Hacker News Reader

Comment by hinkley | Hacker News Reader

There are a lot of big organizations that seem to understand something that few of us in the software industry ever did:

When you have one vendor they have no reason to give you any special attention. They are min-maxing you against all of their other customers and if you aren’t about to either Give them more money or take it away you fade into the background.

If you have two vendors, they have to pay attention. They have to and get to stay engaged.

If SpaceX is ahead or Boeing is ahead doesn’t matter. What matters if SpaceX is still at the table. If they drop to third or fourth place then they have a real problem.

NASA doesn’t want to give all their contracts to one vendor, and neither should we. That makes things easy this year but fattens then for the slaughter.

As long as Boeing is getting some noteworthy fraction of the contracts they stick around as a goad for SpaceX to try harder. And SpaceX does the same for Boeing.

View on HN

>If SpaceX is ahead or Boeing is ahead doesn’t matter. What matters if SpaceX is still at the table.

Not only that but SpaceX is delivering commercial payload after commercial payload to orbit while still TESTING new stuff on their rockets while they do all the planning for ITS/BFR... how many launches has Boeing been doing while designing SLS?

That may be useful for stuff you can afford to lose, but during the same period how many spacecraft have Boeing lost.

When a mission takes a decade to plan a build, you go with the launch vehicle most likely to get you to destination successfully. Fail rates matter immensely. You don't want a launch vehicle iterating on your launch.

Whether or not NASA (and by extension Boeing) can afford to lose a mission isn't dependent on how much it costs but on public relations and legislation.

Money has diminishing returns when it comes to mission success rates. A common ballpark example I've heard from JPL PMs: you can spend $500 million on one launch with a 95% chance of putting one satellite in orbit - or - you could spend $250 million per launch on two launches, each with a 80% chance of putting up one satellite. The probability of getting at least one satellite in orbit with the latter scenario is ~96%, higher than than the $500mil/launch option, with an 80% chance of saving 25-75% of the budget for the second launch (depending on scheduling and staff retention costs for that mission).

The problem with the latter option is that the public is swayed more by the absolute number of failures rather than the percentage, especially since NASA's missions usually draw a lot of global publicity. This kind of public pressure effects legislators and eventually starts changing the "cost-benefit" analysis agencies like NASA do until they're chosing the one launch option every time.

The SLS is inseparable as a project from NASA as far as the public is concerned. Even though SpaceX is also heavily funded by NASA historically, Musk's cult of personality has largely saved them from feeling the kind of public pressure NASA has had to deal with in their usual missions. That ability to experiment without PR consequences is invaluable but we'll have to see whether their failure rate is big enough to scare away NASA for manned missions.

The numbers are even better than you're throwing out there. Gotta keep in mind two big factors. A lot of the cost in these missions is not just the satellite hardware but the R&D costs. And the second is that launches are insured, so you'd need to compare the relative rate of insurance more than the expected value of fail rate x satellite cost. I have no idea on insurance costs for e.g. SpaceX vs Boeing launches but I'd be quite surprised if they're not extremely comparable.

But yeah, I think you're really hitting at the heart of the problem that's going to become even more troublesome once we start putting people on Mars. There are a practically unlimited number of ways for something to go wrong. And even if we do everything we possibly can, we're still playing an odds game with extremely high stakes. A sensationalizing media and an ill informed public creates a nasty mix for issues like this.

Insurance doesn't matter for this tradeoff, except maybe in the very short term. If they fly with a high failure rate, their insurance price will go up. At their price point, insurance companies pay attention.

I assumed the parent meant that you don't get the benefit of a lower failure rate until insurers have noticed a pattern of lower failures, so for rare launches you might not achieve a financial benefit.

When a mission is $250m, I wonder how much additional a second identical copy would cost. For example, if $200m of that is R&D, it would only cost $300m for two missions.

The marginal cost of an identical copy is almost always "just" the cost of assembly and testing - which can be astronomical if the mission is pushing the boundaries of technology like the Mars Science Lab or the NuStar x-ray telescope - but for most missions the major cost is R&D leading up to final assembly. For the majority of missions, that second $250 million goes towards engineering an extra 15% success chance for the satellite itself (success rate of the launch vehicle is known). The first $150 million gets you 80%, the second 250 million gets you an extra 15% with a flat $100 mil for launch.

If you budget for two 80% launches, however, $150-200 million gets you two satellites, you spend $125-200 million on launches (you cant wait until your first failure to put money down for the second launch so it's at least a 1.25x multiplier), and $25-100 million on extra staffing costs (you can't risk losing your top PIs/scientists between the first and second launch so you have to pay them the entire time).

Ideally, both options would be evaluated independent of popular sentiment. For manned exploration or cutting edge missions like the Mars rovers, we should strive for that 95% single mission success rate. Otherwise we should err towards getting more science done faster and cheaper.

He's probably referring to tests that don't significantly affect the probability of a successful mission. I don't know anything about how Boeing does this. But I can imagine a contractor going another route where they offer to come up with a new technology and then make the customer pay for all of the testing.

Why kill two birds with one stone when you can get them to pay for you to throw two stones?

+ the successful landings of the early stage reusable rockets. Pretty monumental stuff.

This article isn't about SLS (or BFR). It's about commercial crew, Boeing's CST-100 capsule vs. SpaceX's crewed Dragon. Boeing may well be ahead because unlike SpaceX they consider themselves having one job, and one job only.

And without a reliable rocket to get them up there, those crew capsules are useless.

They have reliable rockets. The CST-100 can launch on any of the Atlas V, Delta IV, or Falcon 9 rockets. All of which are in commercial use.

I think maybe you're confusing with the Orion capsule?

The standard failure mode is that the two vendors reach an understanding and split the market at a comfortable level of effort and progress.

I wonder if it's an understanding of just a natural equilibrium found when two companies iterate on focusing where their money seems to be most successful at earning contracts. Soon enough one company is great at one thing and the other another.

I won't pretend to have studied this at any depth, but my possibly overconfident intuition is that it can be either an outright conspiracy or an unspoken understanding of not stepping on each others toes.

It doesn't matter much; you don't have to be a genius CEO to see what is in both companies interest, conspiracy or not.

The more companies are added to the mix, the less stable the truce becomes, and somewhere around 4-5 I think ruthless competition is empirically the norm.

You see this in Germany vs the US when it comes to grocery store. In DE there are vastly more grocery stores and more grocery store companies. Prices are crazy low. In the US, there is very little overlap, with mega-stores carving out a multiple square mile section of the town.

FWIW, grocery chains in the US have low single digit margins. I don’t think they’re gouging anybody. Also, prices at my local grocery store are frequently lower than at Costco (a warehouse-style store).

Gross Profit vs Net Profit. Gross Profits are in the 20-40% range, Kroger (22%) [1] Safeway (28%) [2] Whole Foods (34%) [3]

Those Net Profits of 1.5% look like a target, not a byproduct of a cut throat business.

Smart and Final, a warehouse and restaurant supply store [4] has a GP of 14%.

[1] https://csimarket.com/stocks/singleProfitabilityRatios.php?c...

[2] https://csimarket.com/stocks/singleProfitabilityRatios.php?c...

[3] https://csimarket.com/stocks/singleProfitabilityRatios.php?c...

[4] https://csimarket.com/stocks/singleProfitabilityRatios.php?c...

You need overhead to run a business. Pointing to GP is nice for a company that can expect overhead to stay the same while they reap the benefits of increasing scale, but all the large grocery chains in the US are already at scale. If there was extra net profit to be had by reducing overhead, they would have done so already.

Or they pull a ULA and merge.

Cartels are illegal in the US because of anti-trust laws?

You would think so, and yet...

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

Can this be summarized as "competition is good for the marketplace"?

Enterprise systems are always like that. You pay for it, then you are invisible and communicate through support tickets.

Reply on news.ycombinator.com