy.

I could see ULA giving up the launch business entirely in the 2020s.

Arianespace will probably end up limping along on European government charity launches.

SLS is just a pork buffet that will never do anything.

Blue Origin will probably be SpaceX's biggest competition. I could also see the Chinese deciding dump large amounts of funds into a reusable state champion launch service, but there is still a large gap to cross there.


It's pretty amazing the turnaround the US rocket business has had. Before SpaceX, circa 2009, the US launch business was falling way behind. We were pretty much about to sink into 4th place globally after the Russians, Chinese and Europeans. Now it's possible that the US could be launching three quarters of the world payloads into space by the early 2020s.

[ZachF]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

Wow, great article!  Thanks for making it available.

One gem on Outcome Bias (emphasis mine):

Near-misses offer evidence of both a system’s resilience (failure averted) and its vulnerability (failure
narrowly averted) (March et al. 1991). Therefore, a near-miss resembles a success in terms of ultimate
outcome, but resembles a failure in other details, such as problems arising before the outcome. However,
consistent with outcome bias (Baron and Hershey 1988), we believe that decision makers will evaluate
a near-miss like a success based on the ultimate outcome.  Thus, we expect near-misses to be judged in a
similar manner as successes and in a significantly different manner from failures despite the fact that such
near-misses would have been failures except for an element of chance.

Sounds familiar...

We also shouldn't discount the huge advantage of being able to land your boosters, inspect them, and compare observed wear rates/etc against the predicted/modeled ones, and using it to improve your design. This is an advantage that right now only SpaceX enjoys.

[envy887]

Crawling from under my rock for a moment... I usually do not feel qualified to participate in many of the discussions here, but this one is approaching my area of interest.

I have seen a couple of posts trying to relate past performance to reliability. That is not entirely correct. I saw one post mention the Columbia accident, which is germane to this discussion but IMHO to prove that historical success (or failure) is not a sign/confirmation of reliability. There is an interesting paper on hindsight bias and near-misses by Robin Dillon and Catherine Tinsley that relates to this topic: http://www18.georgetown.edu/data/people/tinsleyc/publication-40246.pdf.

One could argue that, prior to the AMOS-6 incident, the reliability of the Falcon 9 rocket was x. Following the failure investigation and accounting for all resulting corrective actions, the reliability of a Falcon 9 rocket now is >x. In other words, although it may sound counterintuitive on the surface, a failure (and subsequent corrective actions) usually result in increased reliability.

Wow, great article!  Thanks for making it available.

One gem on Outcome Bias (emphasis mine):

Near-misses offer evidence of both a system’s resilience (failure averted) and its vulnerability (failure
narrowly averted) (March et al. 1991). Therefore, a near-miss resembles a success in terms of ultimate
outcome, but resembles a failure in other details, such as problems arising before the outcome. However,
consistent with outcome bias (Baron and Hershey 1988), we believe that decision makers will evaluate
a near-miss like a success based on the ultimate outcome.  Thus, we expect near-misses to be judged in a
similar manner as successes and in a significantly different manner from failures despite the fact that such
near-misses would have been failures except for an element of chance.

Sounds familiar...

We also shouldn't discount the huge advantage of being able to land your boosters, inspect them, and compare observed wear rates/etc against the predicted/modeled ones, and using it to improve your design. This is an advantage that right now only SpaceX enjoys.

Ariane can and has recovered SRBs for inspection. But this in only an advantage if the operator actually implements changes and does not normalize deviance.

[Lars-J]

It seems more and more like the strategy for SpaceX's competitors is "hope they get RUDs and we don't"

However, using this 'strategy' is going to have diminishing returns. ULA or Arianespace could have a RUD, and the business case for their higher prices disappears.

Agreed. Betting that SpaceX is going to fail is a fool's strategy. Now even more so than in the past.

Will SpaceX eventually have another "RUD"? With how many flights they are wanting to fly in the next decade(s), very likely. (statistics and probability makes that clear) But with every success they have a stronger foundation for handling setbacks. They've had two F9 accidents, has that stopped them? They are stronger than ever now.