Categories

The job the iPhone is hired to do, part II

I repeat what I’ve mentioned before: The iPhone is primarily hired as a premium network service salesman. It receives a “commission” for selling a premium service in the form of a premium price. Because it’s so good at it, the premium is quite high.

The job the iPhone is hired to do 

The original post on the hiring of the iPhone by operators was anchored in data about the revenue per unit (or price) that the product was able to obtain. The remarkable resilience in the exceptionally high average price showed that the iPhone was still getting a premium for moving users to higher levels of spending on network services.

The evidence was circumstantial however: By knowing the price and knowing it was far higher than competing products and knowing that much of it was paid by the operator and not the consumer (at least not up-front) implied that it was the iPhone, and only the iPhone, that was hired as a network service sales tool.

Now we have more evidence thanks to Ben Thompson (@monkbent). I illustrate the data here as an x-y scatterplot.

Screen Shot 2013-04-23 at 4-23-11.21.37 AM

The data consists of ARPU (Average Revenue per User) ITU data from 2010 for 60 countries combined with browser share data on a country-level from the current period (sourced from StatCounter).

The first observation is that there is a modest correlation between iOS usage and ARPU. It may not be particularly interesting by itself but it is interesting when compared with the lack of correlation between Android usage and ARPU.

The presumption behind smartphone usage is that it leads to more browsing which leads to more network usage which in turn, leads to more network revenues and, finally, more network investment. If there is no additional browsing then there is a far smaller economic incentive to network operators to invest in infrastructure. It is this link between usage and revenues which I hypothesize drives operators to carry, subsidize and promote the iPhone. And the resilience of this link indicates that it still works.

The ASP data seems to support this hypothesis and so does the ARPU data. The more iOS is in relative use, the higher the network revenue per user. Correlation is not causation, but when combined with data about Android, we can also say that the presence of the primary iOS alternative does not lead to more revenue. In other words, if a lot of iOS is present then revenues are higher and if there is little iOS present then revenues are not higher.

I’m not suggesting that Android users use less bandwidth—the share data is after all relative—but rather that Android use itself does not seem coupled to paying more (or less) for service.

If there is something to be concluded it’s that not all usage is created equal. The data regarding how apps and ads are consumed is consistent with a qualitative difference between iOS and Android. This data is more of the same.

It’s something of a cliché to say that audiences vary in terms of quality, but that does not make it less true. Indeed, network operators depend on it being true.  And so does Apple.

 

  • http://www.isophist.com/ Emilio Orione

    So long to the low price iPhone?
    It seem that in countries with low iPhone presence the lower ARPU should push operators towards deploying more iPhones, while in the opposite countries operators should be happy for the price they pay to Apple since the product perform.
    Perhaps the low iPhone country carriers are asking for low price to be able to increase the share more easily and perhaps Apple is listening to them, we will see.

    • handleym

      ARPU is a less than ideal measure for what you are talking about because it doesn’t incorporate ability to pay.

      Even in poor countries with low ARPU, there is no a priori reason to believe that users would not spend more on telco services IF using those services were more convenient or felt safer (cf the epidemic of cramming the US experienced in the past).

      The real question is something like:
      Would a developing country person, who has access to an Android phone today, and who uses it pretty much only for calls, utilize data (as web browsing? Skype? Messages rather than SMS? FaceTime?) if they had access to a low price iPhone?
      This is the hypothesis, based on experience so far. But it is admittedly a hypothesis, and based on rich country data more than developing world data.

      • Walt French

        I agree that rather than ARPU, some economic measure — perhaps, the fraction of the population with annual income above 50X the iPhone’s cost — is a better measure. Still, ARPU is a measure of the economic value of communications in a market, and so proxies for the value of services that the iPhone delivers.

  • oases

    Perhaps you shouldn’t use ‘usage’ as a synonym for ‘market share’ in case they turn out to not be the same…or am I just being pedantic?

    Also, if you could plot profit per user instead of revenue per user would there possibly be even greater variance between iOS/Android? In other words, do networks incur more costs for Android due to having to deal with upgrades etc.?

    • http://www.asymco.com Horace Dediu

      What the market share measures is share of browsing which is usage in my mind.

      • Chaka10

        Would it be useful to weight the country data points by size of the markets?

      • http://www.facebook.com/dougal.watt.7 Dougal Watt

        Horace, I don’t think your graph actually shows anything meaningful. You are plotting 2010 data against 2013 when they both vary in unknown ways over time. Trying to fit a meaningful regression coefficient and infer relationships when you have highly collinear data like this is highly problematic.

      • http://www.asymco.com Horace Dediu

        Variance in ARPU is not unknown. I have seen the data at various levels of aggregation but not as specific as the data set I require. It shows a mostly uniform reduction and I don’t believe there would be an effect on the correlations shown.

      • http://www.facebook.com/dougal.watt.7 Dougal Watt

        Do you mean that ARPU has been constant since 2010? Surely that’s not true. ARPU is likely to have changed significantly since 2010 due to several factors, but the most obvious would be platform market share. If you think back to 2010, Android and iOS would have different market browsing share to today so ARPU would be different to todays figures too.

      • http://www.asymco.com Horace Dediu

        No I don’ t mean that ARPU has been constant, but that the variance in ARPU is known at a higher level of aggregation, i.e. regional or a select set of countries. We also have complete data from 2008, 2009 and 2010 and can extrapolate.

      • http://www.facebook.com/dougal.watt.7 Dougal Watt

        Ok, interesting. Is it possible you could show this variance? I only ask because I highly value your work and want to understand this particular point.

      • http://www.asymco.com Horace Dediu
  • obarthelemy

    Does correlation = causation ?

    I think the causation is the other way around. The iPhone sells more in countries were service is more expensive *because* more expensive contracts make it easier to subsidize the iPhone to lower its perceived cost, and make that extra cost more palatable because since the service is more expensive, the iPhone’s share of the total cost is less prominent. A $199 iPhone with a $100 contract seems more interesting than a $800 iPhone with a $30 contract; and a better deal compared to, say a $199+$100/mo or $550+$30/mo GS3, or an even cheaper handset (OK ones start at around $200).

    As contract costs are going down, that will be a medium-term issue for Apple.

    Also, you’re assuming browsing habits/shares are the same in all countries, and that more browsing = more revenue. Both assumptions seem doubtful, especially with many countries being mainly on “unlimited” contracts with generous thresholds.

    • weird

      If I’m understanding your argument correctly, I don’t see how it explains why carriers are willing to pay the increased subsidy. Just thinking about US carriers, they could have tacked the extra $200 or so onto the upfront cost, or added a monthly handset repayment fee for the life of the contract. If they’re not getting the benefit of increased ARPU then I don’t understand why they’re willing to pay more than they seemingly have to. Is it solely that the iPhone is so popular and desired that a significant number people will switch carriers to use one and won’t accept an Android alternative? Would they lose too many customers shopping around on iPhone prices if they didn’t subsidise as much as other carriers? Does all this mean they’re forced into carrying the iPhone solely to stem subscriber losses? I need some help with the full picture here.

      • obarthelemy

        I think you’ve got it: all of the above. Plus, maybe their cost is not full retail $200 extra.

        And from the carriers’ point of view, this is a virtuous cycle: end-users are more sensitive to handset price than to service charges, so giving the impression of a good deal on the handset while hammering on service cost is an easy and lucrative sale, with lock-in on top.

      • weird

        I was using the iPhone ASP against other “premium” smartphones for that approximate $200 extra. ASP already shows that carriers pay less than full retail.

        Why does it allow them to “hammer on service cost” more than equivalent android phones? I want to understand why they’re choosing to subsidise iPhones more heavily. This seems to be leading me dangerously close to returning to the point of the article, something like the increased usage of an iPhone – seen in web browsing stats, app usage stats, etc – leads to greater value for the customer, in turn leading to a willingness to pay more for service.

      • obarthelemy

        Because they don’t care: the extra cost for an iPhone is a drop in the bucket compared to the 2-yr rent they are then extracting.

        Increased usage doesn’t matter, or even is a drawback: I pay a fixed monthly rent for unlimited everything, so the less I use my phone, the more money the carrier actually makes from my custom.

      • weird

        Not really a “drop in the bucket”. Even next to a $2000, 2-year contract that’s a significant hit to take if they don’t have to.

        “Increased usage doesn’t matter, or even is a drawback”

        For the carrier, sure, but my comment was obviously about the customer which it matters a great deal to.

      • http://twitter.com/WalterMilliken Walter Milliken

        Believe me, VZ would squeeze every penny they could out of the subsidy, if they could. There are a *lot* of drops in that bucket. They’re not doing it because they don’t care, they’re doing it because they have no choice.

        And increased usage is a big carrier profit driver in the US, at least, given the tiered data plans. There’s a reason they dropped the unlimited data plans.

      • emole

        You must be new here. Obar is a resident troll on this board, just ignore him.

      • Chaka10

        Exactly. His thesis, stated over and again, is that the US carriers are non-competitive (other like posters have used the word “cartel”) and disguise high phone costs thru the “high” monthly fees in the 2 year contract. What’s not been explained, though asked repeatedly, is (a) why the carriers would engage in this alleged non-competitive behavior IN ORDER TO PAY APPLE the excess profits (subsidies), which are a main expense affecting the financial performance of the carriers and (b) what are the market forces that allow this anti-competitive behavior?

      • obarthelemy

        It’s been explained, but I can rehash.

        - customers are more sensitive to upfront cost than to total cost over a period.
        - phones are more of a motivator than contracts. People want the newest gizmo, not the newest contract.

        so

        (a) Carriers are OK with giving an extra $200 to Apple, and keeping the other extra $800 or so for themselves. High service costs mean the focus is on keeping the customer, even at the cost of an initial “bribe”.

        (b) barriers to entry, regulatory laissez-faire, consumer apathy. We’ll see how T-Mobile breaking of ranks pans out.
        What makes the US market not free and the prices artificial (I never said “not competitive”) is a series of artificial and rather unique barriers to switching:
        - each carrier use their own radio technology
        - there’s no legal framework for early termination,
        - nor for unlocking devices,
        - nor for changing carrier while keeping the same phone number
        So customers end up with one chance every 2 years to switch, providing they are willing to dump their current hardware and change phone numbers.

        I think this makes it hard for US readers to understand how different the market looks when you buy your phone separately from a contract, and you can change providers at will, at a moment’s notice, with no cost no other adverse consequence. Both the markets for contracts, and for hardware, suddenly become more lively :-p

      • Chaka10

        From Wikipedia:

        “In the United States, the Federal Communications Commission (FCC) first mandated LNP among wireline carriers in 1997, as well. LNP was first implemented in the U.S. upon the establishment of the original Number Portability Administration Center (NPAC) in Chicago, Illinois in 1998. This service covered select rate centers in the Ameritech region. Thereafter, as switches and telephone networks were upgraded with Local Routing Number (LRN) capability, LNP was deployed sequentially to the remaining Regional Bell Operating Company (RBOC) areas. The U.S. FCC since has mandated Wireless Local Number Portability starting November 24, 2003 (in metropolitan areas), and allowed operators to charge an additional monthly Long-Term Telephone Number Portability End-Use Charge as compensation. On November 10, 2003, the FCC additionally ruled that number portability applies to landline numbers moving to mobile telephones and, on October 31, 2007, the FCC made clear that the obligation to provide LNP extends to VoIP providers.”

        From the VZ website:

        Switch to America’s most reliable wireless network and keep your number. Bring your 
existing number from your current wireless or landline service provider to Verizon Wireless.

        – No new number to remember.

        –No need to contact dozens of family members, friends and associates.

        –No unnecessary expense printing new stationery or business cards.

        From the T website:

        Q. Can I switch to AT&T but keep the phone number I already have from another carrier? 
A. Yes, it is possible to keep the number you already have from another wireless or wireline carrier. First, check to see if your existing number is eligible for transfer to AT&T. If it is, then all you have to do is follow the instructions to authorize the transfer. We’ll do the rest.

      • obarthelemy

        Thanks, I wasn’t aware of that change (that’s how old I am). Good for the US. Anything on the other items ?

      • Chaka10

        I’m reluctant to get drawn into another tiresome debate with you, but OK:

        Your argument still boil down to customer preference. Not sure how we fault carriers for reacting to customer preference or for find anything “artificial” in that. In fact, if anything, paternalistic regulatory bans of long-tern contracts are in my book “artificial”.

        Don’t see what’s unique about the US in different carriers having different radio technology.

        Legal framework for early termination/unlocking — it’s a contract.

      • obarthelemy

        Yes it does. Horace keeps repeating it’s carrier preference. The artificial part in that is that carriers are eating the price difference. When luxury costs the same as workaday, of course everybody wants luxury. As to whether the state needs to intervene in markets with only 3 players, I’ll leave that to Adam Smith ^^

        I’m not aware of any other major country where each carrier have their own tech. Not in Europe. Maybe Canada, I think. That’s on top of binding contracts, too.

        Exactly, it’s a contract. Again, in most of Europe, there are *laws* framing those contracts. For France which I know well, unlocking is not only legal, but mandatorily (is that a word ?) free 3 months into a contract (I think, I have bought a locked phone, nor 2-yr contract, in ages); early termination fees are lilmited too (I think 25% of the remainder, after 1 year out of 2).

      • Chaka10

        When I’m tempted to comment on French capitalism I know it’s time to stop (no offense and said with a smile).

      • obarthelemy

        Indeed. But it’s not all black and white. Where the robber barons the epitome of capitalism ? or MS ?

      • Chaka10

        Not. VZ and T make it easy. See link for the Local Number Portability Administration Center for more info, or call them :)

      • JohnDoey

        Number portability does not mean there is competition. There is no phone portability, and no contract portability, and no functional competition. In most of the US, Verizon is your only choice because they are the only one pushing signal into your neighborhood.

      • Chaka10

        Hmmm. I admit I’m a bit shocked to see that claim. It’s not the case in my experience, and it’s the first time I’ve heard of that. If there is no competition among the carriers, then I don’t understand why any carrier needs to pay Apple handsomely (and to advertise in national media — on their network coverage and otherwise) to compete with other carriers for customers. When you say “most of the US”, do you mean by geography or by population?

      • Mark Jones

        (a) Cheapest MobileShare/Share Everything plan is $85/month (on AT&T, and $90 on Verizon Wireless, so $2040 for two-year contract plus $200 for iPhone. Apple gets $650, so carrier gets $1390. So a bit more than $800.
        (b) Each of the big four uses own radio tech/frequencies/speeds. But some prepaid (i.e., unsubsidized) use the same. So one can bring your phone to a prepaid (by breaking contract for a fee, or at end of 2-year contract)
        - Not sure what you mean about no legal framework. Pretty much all go with $325-350 for a smartphone, and decreased by $10 per month.
        - Unlocking phones for use on a prepaid is possible, One can unlock and move from AT&T to T-Mobile or vice versa but can’t use 3G or LTE due to different frequencies used. All Verizon phones are unlocked but case-by-case as to whether can use on Sprint. Sprint phones can be unlocked after a period of time, but case-by-case as to whether can use on Verizon. So a mess.

        - Phone number portability is not an issue.

        US readers who explore prepaid plans can know somewhat what it’s like to have the freedom to switch.

  • obarthelemy

    Plus, having worked for several OEMs and software houses, I think you’re overstating the power and motivation of resellers. Resellers don’t pick favorites, they slavishly try to give their customers what they want. Contracts cost the same whatever the phone nowadays, so any carrier motivation/bias is too weak to counterbalance end-users desires. What little motivation there is, I’m not even sure it goes Apple’s way: I’m fairly sure Apple devices cost more even to carriers, thus lowering margin; and I’m not sure the higher satisfaction of Apple users (if any) carries over to the carriers.

  • Rob Scott

    iPhones use more data compared to Android phones. They also use move voice minutes. Android phones of similar capability to iPhones use similar amounts of data but lower voice minutes.

    Subsidies paid by networks for iPhones are not that higher than subsidies paid for similar high end devices (esp. for high end plans).

    The higher data and voice usage results in higher ARPU, thus higher device margin (total revenue less device cost) relative to other phones.

    For the most part users pay for their phones overtime. The subsidy is in someway a smoke screen. It is typical that networks recover the device cost for iPhones in 5-6 months.

    If there is a cost to selling the iPhones it is – what they would have made if they invested the money somewhere for 6 months. Not much.

    • interesting

      “Subsidies paid by networks for iPhones are not that higher than subsidies paid for similar high end devices (esp. for high end plans).”

      This doesn’t ring true to me. Do you have something that verifies this?

    • obarthelemy

      Except for unlimited-everything contracts, which are getting so cheap ($20/mo in my case) that getting any other contract doesn’t make sense.

      Strangely, carriers then start to neither include nor subsidize phones: they not only have no more reason to push us to higher-end contracts, but actually have a reason to *not* push us to devices that consume more data.

      • http://twitter.com/WalterMilliken Walter Milliken

        Unlimited everything plans in the US lasted only as long as customers didn’t actually use the data much. It’s possible there’s a chicken-and-egg thing going on here….

        I don’t know if the same situation will play out similarly in Europe or not; the carrier situation is very different there. My suspicion is that they’ll simply throttle peak bandwidth if they have “unlimited” plans that chew up too many resources, unless the EU regulators stop them. This is what the US carriers do on “unlimited” plans that are still alive.

      • obarthelemy

        We do already have “fair use caps” (how do they mangle concepts ! ^^), I think it’s 3GB in my case, after which carriers “have the option” to drop us back to 2G speeds, or we can buy more high-speed data.

  • http://twitter.com/jeremyjustice Jeremy Justice

    Is iOS here including iPhone and iPad? The latter is a more popular tablet and web browsing device. Would that skew the iOS data?

    • http://twitter.com/ggruber66 ggruber66

      Jeremy, my guess is that it may include both iPhone and iPad, but the iPad rates would have a negligible impact on the numbers as they sell in far fewer numbers compared to iPhones and the majority of iPads are sold WiFi only and therefore don’t have carrier contracts associated with them.

      • http://gravitationalpull.net/wp/ ampressman

        Jeremy, your assumptions about the data are incorrect. The browser share number is NOT limited to browsing over cellular networks – it includes any and all browsing and the tablet issue is significant in skewing the usage number. I have only seen figures for the US but the huge iOS advantage in browser share melts away when phone-only data is used.

        The original poster had a more sound point: “the iPhone is highly competitive versus Android in subsidized markets (the left side), but is simply too expensive in non-subsidized markets (the right side)”

  • Rick K.

    A note about ARPU:

    ARPU is a really bad indicator… It does not take in to account service delivery costs and is based on SIM card sales versus active mobile accounts.

    ARPU is usually a good measure of a provider’s service area’s population density. The lower the density, generally, the higher the ARPU. Compare Canadian and UK ARPU, then compare population density. You see?

    I would instead suggest AMPU, Average Margin Per Unit.

    • ARPU

      It depends what you want to look at, I guess. ARPU tells you something about how much users value the combination of device and service, and are therefore willing to pay for it. There is some degree of inelasticity because having a cell phone is almost a necessity, though.

  • http://twitter.com/monkbent Ben Thompson

    Horace, first of all thanks for the link, and for the original data on ARPU.

    I really like this theory, and agree it is why carriers tolerate iPhone subsidies.

    However, I don’t think it’s the primary factor behind carrier adoption of the iPhone. Rather, I think consumers demand the iPhone, and vote with their feet.

    That’s why the iPhone generally launches on the 2nd or 3rd place carrier in a given geography.

    I laid this argument out more extensively in a follow-up article: http://stratechery.com/2013/why-do-carriers-subsidize-the-iphone/

  • http://twitter.com/WalterMilliken Walter Milliken

    I think there’s a large potential flaw in this analysis, if my understanding of StatCounter is correct. They appear to count web page hits, but those aren’t necessarily correlating with overall traffic volume. I’m particularly concerned about the qualitative differences in traffic between web (text/pictures), audio streaming, and video streaming traffic. In general, I’d expect each traffic type to count as one “hit”, but the subsequent bandwidth due to the hit is *very* different. (And I’m not sure that Skype-style VoIP services will even be seen at all by StatCounter.)

    It’s overall traffic bits that the US carriers charge for, and profit from, not web hits. It’s possible all three traffic types (web/text, audio, video) have similar profiles across different countries, but I wouldn’t bet on it.

    It would be very interesting to get the stats on traffic volumes and mixes by platform and country. The carriers almost certainly have this information, but I doubt it gets published at all. Third-party web stats collectors don’t really have the necessary visibility unless the service hosting companies allow them to.

    • obarthelemy

      Also, the “hits” don’t differentiate between mobile (3G, 4G) and WAN (Wifi) sources. For all we know, there might be significant differences in patterns by platform, country, age group…
      For example, the older people around me don’t, as a rule, have a data contract. They have wifi, including roaming wifi (they automagically connect to *any* in-range box from their ADSL provider), which sounds enough for mainly home-bound elderlies. I myself barely use 300 MB mobile data/month, wifi is everywhere it seems. Teens on the other hand, can’t cut the virtual cord even for the half hour it takes them to reach school.

  • JohnDoey

    I still don’t get why people expect Android numbers of any kind to match iPhone numbers. The weird thing would be if they matched because they are 2 different animals. If we compare 2004 iPod metrics with 2004 phone metrics we should also expect them not to match. iPhone is also an iPod (“iPod phone” shortened to “iPhone”) and although many Android phones attempt to look the same, they are not the same, they are just highly-evolved 2004 phones, with carrier control, no software updates, and baby Java apps.

    I think took my people say, well Apple is from Silicon Valley and Google is from Silicon Valley, ergo Android is the same as iOS. It is not. Android, Inc is a traditional phone company that Google bought. iPhone comes out of pure Silicon Valley tradition. For example it is running the kernel from the 1988 NeXT workstation upon which the World Wide Web was created, and uses the same developer tools and BSD Unix and relies on the same Mac heritage.

    So we should expect iOS and Android devices to be different in every way. Where they are the same — that is what is surprising.

  • mieswall

    Fantastic info!
    What about weighting the dotted pairs (iOS-android) per number of users in each country? The correlation would be even more clear, is my guess.

  • http://www.facebook.com/dougal.watt.7 Dougal Watt

    I don’t think it’s correct to plot one axis with 2010 data and the other with 2013 data. Because you are trying to correlate two sets of variables that are likely to be time-dependent, you should use the same time series for both data sets.

    Using your approach, you run the risk that you conflate correlation of variables that have independently varied across the interval. Take this example for the sake of argument. We know that Android has penetrated more into lower cost markets, using the proxy that the upfront cost of Android devices now around $100-150 for a decent model hence they are likely to be customers with less disposable income. Thus is may be reasonable to assume that ARPU for those users has moved lower so the 2013 ARPU is lower than 2010 ARPU, while market share has increased. Therefore if you try to correlate 2010 ARPU to 2013 market share you won’t be measuring anything real.

    Having said this, this would be very interesting to plot with current data.

  • Pingback: stratēchery | China Mobile Customers Making Fewer Phone Calls, Telecom Companies At Risk

  • Pingback: stratēchery | China Mobile May Need the iPhone

  • Pingback: Can Carriers Handle a Low-End iPhone? | Christine's Blog

  • Pingback: Carriers Still Think We’re Idiots | Monday Note

  • Nick

    It would be best (if possible) to compare high end Android devices to the iPhones to see if the delta in R^2 is consistent. Probably unfair to compare a low end Android to iPhone, especially since Android share is “artificially” high due to lack of other options

    • Nick

      …at the lower end of the price band

  • http://twitter.com/PeterTanham Peter Tanham

    Great analysis (as usual). It’s definitely true that iPhone users have the highest ARPUs around (and not just because rich people have high ARPUs and buy iPhones – the ease of use for email, browsing & apps makes for up to 5x higher data usage when a user switches from some types of Android to an iPhone).

    However, I don’t think this accounts for all (or even most) of the job that the iPhone is being hired to do. I think it’s mostly about acquisition and retention of users, not ARPU uplift.

    I worked for a network that didn’t have the iPhone until the 3GS and believe me, if you don’t pay Apple’s asking price it’s not that you miss out on the “premium” upsell – it’s that all your high end customers leave you for the network with the iPhone!

  • Pingback: Mobile phone networks must think consumers are idiots | Padroni