The Dual Optimization of Air Freight: How Many iPhones Fit in a 747? | Hacker News Reader

The Dual Optimization of Air Freight: How Many iPhones Fit in a 747?

learn.flexport.com

75 pointsthedogeye52 comments

The good news is that Apple certainly isn't just shipping iPhones and nothing else. It wouldn't surprise me (if we hold the assumption that Apple charters planes themselves for freight) that Apple also mixes their freight (i.e.: heavy iPhones and lighter 'other stuff') precisely to optimise this problem. That's Cook's bread and butter, after all.

For other operators of lower volume, of course outsourcing that problem represents a win-win for both customer and supplier. The customer gets overall lower costs of freight and the supplier leverages their breadth of sources to make a cut of the difference they can save. It's a good example of where outsourcing certain tasks can make sense for some parties but not others.

I wonder what the lighter 'other stuff' would be from Apple. Store window displays and boxes of software? Do they still sell software in boxes?

Little things like cables, headphones, and other accessories that don't take up a lot of weight but could be packaged voluminously to fill the remaining space.

Traditionally, in the airline industry, the answer was "tulips". Out of China it looks like it might be roses: http://www.nytimes.com/2006/09/25/business/worldbusiness/25f...

I'd venture that the odds of Apple having a product that is the best available match for its cargo is very low. The more scale and diversity you have, the better your mix. There are real economies of scale in air freight resulting from this.

This is, of course, just The Knapsack Problem: http://en.wikipedia.org/wiki/Knapsack_problem

It's proven to be NP-Hard, but with a well known "good enough" approximation ("the knapsack problem has a fully polynomial time approximation scheme (FPTAS)")

Actually that's not quite correct. The knapsack problem has 2 constraints: weight and (traditionally) "utility." This problem has 3 constraints - weight, volume and utility. That probably makes the problem more tractable.

It's not the knapsack problem. The task is similar: maximize (some relation between) the carried volume and weight. But instead of a finite collection of different items, you have an essentially infinite set of a few items. Also, the weight and volume of each individual item is negligible, so you don't have to worry about the integer nature of the problem, i.e. it's sufficient to give a fractional answer and round down. (That's what makes the knapsack hard - the fact that you can't add fractional items.) This seems like a straightforward linear programming problem, solvable via the simplex method.

Yeah linear programming was my first thought, too. How is that "one of the best kept secrets in the air cargo industry"?

> How is that "one of the best kept secrets in the air cargo industry"?

I guess it's only part of the problem, solving for one plane.

In practice you would want to solve for varying demand, for multiple customers, for planes that eventually are delayed, crews that get sick/delayed, for different planes, for routes that pick/drop different cargo at different places.

And of course, you want to price all of that in a way that makes you money.

It is not a secret. It is a basic day to day task to optimize this. If you didn't you'd be out of business very quikcly.

In addition to the other comments, keep in mind that this is such an enormous knapsack that dumping in items at random is just as efficient as any other method.

So at the scale of Apple, I wonder if it would make sense to actually own 747s and sell the remaining space themselves.

I guess the problem is that they must have a large spike in demand around launch dates; perhaps not enough the rest of the time to make all the overhead of operating the fleet worthwhile.

Precisely why they charter flights. In fact, if I recall correctly, there was a global lockup of all available freight flights out of China going to Apple and no one else during the last iPhone spike. Another competitive advantage for Apple.

They could operate their own aircraft for typical shipment volumes, then augment that by paying Fedex/UPS/etc during spikes.

I don't know what the margins are on air freight, but it seems like there's enough competition to keep them low. Also, Apple is big enough to negotiate better deals than typical customers. For those reasons, I don't think it's worthwhile for Apple to do it themselves.

For sure; my meaning was that since their volume wouldn't be as high typically, it may not be worth operating the fleet (whereas if every day was launch day, perhaps it would be). Basically I agree with you.

Possibly from a purely financial perspective, but unlikely from a holistic perspective.

The headaches involved with owning aeroplanes would be immense. They're large expensive heavily-regulated mechanical beasts.

    > So at the scale of Apple

I suspect Apple's "scale" is a rounding error of air freight volume, and unlike - say - retail, there's very little consumer value-add for owning your own freighting service.

The unit used in the air cargo industry is the RTK -- revenue tonne-kilometer.

On PVG-ANC, the route by which iPhones enter the US, a single 77-tonne load of iPhones (what the article estimated) is around 530,000 RTK. Boeing estimates that in 2013, worldwide air cargo was 208 billion RTK. In other words, that plane-load of iPhones is around 25 ten-thousandths of a percent of global air cargo capacity.

So yes, iPhones are absolutely a drop in the bucket. Just saturating the existing scheduled cargo capacity into the US for a few days would probably exceed Apple's manufacturing capabilities.

I'd love to chat about how you calculated their share for my upcoming post on Apples share of transpacific freight. Email me if you'd be down?

Ryan@flexport.com

PVG-ANC = 6,933 km great circle distance.

77 tonnes of cargo on that route is 6,933 * 77 = 533,841 RTK.

Boeing reports[1] 208 billion RTK worldwide in 2013.

533,841 / 208,000,000,000 = 2.566543269230769e-06. So that's the percentage of worldwide air cargo a 77-tonne load on a PVG-ANC cargo flight would represent.

[1] http://www.boeing.com/assets/pdf/commercial/cargo/wacf.pdf

I think the hang up here is on Apples scale. That single shipment of 300k phones weighs just 77 tons. But Apple sold 74.4M phones last quarter of which about 30% went to North America.

So, go with 77 tonnes = one 747 = 300k iPhones.

30% of 74.4m iPhones would be 22,320,000 iPhones. Divide by 300k (which is a planeload) and you get roughly 74 planeloads of iPhones per quarter.

Currently there are about half a dozen 747 freighters each day flying PVG-ANC. So at most it's about 12 days' capacity of that one route each quarter, meaning it tops out at 10% of the overall capacity on the route each quarter.

And it still is a drop in the bucket compared to worldwide air cargo. Remember one planeload of iPhones is 533,841 RTK. We're talking now about 74 per quarter, which is 296 planeloads/year. 296 * 533,841 = 158,016,936 RTK represented in iPhones coming to the United States in a year.

And global air cargo is 208,000,000,000 RTK/year. Divide 158,016,936 by that, and you get a figure of: 0.0007596968076923077. So if we add up a reasonable estimate of all the iPhones that come to the US in a year, and a reasonable estimate of how many 747s it takes to haul them from China, we get a figure that is... just over seven hundredths of a percent of a year's worth of global air cargo. In other words, all the iPhones that come into the United States in a year, combined, are still basically a rounding error of global air cargo.

What's with the PVG-ANC route? Is Apple doing U.S. fulfillment out of Alaska? Those phones still need to reach the mainland, so I think you need to double the ton-kilometer number to get them down here.

Shanghai is the point of departure from China, and Anchorage is the point of entry in the US (both are significant cargo hubs). So that's the route you can rely on the shipments taking; from Anchorage they could go anywhere in the US, and so it's not really possible to estimate how much further they travel on average.

Even if we assume worst case for 100% of iPhones -- that is, that every iPhone bound for the United States ends up in the southeastern corner of the country in order to maximize the distance flown (since that's the furthest point from ANC) -- the incoming leg from Shanghai is still longer than the onward journey from Anchorage. Which means that the estimate for PVG-ANC represents the majority of the RTK of iPhones bound for the US, which means the hard ceiling on the total RTK even if we assume every single iPhone is delivered to someone in Miami to max out the distance, is around 300m RTK, which is still a rounding error of the global air-cargo total of 208 billion.

No matter how much you try to push it to be a bigger number, "the scale of Apple" is just not significant in terms of percentage of global air cargo.

Are you sure there are 100 freighters operating on the PVC-ANC route daily? 60% of air cargo is flown in the belly of passenger planes. There are not a huge number of dedicated freighters out there.

A 77-tonne load would represent about 6% of the capacity of the scheduled freight-only flights that ran PVG-ANC today (Wednesday). Depending on final aircraft arrangements, it would represent at most 4.8% of tomorrow's (Thursday's) scheduled freight-only flight capacity.

Very cool. Where do you get your data?

Apple is flying 1.4 full freighters per day just for the iPhone (well, maybe it's like 1/3 of 4 freighters). So they are a significant percentage of transpacific cargo.

Plenty of flight-tracking sites let you look up scheduled flights on any route and see how many flights and what type of aircraft.

New blog post coming at the end of the week but my initial estimate pegs the iPhone alone at 5% of air freight crossing the pacific in the first quarter of last year.

But still they are not likely to go into the airline operating business given all the regulations, airport slots, and operating expertise required.

I would expect that they are using some of that insane cash hoard to provide awesome financing terms to their preferred airlines. That could come just through guaranteed contracts for space, but if I was at Apple I'd get creative offering airlines some loans to get better planes and retrofit them to my needs.

I would guess they calculate shipping cost on 'per phone'.

14 hour flight from China would cost give or take $300,000 on a large cargo jet. If a plan can pack in 300,000 phones, the cost of shipping would be $1 / phone for a product that costs $800. If they sell 250 million phones per year, it's $250 million in air freight cost for iphones. Even if we are generous and pretend margin for air cargo business is 20%, Apple would say maybe $50mil/year by operating their own fleet.

EBIT margins of a well run freight forwarder are closer to 5%, so this looks aggressive. And it's so far outside their core competency, to save even $50M that'd be a big risk to take.

Freight operators are extremely good at selling space and solving the problems described in the article and operate a dispersed global network in a difficult space on extremely thin margins. It barely makes sense for Apple to charter, except perhaps to max out loads to the US and Europe in the run up to new product launch. Apple's iPhone sales vary between 12 and 40m units per quarter, so it makes far more sense for third parties to balance that variable demand for them.

Now I'm off to pick up a MBA with a new SSD that apparently took four days to get to London!

>Airlines consolidate heavy freight with light freight to get the ideal mix that lets them charge the companies with heavy goods by weight and the companies with light freight by volume. The result is called “free margin,” and is one of the best kept secrets in the air cargo industry.

This assumes that the cost of fuel doesn't increase with the mass on the aircraft. It does. But the cost of fuel doesn't increase with the volume of the payload because the fuselage is statically sized. In other words, they aren't losing money if they are shipping around a bunch of lead bricks that only take up 15% of the available cargo space.

"In other words, they aren't losing money if they are shipping around a bunch of lead bricks that only take up 15% of the available cargo space."

They may not be literally losing money, but it's a pretty clear-cut opportunity cost if there was anything they could do with that volume that produced profit meaningfully more than $0. I wouldn't care to stand in front of the airline's financial officer and try to make your argument.

The opportunity cost you are paying would be in not having enough flight capable planes, not in having the cargo hold half empty by volume.

The fuel use increases with weight, but not as much as you might think, for two reasons:

1. Even at max weight, the cargo is only half the mass of a 747; put another way, you only decrease the mass by 50% when going from full cargo to empty.

2. Reducung weight reduces induced drag (the drag that you get in exchange for lift) but leaves parasitic drag unchanged. I don't know for sure, but I'd be willing to bet the parasitic drag dominates.

Pilot here. Induced drag dominates at low airspeeds (i.e. takeoff and landing). At higher speeds parasite drag does dominate, so for most of the flight its what matters the most.

I don't know the specifics of the 747 mass characteristics, but you are correct in saying that a good chunk of the mass is the plane itself. Not the cargo. But of the plane's mass, a huge amount of that is just fuel. Whenever they take on fuel, they measure it in thousands of pounds (in the US at least). So by reducing the cargo weight, you reduce the fuel weight as well because you can takeoff with less fuel.

In other words, removing a pound of cargo can reduce the total weight of the plane by several pounds. For extremely short trips this would be irrelevant, because fuel costs are dominated by takeoff, landing, and waiting for clearance for those two things.

> In other words, removing a pound of cargo can reduce the total weight of the plane by several pounds

Thats only true if a significant amount of fuel is used to overcome induced drag when in cruise. If most of the fuel is used to overcome parasitic drag, then reducing your weight doesn't significantly affect how much fuel you need to carry.

Am I the only one who finds this a bit grating?

> If it was any company besides Apple I’d probably say 10% inefficiency, but they are pretty much the best at everything they do.

I think the impressive bit is not just having the scale to be able to afford to be worrying about that level of detail. There are a lot of companies that would benefit from the 5% of reduced waste in packing their devices into crates, but it does seem like Apple is uniquely adept at achieving fringe optimizations which turn around incredible ROI.

Not really, Apple makes the best Apple products in the universe. How dare you question the legacy of the late Steve Jobs?

I do too.

i am very skeptical that they use air freight. having worked in large scale logistics, its like the very last choice you use when you are going to be guaranteed a lot of money because it is expensive. there are not that many planes, they are tiny and relatively expensive to run. i can imagine that apple plan ahead, and make sure that the long boat journeys are factored into their plans.

is there any evidence of what they actually do? i'd put my money on boats. iphones don't magically devalue in transit afaik and the lead time is short enough to be practical... it just makes no sense, from what i understand, to use planes unless its an emergency.

I know a pilot that has flown iPhone shipments many times for a major carrier. He's not a fan due to the potential (heh) in-flight battery issues.

The cost of bulk shipping an iPhone by air from China to the U.S. should be around $0.25-$1.00 and by sea would be about 1/5th that price.

Assume they can earn 5% on their cash hoard with smart, conservative investments.

If they have a $600 phone (they sell them all for that price) at sea for 1 month they would have earned about $2.50.

They sell the phones as fast as they can make them, and if they don't have a phone in stock and someone desperately needs a phone, they may buy an android, so that's revenue they'll never see again.

When we're talking about $0.40 savings to have to delay $600+ in guaranteed revenue by a month. That's all I need to see to know do air every time.

Of course we also know that Apple manages their entire supply chain with just 5 days of inventory in stock at any give time. Much harder to do that with ocean freight.

this only works if there is a delay. part of organising this sort of thing is doing it upfront to avoid the delays, the deliveries can all still arrive at the same time. the trade off is that you can not react to supply/demand change so quickly... even then surely you would use planes to fill that gap only, not for everything...

one thing that probably skews me is time... air freight has probably gotten cheaper continuously over the last 15 years..

given the problems when there is a lack of supply of phones, i'm still dubious that air freight is the standard option... such shortages last longer than a week.

How is that a secret...

You would have to be a complete moron to not realize that solution when faced with the issue.

Ask any friend what the secret to making money in air freight is and see if that's true. Even people who ship cargo by air every day have no idea the game being played on the other side of the transaction.

Well that answer to that question could be anything.

I would say the secret is to have people pay you for shipping cargo....

If i ask some fright company to ship some cargo from point A to point B i assume the freight company wont waste money doing it.

They don't know exactly how stuff is done cause it does not mater to them.

Do they know they are being charged by weight if their product is dense, and by volume if their product is not-dense?

Reply on news.ycombinator.com