Research

My brief appearance on Bloomberg Surveillance last week:

Apple Success All About Customer Loyalty: Dediu: Video - Bloomberg.

My appearance on Bloomberg West using an iPad to illustrate what it takes to make an iPad.

The Numbers Behind the Apple vs. Samsung Rivalry: Video - Bloomberg.

IBM Digital Analytics Benchmark (report is here) has published, for the fourth year in a row, US online shopping traffic data with a split between mobile and fixed online traffic. It reveals a pattern of consumer behavior which is quite startling: people seem to prefer to shop using mobile devices.

The data is shown below:

In terms of sources of traffic, mobile devices (including tablets) have increased from about 5% of all online traffic to 40% in four years. Last year the volume was 24% implying that 16 points of share were gained in one year. If that is extrapolated to next year then more than half of holiday shopping traffic in 2014 will be on mobile devices.

Put another way, next year less than half of users will hire a PC for the job of online shopping.

If this conversion from PC-based shopping to device-based shopping follows a logistic substitution curve then the conversion to devices will probably be 90% complete in another 5 years or by 2018 or 2019. I see no reason why this particular job to be done (shopping) should not be considered a proxy for other jobs as well. Therefore it's not a stretch to imagine that the transition to post-PC consumption will also be practically completed by 2020.

Within the data there is also a significant amount of detail beyond the mobile/non-mobile split in behavior. There is a distinction drawn between tablet and phone traffic and a further distinction between Android and iOS devices. ((The split between Android and iOS for tablets and phones was not published for 2013 but I extrapolated from previous years and assumed that 80% of tablet traffic was sourced from iPads vs 88% in 2012 and 99% in 2011. Note also that non-iPad traffic is assumed to be Android but that includes the Kindle and Nook products which are using partial and unsanctioned versions of Google's Android.))

This more detailed view of the traffic is shown in the following diagram:

A few observations:

• The pattern of slowing tablet growth is likely due to the slowing growth in iPad sales. This itself is due to the lull in product launches as the shift to fall only iPad releases took place in 2012.
• The consequence is that iPad traffic contribution only increased by 2 percentage points, from 10% in 2012 to 12% in 2013. Compare this with the doubling the previous year (from 5% to 10%).
• Among phone users, both Android and iOS grew their contributions to traffic by roughly 100%. Android phone traffic increased from 4.2% of all traffic to 8.5% and iPhone traffic increased from 8.7% to 16.7%.

As we also know the approximate number of phone users for each platform (via comScore) then we can also determine the per user contribution of traffic of each platform. ((Note that this is phones only.))

The measure is "basis points of traffic per million users" and it is the result of taking the traffic data (in percent of total online traffic) and dividing by the population of users of a particular platform. The result is shown below:

The way to read it would be that each million iPhone users generated about 28 basis points of traffic while each million Android users generated about 11 basis points of traffic.

A way to compare these audiences would be to define the ratio between traffic usage by platform user. That ratio was

• 1.83 (iOS:Android) in 2010
• 2.16 in 2011
• 3.15 in 2012
• 2.51 in 2013

So, as far as phones are concerned, US iPhone users are about 2.5 times more likely to shop online than US Android phone users. The data is reliable: with the US population now over 65% penetrated with smartphones, these populations of users (totaling over 130 million) are vast and representative. As the technology is reaching deep into the late adopter segments the difference in behavior between iOS and Android usage is becoming easier to explain. We have to eliminate quirks of demographics or income or other possible causes.

The platforms themselves are becoming the last remaining suspects as causes for differences in the behavior of their users.

It may seem preposterous to suggest that a platform can affect shopping behavior but let's remember that the PC, the subject of this article, is a platform in itself. The migration of usage away from the PC to alternatives can only be explained by the alternatives being superior at the particular job that needs doing. The argument is naturally extendable to each alternative individually.

iOS unit sales crossed over 700 million units last month. That is a significant milestone but the total number of units in use is likely to be lower. My estimate based on device replacement assumptions is that about 500 million are still in use.

The estimated break-down of units sold and in use by device type is shown below:

When I looked at the penetration of iPhones in the US (as a part of smartphones in total) the surprising pattern was that it followed a logistic curve (while other platforms didn't.) If this is indicative of a particular learning process to be observed in the market, then can we extend the logistic growth pattern to iOS overall?

I tried it with the assumption that at 500 million, iOS is at the half-way point in its growth. This assumption comes from growth being near an inflection point. It could be that iOS might accelerate further in which case peak penetration could be higher than 1 billion, but might take longer to reach.

The 500 million as 50% penetration point assumption fits the model as shown below.

 

 

The fit is pretty good for the period after the iPhone 3G launched. This assumption would therefore imply (nearly) one billion iOS devices in use by the end of 2018.

Note that this does not mean one billion iOS users. The attach rate for iOS devices per user could well reach 2 by then placing total user base at 500 million. Note that this size of audience is very nearly the existing total Apple audience already (counting iTunes/iCloud accounts).

This means that we already have "existence proof" that Apple can reach half-a-billion users. Apple does not need to discover a large new set of non-consumers. The idea of 1 billion iOS devices is simply the projection that each current user could some day use two devices.

Benedict Evans explains well the problem with measuring Android tablets. There are no reliable data collected because many of the devices are invisible through the regular, measurable channels:

• There are no firms which report their shipments
• They are not sold through retail chains which normally are sampled in the US and Europe (NPD and GfK respectively.)
• They don't show up in browsing or ad transaction data
• Google Play statistics are missing most of the activations since they are not sold as bona fide Google-sanctioned Android.

The only measured statistic happens to be component shipments. Items such as screens, CPUs or perhaps memory might be visible to market analysts. It's therefore tempting to add up tires manufactured to determine what's getting sold in auto dealerships.

But it's also hugely problematic.

Before we go into that rat's hole, I want to re-visit my original criticism for market analysis based on device attributes. When the iPad was new it was famously labeled a "media tablet" by Gartner and excluded from being counted (and hence compared) with PCs. This created an implicit signal to the analyst's clients to ignore the entrant as irrelevant.

That actually accelerated their demise (and the revenue prospects for the analyst).

In contrast to that exclusion--of a clear and present danger based on an arbitrary definition of form factor--the analyst who defines categories of products purely on the definition of their components is guilty of the same sin.

Snowing the audience with irrelevant comparisons is just as bad as withholding relevant comparisons. This is explicitly asked of witnesses: to tell the truth, the whole truth and nothing but the truth. The "nothing but the truth" means the exclusion of irrelevance.

The correct approach is to identify which devices are comparable (and thus in the same category) based on the jobs they're hired to do. If a tablet overlaps considerably in the work it does with a PC (e.g. browsing, email, media play, and other applications) then they should be considered comparable. If a tablet does nothing that a PC does or nothing that an iPad does (e.g. no browsing, email, apps or ecosystem activity other than piracy) then they should not be considered comparable.

Which brings me back to the question of what these dark matter "Android tablets" actually do.

To answer I want to recall another "dark matter" technology that seems to have gone largely unnoticed but was exceptionally popular. This was the VCD or Video CD. It was an alternative to the DVD without the licensing and DRM overhead it entailed. It was extremely popular in Asia as a way of not only buying but also sharing video content. The cost of content was a fraction of sanctioned versions, essentially taking the role that mass-copied VHS had in previous years.

The "job to be done" was to enjoy video content cheaply and conveniently for the equivalent of pennies a day. A job very similar to what we think of as cable TV today.

These devices and discs were largely invisible mainly because they did not exist in economies which adopted protected DVDs and because the content consumed did not hit against any databases that tracked consumption. Nevertheless there might have been hundreds of millions of these devices sold and billions of discs viewed.

Now we are in a situation where devices whose primary purpose is to watch video content (cheaply and conveniently) are being counted. By their counting, their implicit competitiveness is asserted. But it means nothing if they are not hired for the same jobs. In fact it misdirects attention.

Sometimes these "low end" devices do grow up to swamp the higher-powered alternatives with good enough functionality. That is the innovator's dilemma for low end disruption. But in some cases they don't because other factors are at work. Because ecosystems have complex multi-sided markets and because regulations and channels and consumer behavior conflict with the technical specification.

For this reason the VCD did not "disrupt" DVD or grow up into a Blu-Ray equivalent that caused Hollywood to crumble. It faded away as an anomaly.

Whether the dark matter Video-only Android device will come to swamp the iPad will depend not on just volume shipments in select geographies. It will depend largely on the ecosystems built around it. The ecosystems of VCD were largely unsustainable because there was no value placed on the content itself. The value chain did not strive to sustain the technology. When something better came along, it got dropped.

In contrast, content-based value chains sustain technologies which keep the revenues coming. And we can measure this revenue.

You don't need to look too hard for that in tablets. Apple states it quite frequently: total payments to developers.

In the post Seeing What's Next, I showed how the rate of change of adoption of technology varies with time and asked what might be experienced by present and future generations.

It turns out that knowing how what innovations become universal and the speed at which these technologies are replaced can give us an idea of what individuals might experience in their lifetimes.

Here's how to think about it:

You can use the model above to visualize a lifespan, given a year of birth. Follow the line for your year of birth and measure an expected life (the life expectancy in green is a guide). You can skip over childhood years and decide when you might have reached an age when you start buying and using certain products as well as when you think you might stop.

Trace a rectangle over the grey areas in the bottom graph spanning the years you will start to consume and when you might expect to stop.

The area within the rectangle will include known products and their expected lifespans (marked in alternating shades of grey with a figure for the longevity of service life.) The total number of segments represents your expected consumption.

I marked the areas for a person born in 1913 and one born in 1968 ((Corresponding to my grandmother and myself though we were not born in the US.)) What the analysis shows is that I would be expected to own:

• Three stoves (I've owned one so far.)
• One regular landline phone service. Mine was discontinued in the late 90s.
• One electricity subscription. (It seems banal to give electricity a product lifespan but will make more sense if the electric grid is re-achitected.)
• Five cars (for my household). I've owned more than five but mostly due to vanity.
• 8 Radios/Stereos. I owned about 5 "sound systems" so far
• 4 Washers. My household has owned 3 so far.
• 4 Refrigerators. I've owned three so far.
• 4 Dryers. Three so far.
• 7 Air conditioning systems. I've only owned about 3 but I lived in cooler climates.
• 5 dishwashers. I've owned 3 so far.
• 4 CRT Color TVs and three HDTV. I've had 4 CRT TVs and one HDTV so far.
• 4 VCRs. I've owned 4.
• 11 PCs. I lost track but I think more than that so far.
• 5 Game consoles. I've owned none. To be fair, consoles are not an universal technology as they have not reached saturation.
• 7 voice-oriented cell phones. I've had a few more but it's close
• Three internet service contracts. (This assumes dial-up as first.)
• 15 Smartphones. 6 so far (excluding company purchases).
• 10 Tablets. 4 so far.

A couple of caveats:

1. My life is not yet over, so there is still time to fulfill my consumption quota.
2. There will be additional things to count within my life span.
3. There are many technologies not counted but few of those I can imagine are universal, or saturated. This means that there is little certainty this model can be used reliably.
4. Some people are early adopters and some are late adopters. Depending on which, it might add or subtract certain product cycles but as adoption rates increase and product lifespans decrease.

I did the same tally for my grandmother and it matches prettily closely her consumption with the caveats that she did not reside in the US and that she lived almost twice the expected lifespan.

The obvious next step is to multiply the model by population forecasts.

 

 

The adoption of smartphones in the US is on track for reaching 90% of the available audience by August 2016. This is a mere eight years after smartphones reached 10% penetration. As far as technologies go, that's pretty fast. To get an idea of how rapid, I plotted a few other technologies and the time they took to grow within the US market.

A few things to note:

1. The time scale has a yearly resolution spanning 1900 to 2074
2. The penetration scale is logarithmic and shows the ratio of Penetration/(1-Penetration) allowing a linear view of logistic growth.
3. The blue bars summarize the growth pulses and give a starting year and duration in years.
4. Note some traumatic events are marked on the time frame. They partly explain innovation gaps or decreases in adoption.
5. The life expectancy of Americans born from 1900 to 1998 is also shown as green bars. This allows the reader to trace which technologies emerged during the lifetimes of people they might have living memories of. I overlaid one specific lifetime beginning in 1968 and color coded with Childhood, Adolescence, Higher Education, Work live and Retirement as a guide. You can note how the PC emerged when that person would have been in college.
6. Some technologies have not yet saturated and may never do so. They are marked with a "Z" on the bar graph. It might be instructive to understand why.

You'll also note that the time frame does not end today but that the technology curves do. The time frame includes the expected lifespan of someone born in 1998.

What will they see?

There is a lack of visibility or certainty among observers that anything worthy of inclusion on this graph will ever emerge.

And yet, when glancing at this graph, it would seem that the addressable market, the rate of introduction of new technologies and the speed of adoption have all increased. The area to the right of 2013 looks inviting to a plot of many lines.

Indeed, for the canvas to remain blank would be the riskiest of bets.

What would you put upon it?

The Transportationist, David Levinson @trnsprttnst joins us to discuss the technical, human, environmental and economic factors driving change to the auto-ecosystem.David helps us smartly survey the landscape via:

1. Human behavior
2. Technology lifecycles
3. Urban transportion evolution
4. Network capitalization
5. Congestion

via Asymcar 7: The Transportationist | Asymcar.

Photo credit: Wikipedia, CVN-78

I've been writing about Apple's capital intensive operations for some time. The difficulty has always been in explaining the scale involved. I've compared the spending to that of Samsung, Microsoft, Google, Intel and Amazon. But these numbers still can't be easily grasped. You can't point to any comparable objects when you try to explain the figure. I struggle to create a less abstract notion than that of a "sea of tooling and servers."

Instead, I've used the analogy of US aircraft carriers. Historically, Nimitz class aircraft carriers have cost a less than $5 billion. The USS Ronald Reagan, christened in 2001, cost $4.5 billion. Therefore I was comfortable saying that Apple spends the equivalent of about one Nimitz class aircraft carrier every six months (and that the Navy takes about six years to put it together.)

Unfortunately, costs for aircraft carriers have gone up. The USS Gerald Ford will take about $13 billion to complete. That places Apple and Samsung capital spending in the following context:

Note however that planning and construction for the latest aircraft carrier takes 12 years, rather than the 12 months spending schedule for the companies shown. Note also that Apple's spending is unspecified and included several categories while Samsung's categories are partially specified.

Due to this inflation in aircraft carrier costs, my denomination of capex is beginning to be strained. Nevertheless, the question of what Apple spends $10 billion a year remains knowing what that could buy the Navy (or Samsung.)

Apple has sold 700 million iOS devices. Google claims one billion Android device activations. Microsoft has about 1.5 billion Windows users and Facebook about 1.19 billion. LinkedIn has 259 million users and Twitter has 232 million. Amazon has 215 active account holders and PayPal 137 million.

Markets place a value on these users implicitly when company shares are priced. For example, Twitter whose users are worth about $110 or FaceBook's $98 and LinkedIn at $93.

This consistency suggests a universally accepted value per social media user but what is the value of an ecosystem user? Apple, Google, Microsoft and even Amazon aspire to enable ecosystems which should be seen are more valuable than mere communities. Ecosystems enable a higher level of economic activity because they are unbounded by the medium itself. Any number of media can be created. Or so the theory goes.

If we could determine a value for an ecosystem user we could test it against the going value of a social media user. Fortunately we have enough data to do so.

The total number of iOS devices sold per quarter allows us to measure the install base of device users. With some assumptions regarding the retirement and attrition rate we can get the following history:

Since the total number of iTunes accounts is updated with some regularity I've added it to the graph. I've also shown on the same graph the total number of iCloud accounts. For calibration, I included survey data showing the number of iPhone users in two regions (US and EU5).

These combined graphs give us an upper bound of active users (iTunes accounts) and a lower bound (iCloud) calibrating our device-based estimate. The regional iPhone data offers some additional test on growth.

The result is "about 500 million" ((482 million by my count but with quite a bit of margin of error)) with a growth of about 11 million per month.

So if we know with some confidence that there are about 500 million active users as part of the iOS ecosystem, how do we know how much value they generate? The answer is in the consolidated iTunes revenues as reported by Apple: $4.26 billion/quarter. With some attention to the reported media retail and purchases, the following picture of the average iOS user's quarterly spending emerges:

The bottom line is that in the last 12 months the average iOS user contributed  about $48/yr to the ecosystem via Apple's own properties. They also bought the device and paid for cellular service in most cases. They may also have engaged in quite a bit of browsing and online purchases which fed other ecosystems. But as far as Apple is concerned, it transacted about $48 per user.

As Twitter revenues last year were $317 million, the Twitter user today generates about $1.36 of revenue per year or one thirty-fifth of what an iOS user generates through iTunes and iCloud alone. And note that the number of iOS users is about twice that of Twitter.

Of course one could argue that the iOS user's participation is not profitable for Apple. It does obtain massive cash flow from iTunes which finances its fashionable data centers and bandwidth but it does not presume to value itself primarily through this revenue.

Or at least the market doesn't think so. We don't hear much about how Apple is worth so many dollars per user. ((I suppose the corollary to this is that Apple users are disloyal or fickle or fleeting, and quite contrary to observed behavior.))

The paradox is that while the device which enables this commerce is where Apple guides the consumer's perception of value, markets discount devices and bid up services. In other words, users continue to pay for the devices which delight them enough to participate in services (especially the monthly billable kind) but markets walk right by these shiny baubles and and shower money on services which have negligible sales and even fewer profits. There must be some serious misdirection going on here. The question is who's being fooled.

The graph below shows the Revenue and Operating Income for a select group of companies. The large numbers represent the share price to earnings (trailing twelve months) ratio (P/E or PE ratio).

Of course the P/E ratio hides a lot of subtlety. It mostly fails to account for the fact that earnings are largely a matter of opinion. A company can defer income (as Apple and Microsoft do), it can invest earnings (as Amazon does) and can otherwise avoid declaring it since it's taxable.

A better measure is the Enterprise Value to Free Cash Flow ratio (EV/FCF). It reflects a "harder" measure of performance in the form of free cash flow. It is not a matter of opinion but a fact. Enterprise Value also excludes cash which is a measure of past performance.

Nominally both P/E and EV/FCF measure the valuation of a company. They give what amounts to the number of years that current performance can be expected to last. For example a P/E or EV/FCF of 10 can be read as "If I purchase a share today, I should expect the company to pay me back in earnings in 10 years." Any earnings after 10 years are essentially "profit" for the investor.

Inverted these ratios are a measure of Yield or how much of the total investment is paid back in a given year. In the example above, the Yield would be 10% or "If I purchase a share today, the company will pay me back 10% a year in earnings."

Relying on P/E is troublesome since, as mentioned above, it's deceptive. We get into difficulty when P/E and EV/FCF differ greatly. If a company has minimal cash and declares all free cash flow as earnings then the ratios are the same. However, consider how different these two ratios are for Apple:

The P/E ratio is 13 while the EV/FCF is 7.5. ((It would be very interesting to have both ratios available for all the companies in the graph above.))

This means the investment horizon is nearly half of what P/E suggests. (The cash flow yield should give pause). Put another way, the market suggests that Apple has 7 more years of current profitability and not 13. Note that the P/E ratio for the S&P 500 is above 19.

As Apple's deferrals are increasing investors should look again at what the company's worth. We had similar confusion when the company was deferring most of its iPhone revenue.