Category Archives: Advertising

Yet Another Reset for Twitter

Here we are, almost eleven years into Twitter’s history and a little over 18 months into Jack Dorsey’s second term at the company, and Twitter is heading for yet another reset. The company says it’s already been through a reset on its consumer-facing product, and that the changes it’s made are delivering results: positive year on year growth in daily active users, though Twitter still refuses to provide the underlying metric. It now says ad products need to go through a similar reset and re-focusing process. As a result of all this, the company isn’t even providing revenue guidance for Q1.

Here’s a quote from Twitter’s earnings call:

we remain focused on providing improved targeting, measurement and creative for direct response advertisers

Specifically, that’s from Twitter’s Q1 2015 earnings call, almost two years ago. But on today’s call, Anthony Noto said almost exactly the same thing again – some of this stuff has been in the works for over two years, and Twitter still doesn’t seem to be making meaningful progress. Rather, it’s now evaluating its direct response ad products to figure out which are delivering an appropriate return on the resources invested in them, with a view to killing some off.

Why is this all taking so long? It seems Twitter has been unable to focus on more than one big project at once, despite its arguably bloated workforce, and it’s hard to avoid the sense that this is mostly about management. It starts with Jack Dorsey, who is trying to run two public companies at once, but it continues with the next layer of management, where there’s been huge turnover in recent years and where product management seems to have been a particular challenge. It feels as though Dorsey at once wants to own product, because he has the authority of a founder in this area, but doesn’t really have the time to do it properly, which means both that things don’t get done and nominal heads of product get fed up.

The other big problem is that Twitter’s big competitors for direct response advertising – notably Facebook and Google – are just way better at this stuff than they are, and Twitter simply hasn’t made anywhere near enough progress here over the last few years. As a result, Twitter is enormously susceptible to competitive threats – its guidance for Q1 is so broad because there was a meaningful difference in competitive intensity between the beginning and end of January alone. Any company that can’t predict its revenue a quarter out with reasonable confidence because of the competitive environment is really struggling.

In the meantime, ad revenue is actually falling year on year, despite the modest MAU growth and apparent growth in DAUs. US ad ARPUs dropped 8% year on year in Q4, and total US revenue was down 5.3% despite flat MAUs. The supposed increased engagement simply isn’t translating into revenue growth. The revenue growth trend for Twitter as a whole is pretty awful:

In percentage terms, the growth rate has been falling since Q2 2014, but even in pure dollar terms, growth has been slowing for a year. The EBITDA guidance for Q1 suggests a pretty big drop in revenue in the quarter, extending the streak here.

What Twitter’s management said today in their shareholder letter and on the earnings call is that it will simply take time for the increased user growth and engagement to flow through, and that Twitter essentially has to convince advertisers that it’s making progress in getting users engaged. But advertisers don’t spend money because of user growth trends – they spend money because it’s effective, and stop spending where it isn’t. Twitter seems to have a fundamental issue convincing advertisers that money spent on the platform will actually pay off, and I don’t see that changing just because it tweaks some ad formats.

Facebook, Ad Load, and Revenue Growth

Note: this blog is published by Jan Dawson, Founder and Chief Analyst at Jackdaw Research. Jackdaw Research provides research, analysis, and consulting on the consumer technology market, and works with some of the largest consumer technology companies in the world. We offer data sets on the US wireless and pay TV markets, analysis of major players in the industry, and custom consulting work ranging from hour-long phone calls to weeks-long projects. For more on Jackdaw Research and its services, please visit our website. If you want to contact me directly, you’ll find various ways to do so here.

Facebook and ad load have been in the news a bit the past few days, since CFO David Wehner said on Facebook’s earnings call that ad load would be a less significant driver of revenue growth going forward. I was listening to the call and watching the share price, and it was resolutely flat after hours until the moment he made those remarks, and then it dropped several percent. So it’s worth unpacking the statement and the actual impact ad load has as a driver of ad growth a bit.

A changing story on ad loads

First, let’s put the comments on ad load in perspective a bit. It’s worth looking at what’s been said about ad loads on earlier earnings calls to see how those comments compare. Here’s some commentary from the Q4 2015 call:

So, ad load is definitely up significantly from where we were a couple of years ago. And as I mentioned, it’s one of the factors driving an increasing inventory. Really one thing to kind of think about here is that improving the quality and the relevance of the ads has enabled us to show more of them and without harming the experience, and our focus really remains on the experience. So, we’ll continue to monitor engagement and sentiment very carefully. I mentioned that we expect the factors that drove the performance in 2015 to continue to drive the performance in 2016. So, I think that’s the color I can give on ad loads.

Here’s commentary from a quarter later, on the Q1 2016 call:

So on ad load, it’s definitely up from where we were couple of years ago. I think it’s really worth emphasizing that what has enabled us to do that is just improving the quality and the relevance of the ads that we have, and that’s enabled us to show more of them without harming the user experience at all. So that’s been really key. Over time, we would expect that ad load growth will be a less significant factor driving overall revenue growth, but we remain confident that we’ve got opportunities to continue to grow supply through the continued growth in people and engagement on Facebook as well as on our other apps such as Instagram.

Some of that is almost a carbon copy of the Q4 commentary, but note the second half of the paragraph, where Wehner goes from saying 2016 would be like 2015 to saying that over time ad load would be a less significant driver. This is something of a turning point. Now, here’s Q2’s commentary:

Additionally, we anticipate ad load on Facebook will continue to grow modestly over the next 12 months, and then will be a less significant factor driving revenue growth after mid-2017. Since ad load has been one of the important factors in our recent strong period of revenue growth, we expect the rate at which we are able to grow revenue will be impacted accordingly

These remarks turn “over time” into the more specific “after mid-2017”. Now here’s the Q3 commentary that caused the stock drop:

I also wanted to provide some brief comments on 2017. First on revenue, as I mentioned last quarter, we continue to expect that ad load will play a less significant factor driving revenue growth after mid-2017. Over the past few years, we have averaged about 50% revenue growth in advertising. Ad load has been one of the three primary factors fueling that growth. With a much smaller contribution from this important factor going forward, we expect to see ad revenue growth rates come down meaningfully….

Again, it feels like there’s an evolution here, even though Wehner starts out by saying he’s repeating what he said last quarter. What’s different now is the replacement of “less significant factor driving revenue” with “much smaller contribution from this important factor”, and “the rate at which we are able to grow revenue will be impacted accordingly” to “ad revenue growth rates come down meaningfully“. Those changes are both a matter of degree, and they feel like they’re intended to suggest a stronger reduction in growth rates going forward.

Drivers of growth

However, as Wehner has consistently reminded analysts on earnings calls, ad load is only one of several drivers of growth for Facebook’s ad revenue. The formula for ad revenue at Facebook is essentially:

Users x time spent x ad load x price per ad

To the extent that there’s growth in any of those four components, that drives growth in ad revenue, and to the extent that there’s growth in several of them, there’s a multiplier effect for that growth. To understand the impact of slowing growth from ad load, it’s worth considering the contribution each of these elements makes to overall ad revenue growth at the moment:

  • User growth – year on year growth in MAUs has been running in the mid teens, with a rate between 14 and 16% in the last year, while year on year growth in DAUs has been slightly higher, at around 16-17% fairly consistently
  • Time spent – Facebook doesn’t regularly disclose actual time spent, but has said recently that this metric is also up by double digits, so at least 10% year on year and perhaps more
  • Ad load – we have no metric or growth rate to look at here at all, except directionally: it rose significantly from 2013 to 2015, and continues to rise, but will largely cease to do so from mid-2017 onwards.
  • Price per ad – Facebook has regularly provided directional data on this over the last few years, but it’s been a highly volatile metric unless recently, with growth spiking as mobile took off, and then settling into the single digits year on year in the last three quarters.

So, to summarize, using our formula above, we have growth rates as follows: 16-17% user growth plus 10%+ growth in time spent plus an unknown growth in ad load, plus 5-6% growth in price per ad.

The ad load effect

Facebook suggests that ad load is reaching saturation point, so just how loaded is Facebook with ads today? I did a quick check of my personal Facebook account on four platforms – desktop web, iOS and Android mobile apps, and mobile web on iOS. I also checked the ad load on my Instagram account. This is what I found:

  • Desktop web: an ad roughly every 7 posts in the News Feed, plus two ads in the right side bar. The first ad was the first post on the page
  • iOS app: an ad roughly every 12 posts, with the first ad being the second post in the News Feed
  • iOS web: An ad roughly every 10 posts, with the first ad being the fourth post in the News Feed
  • Android app: an add roughly every 10-12 posts, with the first ad being the second post in the News Feed
  • Instagram on iOS: the fourth post and roughly every 10th post after that were ads.

That’s pretty saturated. You might argue that Facebook could raise the density of ads on mobile to match desktop density (every 7 rather than every 10-12), but of course on mobile the ad takes up the full width of the screen (and often much of the height too), which means the ceiling is likely lower on mobile. I’m sure Facebook has done a lot of testing of the tipping point at which additional ads deter usage, and I would imagine we’re getting close to that point now. So this is a real issue Facebook is going to be dealing with. I did wonder to what extent this is a US issue – in other words, whether ad loads might be lower elsewhere in the world due to lower demand. But on the Q2 earnings call, Facebook said that there aren’t meaningful differences in ad load by geography, so this is essentially a global issue.

So, then, if this ad load issue is real, what are the implications for Facebook’s ad revenue growth? Well, Facebook’s ad revenue has grown by 57-63% year on year over the past four quarters, and increasing ad load is clearly accounting for some of the growth, but much of it is accounted for by the other factors in our equation. Strip that ad load effect out and growth rates could drop quite a bit, by anywhere from 10-30 percentage points. Facebook could then be left with 30-50% year on year growth without a contribution from ad load. Even at the lower end of that range, that’s still great growth, while at the higher end it’s amazing growth. But either would be lower than it has been recently.

Of course, it’s also arguable that capping ad load would constrain supply of ad space, which could actually drive up prices if demand remains steady or grows (which Facebook is certainly forecasting). Facebook has dismissed suggestions in the past that it would artificially limit ad load to drive up prices, but this is a different question. Supply constraints could offset some of the slowing contribution from ad load itself, though how much is hard to say.

Ad revenue growth from outside the News Feed

Of course, Facebook isn’t limited to simply showing more ads in the Facebook News Feed. While overall impressions actually fell from Q4 2013 to Q3 2015 as usage shifted dramatically from desktop to mobile, where there are fewer ads, total ad impressions have been up by around 50% year on year in the last three quarters. Much of that growth has been driven by Instagram, which of course has ramped from zero to the significant ad load I just described over the course of the last three years. Multiplied by Instagram user growth (which isn’t included in Facebook’s MAU and DAU figures) and that’s a significant contribution to overall ad growth too. As I understand it, the ad load comments apply to Instagram too, but there will still be a significant contribution to overall ad revenue growth from user growth.

And then there are Facebook’s other properties which until today haven’t shown ads at all: Messenger and WhatsApp. As of today, Facebook Messenger is going to start showing some ads, and that will be another potential source of growth going forward. WhatsApp may well do something similar in future, too, although Zuckerberg will have to overcome Jan Koum’s well-known objections first.

Growth beyond ad revenue

And then we have growth from revenue sources other than ads. What’s been striking about Facebook over the last few years – even more than Google – is how dominated its revenues have been by advertising. The proportion has actually risen from a low of 82% of revenue in Q1 2012 all the way back up to 97.2% in Q3 2016. It turns out that the increasing contribution from other sources was essentially down to the FarmVille era, with Zynga and other game companies generating revenues through Facebook’s game platform. What’s even more remarkable here is that these payments are still the bulk of Facebook’s “Payments and other fees” revenues today, as per the 10-Q:

…fees related to Payments are generated almost exclusively from games. Our other fees revenue, which has not been significant in recent periods, consists primarily of revenue from the delivery of virtual reality platform devices and related platform sales, and our ad serving and measurement products. 

As you can see in the second half of that paragraph, Facebook anticipates generates some revenue from Oculus sales going forward, though it hasn’t been material yet, and later in the 10-Q the company suggests this new revenue will only be enough to (maybe) offset the ongoing decline in payments revenue as usage continues to shift from desktop to mobile.

Of course, Facebook now has its Workplace product for businesses too, which doesn’t even merit a mention in this section of the SEC filing. Why not? Well, it would take 33 million active users to generate as much revenue from Workplace in a quarter as Facebook currently generates from Payments and other fees. It would take 12 million active users just to generate 1% of Facebook’s overall revenues today. And that’s because Facebook’s ad ARPU is almost $4 globally per quarter, and $15 in the US and Canada. Multiplied by 1.8 billion users, it’s easy to see why Workplace at $1-3 per month won’t make a meaningful contribution anytime soon.

Conclusion: a fairly rosy future nonetheless

In short, then, Facebook is likely going to have to make do with ad revenue for the vast majority of its future growth. That’s not such a bad thing, though – as we’ve already seen, the other drivers of ad revenue growth from user growth to price per ad to time spent by users are all still significant drivers of growth in the core Facebook product, and new revenue opportunities across Instagram, Messenger and possibly WhatsApp should contribute meaningfully as well. That’s not to say that growth might not be slower, and possibly quite a bit slower, than in the recent past. But at 30% plus, Facebook will still be growing faster than any other big consumer technology company.

Google’s Increasing Reliance on its Own Sites

After a couple of weeks on vacation, I’m still playing catchup with some of the tech earnings reports that came out while I was gone. Today, I’m tackling an interesting aspect of Alphabet’s earnings, which is the increasing dominance of ad revenue from Google’s own properties versus revenue from third party sites as a proportion of its total ad revenues. The charts I’m using here are largely taken from the Alphabet deck in the Jackdaw Research Quarterly Decks Service, which you can sign up for here. We also discussed Alphabet earnings along with those of other major tech companies on this week’s Beyond Devices Podcast.

Some quick definitions

First off, some quick definitions. Google divides its ad revenue into two categories:

  • Google websites – this includes all revenue from Google’s own websites, including AdWords revenue that is generated on, advertising revenue generated on YouTube,  and advertising revenue generated from other Google owned and operated properties like Gmail, Finance, Maps, and Google Play.
  • Google Network Members’ websites – this includes AdSense,
    AdExchange, AdMob, All DoubleClick-related revenues including DoubleClick Bid Manager revenues, and Other Network products including AdSense for Domains.

Within the core Google segment, these two divisions plus the “Other” category make up the entirety of revenue, and ad revenue from these two sources makes up 90% of Google segment revenue.

Very different growth rates lead to increasing dominance by Google’s own sites

These two segments have been growing at very different rates over the last several years, with Google’s own site revenue growing much faster than its Network Members’ ad revenues, as shown in the chart below:

Google ad segment growth Q2 2016 560px

Though there have been a couple of brief periods (in 2009-2010 and 2012) when Network revenue grew faster than Google website revenue, the pattern has otherwise been fairly consistent: Google’s revenue from its own sites has grown faster. Over the last three years, the gap has been significant – Google website revenue is up 74% over that period, while Network Members’ revenue is up just 17% over three years. That leads to a business that’s increasingly lopsided in favor of Google’s own sites:

Google ad revenue split Q2 2016 560px

In Q2 2016, Google sites passed 80% of total Google ad revenue for the first time. That’s up from 70% in 2011, and 60% around 2006.

Paid clicks growth is the driver

The reason for this discrepancy becomes abundantly clear when you look at the ad metrics Google provides. Every quarter, it reports growth in the number of paid clicks and the cost-per-click (i.e. price) for both the Google sites business and the Network Members business. These numbers bounce around quite a bit, but I find it’s often helpful to index the numbers to a certain point in the past to see the longer-term trends. The chart below shows these figures indexed to the quarter two years ago, Q2 2014:

Google ad metrics Q2 2016 560px

As you can see, there’s a stark contrast between the two businesses here. Let’s start with the number of paid clicks:

  • Google sites paid clicks are up 61% over two years (the number was 67% last quarter)
  • Network paid clicks are at 98% of where they were two years ago, and this number has been relatively flat over that whole period.

Google says that growth in clicks on its own sites has been driven by a combination of growth in the adoption of YouTube engagement ads, improvements in ad formats and delivery, and expansion of products, advertisers and user bases across all platforms, particularly mobile. There simply aren’t similar drivers for the Network business, which Google obviously doesn’t control as directly, and which is in some ways much more mature.

Looking at the cost per click:

  • Google sites CPC in Q2 2016 was only 76% of the CPC for two years earlier
  • Network CPC was 95% of two years earlier, and again this number has been relatively flat over the period, with modest growth in the first year, followed by slight shrinkage since.

The reason for the falling cost per click on Google’s own sites is largely due to growth in YouTube engagement ads where cost-per-click is lower than on Google’s other platforms, as well as changes in property and device mix, product mix, geographic mix, and ongoing product changes, with a smaller impact from currency exchange rates. Some of the same factors drove the modest recent decline in Network CPC as well.

Traffic Acquisition Costs and margins

This all matters a great deal for a couple of reasons, the second of which we’ll come onto in a moment. But the single most obvious reason is that the economics of ad revenue from Google’s own sites is radically different from the economics for Network sites, and that’s because of how Google pays for traffic. On third-party sites, Google pays out most of the revenue to the site owner, whereas on its own sites it keeps the vast majority of the revenue. Google breaks out the traffic acquisition costs (TAC) for both these segments, and the stark differences are shown in the chart below:

Google TAC by segment Q2 2016 560px

As you can see, in the Network business, Google pays out at a rate very similar to the economics of the major app stores, at about 70% of revenue. For its own sites, however, Google’s TAC is a fraction of that, at around 9% in Q2 2016. These payments go to the sources of traffic to Google’s various websites, principally makers of browsers including Apple’s Safari which feature Google as a default search engine.

Interestingly, Network TAC had come down quite a bit for several years, but has recently spiked back up a little, though it’s been within a range from 67-71% for the last five years. Conversely, Google’s TAC for its own sites has been steadily rising, as its cut of revenues under various placement deals has been shrinking. With competition from Yahoo and Microsoft in particular rising over recent years, Google has had to pay more to retain its prime placement in various browsers.

Regardless of the recent changes, TAC remains far higher for third party sites than for Google’s own, though that’s not to say that this somehow translates directly into margins. Obviously, Google’s other costs for running its own sites are much higher than its cost for running ads on other people’s websites. But I suspect the increasing dominance of Google’s own sites as a source of ad revenue is driving the steady improvement in margins we’ve seen over recent years.

The downside of all this

Let’s turn now to the second reason all this matters. Though I’ve just said that the increasing dominance of Google’s ad revenues by money from its own sites is likely good for margins, there’s a downside here too. The problem with this dominance is that Google has to be responsible for essentially all the growth itself, largely by growing its direct audience and finding ways to sell more ads at higher prices. As we’ve already seen, YouTube has been a huge help here in recent years as monetization has really taken off, but I wonder how sustainable that growth will be over time. Google is already attempting to drive revenue through alternative business models like YouTube Red and other subscriptions, and I suspect we’ll see more of this.

But as long as Google is so heavily dependent on revenue from its own sites, it’s going to have to find new sources of revenue which it owns, which might well drive it to make acquisitions (Twitter, perhaps?) and organic investments in new properties. That may be challenging over time, especially as more and more online activity takes place on mobile devices, where there’s simply less room for ads. Better targeting and more lucrative formats like app install ads should help offset that a bit, but it may still be tough to sustain over time. Though Alphabet and Google’s recent results have been very positive, there is here still the core of a bear case against continued growth along the same lines.

Quick thoughts: Microsoft’s ad business

Given today’s news about Microsoft selling its display ad business to AOL and in turn replacing Google as the search advertising provider for AOL, I thought I’d quickly revisit some of my earlier analysis on Microsoft’s ad business.

By way of background, Microsoft has never directly reported the financials for its advertising business, but it has provided enough detail in its past financial reporting that I’ve been able to build a pretty good picture of this business over time. This past quarter, perhaps as a precursor to today’s announcement, Microsoft stopped providing any sort of information about its display ad business, but here’s a quick view of my estimates of Microsoft’s two major ad revenue streams over the past couple of years:

Screenshot 2015-06-29 16.02.44As you can see, Search advertising has been growing very well indeed, almost reaching the $1 billion per quarter mark last quarter, and likely to hit it very shortly, especially with the help of the AOL deal. By contrast, though, Display advertising has been heading south for some time now, and was under a quarter of a billion in revenue for each of the last two quarters of 2014. The split between the two, then, is roughly as shown in the chart below:Screenshot 2015-06-29 16.04.58In other words, search advertising was not only vastly outperforming display advertising in growth terms, but as a percentage of Microsoft’s overall online advertising business. As such, it’s made sense for some time for Microsoft to jettison this part of the business in favor of focusing on the part that’s working: search advertising. Part of the reason for the disparity between the two is general industry dynamics – display has been struggling for other companies too, while search continues to be one of the most effective forms of advertising and to command commensurate rates.  Microsoft’s display ad business, though, was also sub-scale, and hadn’t made the transition to mobile devices and native advertising effectively. Search, meanwhile, has benefited both from positive industry trends and the growth of Bing and Yahoo’s growth in search market share in the last couple of years.

The impact on Google

AOL’s decision to switch from Google to Microsoft is not enormously impactful on Google by itself, but in the context of Firefox’s earlier switch to Yahoo as its default search engine in the US, and the potential for a much more significant switch away from Google by Apple sometime this year, it’s part of a drumbeat of bad news for Google. One of Google’s challenges at this point is that it’s come to compete with many of its erstwhile partners, with Apple as perhaps the most striking example, and it’s arguably starting to pay the price for that strategy.

Apple and Privacy

Apple’s privacy stance has been in the news again this week, mostly because of a speech Tim Cook gave to an event in Washington, which honored him for his (and Apple’s) commitment to privacy and encryption.

Apple’s admirable but over-played privacy stance

My reaction to the speech has been somewhat mixed, as these two tweets indicate:

As I see it, Apple’s commitment to privacy is an admirable one, and one that provides a useful competitive vector as it seeks to differentiate itself against companies like Google and Facebook. But I feel that here, as when Tim Cook repeats the now-hackneyed phrase “when you’re not paying, you’re the product”, he’s overplaying Apple’s hand 1. There’s an underlying truth to both of these claims, but it’s not as cut and dried as Tim Cook makes it seem to be. And I believe that’s likely just positioning, or in other words making the point in the strongest possible black-and-white terms, even though Tim Cook (and the rest of Apple’s leadership) clearly understands that there’s more nuance to this in reality. It’s obvious that many people do value these free services, and are willing to make the tradeoffs inherent in them (if they understand them at all). If Apple really believed everything Tim Cook said in literal terms, I’d be worried about the company, but I don’t actually believe it for a minute, though I absolutely buy Apple’s commitment to privacy and its intention to continue to differentiate on this basis.

Privacy and machine learning

Ben Thompson, in his excellent Stratechery daily email (subscribe here) makes to some extent this same point in his email today, and I agree with those thoughts pretty completely. But he and others have also taken this point further and talked about a supposed downside to all this, which is that by refusing to collect personal data about users, Apple risks not being very good at machine learning. I actually think this point is false, because it conflates three different kinds of data collection and analysis:

  • Data collected on an aggregate basis to allow computers to determine broad trends, better understand text and speech across the entire base, glean information about searches and the best responses to them, and so on. There is nothing user-identifiable about this form of data collection.
  • Data collected about individual users to better customize services and products to their individual needs – i.e. learning favorite places, home and work locations, building patterns of searches to better interpret future searches, and so on. This kind of data collection exists on a spectrum, with some forms of data explicitly provided by the user and others easily inferred, with other data reliant on deeper analysis.
  • Data collected about individual users to build profiles which can be used to target advertising. The only benefit to the user from this form of data collection is making the ads they see more relevant, and the downside is a vague sense of creepiness that third parties are suddenly serving up ads which make use of quite private data from browsing, searches, or the contents of emails (and potentially photos).

The reality is that machine learning takes place across these three different types of data collection, but only the first two are primarily about creating a better experience for users, and Apple has shown itself to be perfectly willing to engage in the first kind with products like Siri and Spotlight, and quite comfortable with at least some forms of the second. It’s only the third category that Apple eschews and which Tim Cook appears to be criticizing other companies for in his public remarks.

Apple’s privacy stance

The table below summarizes my inferences of Apple’s privacy stance on these three categories:Screenshot 2015-06-05 09.37.14If this is accurate, and I believe it is, then Apple isn’t constrained at all when it comes to broad improvement of its services on an aggregated basis, because it’s clearly entirely comfortable with this form of data collection and analysis, and has even recently started crawling websites itself to further this effort. In the second category, it seems very comfortable with building basic profiles of its users through a combination of data users actively pass to its systems and a small amount of data inferred from behavior 2. As such, it’s capable of customizing certain of its services and willing to do so. One good example of this is the customization of its QuickType keyboard on a per-user basis, although Apple is careful to point out that “Your conversation data is kept only on your device, so it’s always private.” When Apple does customize services on a personal basis, it often keeps the information on the device, unlike competitors who use cloud services to make these customizations available across devices, which really does constitute a compromise on Apple’s part. 

However, it’s the third form of data where Apple really seems unwilling to engage in broad data collection and analysis, and that really doesn’t affect its ability to provide its users with compelling services. As such, I think it’s a stretch to suggest that Apple will somehow always be inferior at machine learning because it eschews targeted advertising – the two are for the most part separate, and it’s entirely possible for a company like Apple to engage in both broad based aggregate machine learning and machine learning on an individual user basis without either compromising privacy or engaging in the type of behavior it’s criticizing in others.

Apple still has work to do in machine learning

None of this is to say that Apple is just as good at machine learning as competitors. I honestly believe this is the core of Google’s differentiation as a company and Facebook seems to be becoming increasingly strong in this area too. Apple is currently weaker in these areas, but I don’t believe that its privacy stance is the reason – I just think it hasn’t chosen to invest in these areas as heavily, and to the extent that it is doing so now, it has some catching up to do (and it seems that at WWDC Apple may announce some advancements in this area relating to Siri).

The same applies to Apple’s cloud services in general – this simply hasn’t been a major focus for Apple so far, for a variety of reasons, and its cloud services simply aren’t as strong as competitors’. Google Photos is a great example of that – its seems to do certain things much better than Apple’s Photos product, and a large part of that is about machine learning. And yet Google Photos is also a great example of exactly what Tim Cook is talking about – the inherent unease about sharing such personal data with a company you know would like to use it to target advertising to you. If Apple produced a similarly compelling product, there would be none of that unease, but you’d likely pay for the privilege of using it directly. Therein lies the real difference between Apple and its competitors.


  1. For a more nuanced analysis of the latter claim, please see this post
  2. See this earlier post for an example of this, which Apple hasn’t shouted about much to date

Quick Thoughts: Apple and advertising

I wrote a piece almost a year ago on Techpinions on the subject of Apple and its advertising products. Today, Jay Yarow posted a piece on a similar subject, but reached a different conclusion, at least in the headline. My piece was actually about the inherent conflict at Apple surrounding ads, and I think Jay’s piece largely has the same theme. I wanted to post a quick update here following some discussion on Twitter on this topic earlier in the day.

Jay’s key point is this:

For 0.3% of Apple’s revenue, Apple’s ad business cuts against what Tim Cook calls one of his company’s “core values.” That makes no sense. Apple should shut down iAd.

I’ve actually made a similar point about Microsoft here, and advertising is actually a bigger revenue stream for Microsoft than it is for Apple, so why don’t I agree with Jay’s conclusion?

The biggest single reason is the role that advertising plays in the TV market, a market that Apple seems very interested in and one it is likely to enter at some point. The chart below shows the contribution of ads as a percentage of revenue for three categories of US TV-related businesses based on a sampling of companies that report this split:Ads as percent of revenue TVAs you can see, for each of these sets of companies, ads are over 50% of revenues, and although there’s been a slight drop at the TV stations due to the rise of retransmission fees, it’s still an enormous portion of overall revenue. Now, there are, of course, two prominent video businesses in the US which don’t make any money from ads: HBO and Netflix. But both of them rely heavily on catalog content, mostly movies, with a handful of original series at any given point in time. Neither is a comprehensive replacement for a pay TV service, and both are used largely as complements to, not replacements for, the traditional cable TV bundle. If Apple is serious about getting into video subscription services, it has to offer live, linear programming, and that means ads will be a big part of the business model.

The challenge for Apple is how to embrace this model without compromising its principled stand on advertising. Given the increased focus on targeted advertising, that’s going to get more complicated, but the answer may lie in allowing brands to bring their own consumer data and match it with Apple users, in the way Apple is apparently planning to do with iAd. That way, the targeting is all done by the third party, with Apple only offering to match a user ID with a profile from the third party. This takes all the unpleasant profiling and targeting out of Apple’s hands, while still making its platform useful and attractive to advertisers. That’s a subtle distinction, but it may be a critical one if Apple is serious about TV.

Twitter’s new ad product won’t help

I’m making this part of by earnings series, even though Twitter hasn’t actually reported yet, because it hits one of the key points I expect to come up in Twitter’s earnings this week and gets at one of the fundamental concerns I have about Twitter’s growth going forward.

Today, Twitter announced that it would start offering advertisers the ability to serve their promoted tweets off the Twitter platform itself. This is an important step for Twitter because its key growth challenge has been that the number of monthly active users on the platform itself isn’t growing rapidly, and it sees its biggest opportunity in serving what it terms “logged out users,” or those who see tweets without being explicitly logged in to Twitter or one of its apps. This is something I’ve talked about quite a bit here over the last few months, notably here and here. The challenge as I’ve seen it has been that (a) Twitter hadn’t articulated how exactly it would monetize that usage and (b) Twitter can’t use any of the data it has about its users in these third-party contexts, so the advertising would be only minimally targeted.

Today’s announcement gets at the first of these problems, in that it’s the first meaningful articulation of how Twitter can play outside of the core Twitter experience. But it doesn’t solve the second problem. In fact, the way Twitter has chosen to implement the promoted tweet product is a great illustration of how steep a hill Twitter has to climb here. I’m reposting below a chart I’ve used several times here in relation to the effectiveness of advertising:

Advertising relevance vs timelinessWhat this chart attempts to demonstrate is that there are two characteristics to effective advertising: relevance and timeliness. What makes search advertising so effective is that it hits both of these – i.e. an ad next to search results is relevant now to that user. Most other forms of advertising can at best produce relevance without timeliness, and in some cases neither. The problem with Twitter’s new ad product is that it looks an awful lot like a classic display ad, with little or no connection to the context and with no use of Twitter’s demographic data about the user. It’s likely that these ads will use some targeting from the property on which they appear, assuming those properties have that information, but Twitter itself brings nothing to the table here but the format.

Twitter makes the argument with regard to Flipboard specifically that the app already shows normal (non-promoted) tweets, so these promoted tweets won’t seem too out of place, but it’s a far cry from seeing a promoted tweet in the midst of a stream of regular tweets in the Twitter app or on This is at best very loosely native advertising. And very few other properties show enough tweets in native format that these ads won’t look just like another sort of banner ad.

It seems investors don’t mind, at least in the early reactions: the stock is up 6% today. But to my mind nothing about today’s announcement really gets at solving the fundamental problems solving Twitter. We’ll see how this is presented on the earnings call this week, but given this and recent comments from Twitter founders Ev Williams and Jack Dorsey (ironically, in a tweet storm I can’t easily link to because he didn’t post it right), I’m wondering whether this is a pre-emptive strike ahead of another set of poor user growth numbers. Both Williams and Dorsey seem to be saying that we need to be looking beyond Twitter’s actual numbers at the societal impact it has, which seems like a strange thing to say if the numbers are actually going to be any good.

Thoughts on Twitter earnings for Q3 2014

Note to new readers: this is part of a series on major tech companies’ Q3 2014 earnings. I do similar analysis each quarter for certain companies. Previous pieces on Twitter may be found here

Last quarter I talked a little about the challenge of Twitter’s user growth, and what I uncovered as its renewed emphasis on logged-out users (a theme Dick Costolo had first talked about three years ago and to which he has now returned). I and others have talked about Twitter’s different user groups and on yesterday’s earnings call Costolo explicitly spoke about three concentric circles (he actually said eccentric, but I’m fairly certain he meant concentric) representing Twitter’s users. That’s something I’ll return to below. There will be three major points to this analysis:

  • Twitter has given us metrics to measure its growth and potential, but it keeps backing away from or hedging on these without providing better ones
  • Twitter’s core user growth has slowed in recent quarters, and though it wants us to think about a broader group of users, this broader group is both ill-defined and much harder to monetize
  • Twitter is very successfully managing the same transition to mobile advertising as Facebook, but it’s not yet clear how sustainable this business will be for either company.

Twitter’s metrics aren’t telling a good story

The three main metrics Twitter has established for tracking its progress are:

  • User growth: as measured by growth in monthly active users (MAUs)
  • Engagement: as measured by timeline views per MAU
  • Monetization: as measured by ad revenue per thousand timeline views.

These three may be seen as levers, or multipliers, for growth. To the extent that all three are improving, they should drive faster and faster growth for the company, as each change in the first is magnified or multiplied by growth in the second, which in turn is magnified or multiplied by the third. The challenge is that of these three metrics, only one is heading in the right direction, and that’s ad revenue per thousand timeline views:

Ad revenue per 1000 timeline views

As you can see, international monetization continues to lag significantly behind US monetization, which is a function both of Twitter’s underdeveloped overseas sales channels for ads and the lower ad spend in every other market around the world. But both sets of numbers are moving in the right direction, and in the case of US revenues very rapidly so. I’ll come back to the drivers for this growth later, but this is the single best sign about Twitter’s business: it’s able to monetize what usage there is better and better, and this in turn has driven strong overall growth in ad revenues (in thousands of dollars):

Twitter ad revenue

The challenge is that this growth is strongly tied to direct sales, which is an expensive proposition in its own right. Even as some of Twitter’s other cost lines have shrunk as a percentage of revenue, sales and marketing costs are rising over time: Continue reading

Thoughts on Twitter earnings for Q2 2014

Note: This is part of a series on major tech companies’ Q2 2014 earnings (e.g. Google here, Microsoft here, Apple here and here, Netflix here, Facebook here and Amazon here). My post about Twitter’s earnings last quarter is here.

Non-logged-in users and “Twitter Everywhere”

I posted yesterday evening about the growth challenge that Twitter faces (and which it shares with Foursquare). Given the focus during today’s earnings call on non-logged-in users, I’d say this is very much at the top of Twitter executives’ minds at the moment too. Let’s dig into that a bit. I was reminded of something I read quite a while ago on Fred Wilson’s blog:

Let’s remember one of the cardinal rules of social media. Out of 100 people, 1% will create the content, 10% will curate the content, and the other 90% will simply consume it.

It occurred to me during today’s earnings call that Twitter is experiencing this in quite a different way from some other services. Facebook and other social networks require you to log in to gain any meaningful benefit from interacting with the service, because Facebook posts are largely private and because the benefit is largely derived from seeing posts from people you know. As such, this 1%/10%/90% split Fred Wilson talks about largely applies to logged in users. But with Twitter, the big difference is that the vast majority of posts are public, and therefore people don’t have to be logged in at all to get value out of the service. So, in fact, the logged-in monthly active users Twitter has reported so far (271 million at the end of June) might represent some mix of the 1% and 10%, while Twitter might very adequately serve the other 90% as non-logged-in users. This is what drove much of Dick Costolo’s prepared remarks during the call: that the real size of Twitter is a multiple of the size it appears to be based on the number of monthly active users (MAUs) it reports.

It turns out that I had forgotten the context in which Fred Wilson quoted those numbers, which was an account of a state of the union address Dick Costolo gave at Twitter HQ in September 2011. During that session, Costolo very much hit on the same theme he returned to on today’s call, and I direct you to Wilson’s post for a recounting of some of the key points. What’s very interesting is that back then Costolo was quoting a 4:1 ratio of total users to logged-in users, or about the same 3x multiple of non-logged-in to logged-in users he quoted on today’s call. What’s surprising is that Costolo doesn’t seem to have talked very much about this in the three years between then and now (if you search for Costolo Logged In Users on Google, you’ll find posts from 2011 and today, but almost nothing in between). But I wouldn’t be at all surprised to see some sort of unique users metric emerge next quarter now that Costolo has provided the strategic rationale for measuring it. The challenge will be monetizing these users going forward, something Twitter doesn’t even attempt to do today.

Downplaying the World Cup while learning lessons

Costolo and the other executives on the call were also very keen to both play up and downplay the impact of the World Cup. On the one hand, they want to play it up as an example of what Twitter can do on an ongoing basis by way of creating topic- and event-driven experiences that are different from the classic Twitter feed. On the other, they were keen to downplay the impact of the World Cup specifically as a factor in driving the strong MAU growth it saw this quarter. On the other hand, they suggested the World Cup did have an effect on engagement, which is definitely illustrated by this stark contrast in sequential MAU growth from Q1 to Q2 (turns out the US still doesn’t care much about soccer):

Sequential growth in MAUs US vs InternationalLook at the difference in Twitter’s key engagement metric from Q1 2014 (no World Cup) to Q2 2014 (World Cup). International engagement rose, while US engagement actually fell. You can see this in the shape of the overall graph for timeline views per MAU, which appeared to improve in Q2 but were largely driven by International growth:

Timeline views per MAUThe big question is to what extent the World Cup experience is replicable for other events and topics, which is clearly Twitter’s plan.  The World Cup is either the most or second-most watched event in the world (sporting or otherwise), and along with the Olympics only happens every four years. Putting a lot of investment into creating  a curated experience around the World Cup clearly makes sense, and its major news bites make for good tweets because they’re short and to the point (Brazil 7, Germany 1 takes considerably fewer than 140 characters).

But how does this experience translate to hundreds of other events and topics Twitter would want to do this for, and to what extent can those experiences be machine- rather than human-curated? Twitter has just 3300 employees – how many of them can it afford to attach to human curation projects for the many other events and topics Twitter might want to recreate the World Cup experience for? And would there be any meaningful return on that investment for most of those events? If the World Cup drove a small increase in engagement for international users, how many events that happen more frequently would it have to bundle together to drive a similar impact?

Data and Licensing is growing faster

Twitter, like Facebook and Google, derives about 90% of its revenue from advertising. However, unlike Facebook, which is deriving an ever-greater proportion of its revenues from advertising, Twitter has actually quietly been going in the opposite direction for the previous two quarters and just took a slightly bigger step in that direction this quarter:

Twitter revenue mixThe reason is that Data and Licensing revenue actually grew as a proportion of total revenue over the last several quarters, and  accounted for 11.1% of revenue in Q2. There wasn’t much detail on the earnings call, but the 42% sequential and 90% year on year increase was put down to a combination of the Gnip acquisition (which closed in April) and revenue from the mobile ad exchange (the former MoPub). Presumably, much of the revenue generated by the MoPub business is actually recorded in the advertising line, but a portion apparently comes from Data and contributed to this growth. In my post on Facebook’s earnings I worried about their increasing reliance on mobile advertising and the paucity of other possible revenue sources in the near term, so I think this secondary revenue stream for Twitter (albeit still small) is important.

The limited opportunity for app install ads

Today, Twitter formally introduced its new mobile app install ads product, joining Facebook and Google in what is becoming a crowded space, with Yahoo apparently waiting in the wings too. I had a quick look at this space in the context of Facebook’s Q1 earnings a few weeks ago, but wanted to drill down deeper, especially now that we know more about app revenue through Google Play. The upshot of all of this is that the opportunity for mobile app install advertising, though growing rapidly, is not big enough to provide a significant revenue stream for all these companies. In other words, there’s gold in them there hills, but not enough to justify the gold rush we’re seeing into this space.

First, a quick primer on mobile app economics. Some of the major app companies are public, and report data themselves, while several third parties also report data on the topic regularly, allowing us to draw a few conclusions:

  • Revenue from advertising is a factor for some apps, but the vast majority of revenue today (likely between 80% and 90%) comes from pay-per-download and in-app purchases. As such, the revenue numbers for the two major stores – Apple’s App Store and Google Play – likely account for a significant proportion of total revenues from apps 1.
  • Developers pay Google, Apple or other stores 30% of their gross revenues from these stores, keeping 70% for themselves. Thus, if they’re to make a living, it will be by keeping their other costs contained within that net revenue figure. That needs to cover development costs, ongoing operating costs (salaries, hosting, care, etc.), and costs to promote apps.
  • Sales and marketing costs for most successful app makers sit between 10% and 20% of gross revenues. App install advertising will come out of this budget, and may indeed make up most of it. This percentage may be significantly higher for apps early in their lifecycle and therefore promoting themselves heavily without yet seeing significant revenue, but it will tend to return to that average over time.
  • Thus, anywhere between 40% and 50% of a typical app developer’s gross revenue may go to the store commission plus sales and marketing, leaving about half for all the other costs of running the business.

Given these facts, let’s look at total gross revenue opportunity from the two major app stores. I’ve added 15% to my estimated gross revenue from the two stores to account for the advertising opportunity.

Total app store revenues from Google Play and App StoreNow, let’s think about the size of the market for mobile app install ads, which as we’ve already said will have to come out of that sales and marketing budget. To put it in context, we’ll compare it to Facebook’s mobile advertising revenues, since Facebook is the largest player in this market today and a substantial proportion of this revenue comes from app-install ads today. In the chart below, I’ve plotted Facebook’s mobile ad revenues against two views of the app install ad opportunity – one a bull case and one a bear case. The bull case assumes that the app install opportunity is 25% of total gross revenues, and adds 15% to store revenues to account for ad revenues. The bear case assumes that the app install opportunity is 15% of gross revenues, and adds a smaller 10% to store revenues to account for ad revenues. My own view is that the bear case is likely closer to reality. Continue reading


  1. For simplicity’s sake, I’m excluding the revenue opportunity through other stores, because they account for a tiny proportion of overall revenues. Adding them in would not significantly affect the numbers.