Description

Investment Thesis

Facebook is a network effect business with few competitive threats as the world’s largest social network. Increasing free user generated content reinforces the network effects, allowing the company to take market share of the $600b+ advertising market. I believe that management’s persistent commentary about headwinds from News Feed ad load limits reaching a peak in Q2 2017 has been more about managing the stock price and will be a non-event for Q3 2017 with Facebook delivering a 2H 2017 revenue growth rate of 45% vs. Street consensus forecast of 38%. This is an additional ~$1b in revenue that translates to a full year GAAP EBIT of $21.0b vs. consensus estimates of $20.4b. Management has also been sandbagging expense growth guidance in the prior 2 years, albeit easy to do in the context of >90% incremental margins and high growth. Facebook’s high market cap (#4 globally) and headline risk around future regulation, political ads, fake news, etc. may be factors supporting current mispricing. The bottom line – Facebook is executing. Facebook’s software development teams continue to build new tools that make it easier for marketers to target ads that, along with increasing advertiser competition, drives the price per ad higher (+6%, +5%, +14% and +24% in the 4 most recent quarters). Facebook also has additional ad inventory coming that mutes the impact of News Feed ad load. CMOs continue to arm themselves with marketing and ad technology to measure the ROI on Facebook ads and other digital ad spend. I forecast Facebook trading at 17X 2018 EV/EBIT with high confidence that sustained growth will be supported by both e-commerce and digital transformation trends. Given the cash generating ability and capital light nature of the business model, this is a low multiple for a growth company with an excellent moat.

Business Model Overview

Facebook offers a free place for individuals, affinity groups and businesses to express themselves and share information on one of four properties: Facebook, Instagram, WhatsApp and Messenger. Facebook and Instagram are the focus in this report. Facebook is the top app in the U.S., with more downloads than the Chrome Browser. It is a library of daily activity for 1.3b daily active users (+17% y/y) and 2.0b monthly active users (+ 17% y/y) globally including about ~64% of the US population. The majority of users access Facebook via a mobile device where the average user in the U.S. is spending 50 minutes per day on the platform. Seventy million businesses are listed on Facebook and 5m of them advertise on the platform. Instagram has 15m businesses listed with 2m advertisers (vs. 1m in March 2017).

Facebook owns 7 data centers and leases 2 facilities. The company has 20,700 employees and is based in Menlo Park, CA.

FB largely monetizes Facebook and Instagram today. Facebook is a free website and mobile app for personal and business use to form connections and groups to share text updates, text comments, location, photo and video content. The News Feed is the central viewing area that serves as a personal newspaper crafted by proprietary software every few minutes. This flow of information provides updates on activities of people in the user’s social network. Instagram is a photo-based sharing service where individuals and businesses can post high quality photos and short comments. Instagram has 800m users (500m DAU), up from 400m users only 2 years prior. Users typically “follow” or subscribe to an Instagram feed, thereby forming one-way connections. Instagram and Facebook have a common advertising system so marketers can easily target the same user on both platforms.

While the 20th century heralded the age of mass advertising communication by T.V., radio, and newspapers, the old paradigm is being rapidly supplanted by one-to-one advertising communication enabled by smartphones and social networks.  The company monetizes this attention by helping businesses sell targeted and native advertising to Facebook and Instagram users who natively build profiles about themselves. The primary determinant of advertising ROI is targeting the right message, to the right person, at the right time. Facebook Ads Manager is a self-service ad submission, targeting and analytics tool that allows businesses to advertise and measure the returns on ad spend (ROAS). Facebook also has a Marketing API that allows advanced digital agencies to directly connect to the platform and run programmatic campaigns. Brands uniquely benefit from the platform because they can more easily measure how individuals engage with emotionally driven brand advertising.

WhatsApp (1b DAU, 1.3b MAU) and Messenger (1.2b MAU) are not yet monetized. Facebook only recently released an SDK to allow firms to build applications and chat bots for these messaging tools. Tencent’s WeChat in China has been transformed into both platform and mobile portal for ecommerce and is delivering ~$7 in ARPU on ~1b MAU. It is too early to tell how Facebook will monetize these networks.

Facebook also helps deliver ads across the internet. Facebook Audience Network (FAN) delivered a >$1b run-rate by serving ads to apps at the end of 2016. Facebook Connect is a universal login that allows ID-based tracking (superior to cookie based tracking) so that marketers can follow Facebook users even when they navigate away from the Facebook app or across multiple devices. The ability to follow users around the web is critical for establishing which ads are attributed to desired actions (i.e. purchases, form completions, etc.).

Moat / Opportunity

“Hacker culture” results in the rapid release of targeting tools which leads to advertiser competition and helps drive prices per ad higher –

The ability of software firms to produce new features and functions typically declines as the size of the code base and the number of engineers increase. However, Facebook’s “hacker culture” is consistently executing the release of new features that improve user engagement and keep users on the Facebook platform. There is also significant innovation for targeting tools via the Power Editor ad tools that drive monetization rates (i.e., ARPU). These features are released with very high frequency often without documentation or even notice. Facebook marketers are generally able to deploy these new tools because they work out of the gate and the reward from using new targeting techniques on Facebook can be lucrative given the scale. This dynamic sharply contrasts to the broken innovation cultures at Twitter, Yahoo and to a lesser degree at Snap and Yelp where large audiences are under-monetized because self-service advertising tools are less developed and marketers struggle to test and learn about the platform to increase return on ad spend (ROAS).

Facebook knows more about its users than any other direct advertising medium, and the associated targeting ability combined with advertiser competition is driving prices per ad higher. Facebook allows marketers to target individuals by age, gender, geography, relationships, interests groups, lifestyles, etc. – and marketers know that this information is not inferred from behavior, but is rather directly shared by Facebook users.

Examples of ad targeting that drive ad prices higher:

• Custom audiences (CA) tools have been extremely effective. Marketers upload proprietary customer lists (names, phone numbers, email, address, etc.) into Facebook where up to 60% of these customers can be matched to Facebook users.  Custom audiences then become substitutes for email and direct mail, serving as powerful methods to retarget high value customers. Custom audiences can also be created to track offline activity. Once someone sees an ad, Facebook can evaluate if that person visited a store. There are a variety of targeting tools being developed around this use case. Given retail footfall challenges, these tools are being actively deployed.

• Facebook allows marketers to create CA to target certain users based on their location in the sales funnel via prior interactions with ads, native content, links or even shopping carts. For example, Facebook is beta testing ways to build an audience based on people who have shared a link to certain content. A “Dwell time” metric in beta testing seeks to target users who spent a certain amount of time looking at an ad (without clicking on it) and then retargeting them again in the future. This is similar to tools in YouTube that allow marketers to retarget 6-second bumper type ads to the same users that previously watched an entire longer form ad from the same brand without skipping it.

• Lookalike audiences are also extremely effective. These audiences are constructed by a Facebook algorithm to find people most like another CA. Facebook has sufficient scale and data at 2b MAU to accomplish feats that marketers could only dream about in the past.

• Rich content drives price per ad higher for Facebook and Instagram. Richer ad formats with video and photos are experiencing rapid growth and always command higher CPM from marketers due to better engagement. Brand marketers have only recently focused on developing video content for social media platforms. The strategies and formats for successful social media are vastly different from TV ads which dominate the digital assets of many brands. The gradual shift away from $190b global TV market is likely to continue. Video consumption is an increasing activity on mobile devices and marketers want to deliver commercials. Nielsen, Social Code and other data providers are now able to match in-store purchases to ID-based profiles at Facebook/Instagram to demonstrate brand lift and campaign efficacy. Facebook has target rating point (TRP) buying which allows advertisers to plan, buy and measure video with Facebook the same way they do with TV, including verification via Nielsen. The long-standing trend of TV viewership being crowded out by other digital media makes social media attractive to brand advertisers. According to Alan Gould, CEO of Nielsen AG (2010) “Facebook has the opportunity that Google only wishes it had – the ability to build a credible proposition for the largest brand advertisers”. Early reports from Nielsen as to Facebook’s impact on brand lift are encouraging. I heard from a Target executive who highlighted that too many campaigns start with a TV commercial in mind first but should instead focus on the data to tell them where a concept should go to choose the right media mix. I spoke with one creative ad agency that tests 8 forms of content on social media first before considering investing in TV distribution. Dana Anderson, former CMO of Mondelez, noted that the media world has changed so fast that it is no longer possible to buy one commercial and achieve an objective. Nanigans provides a quarterly report on about $700m in annual ad spend (~1.8% Facebook revenue). In Q2 2017, they report a 49% increase in click through rate for ads. This is resulting from a combination of factors that are allow marketers to deliver relevant ads that have increasing rich and dynamic content that users like.

I have spoken with more than 18 participants in the digital advertising, either by phone or at industry conferences. They are almost uniformly supportive of the thesis that Facebook delivers a high level of ROI to marketers.

• “We’re very positive on the efficiency of Facebook compared to all other paid media”.

• “The CPM that clients are spending today can be much higher and they will be happy with the results”

• “Facebook is the new network like ABC or NBC”.

• “Anybody who is in the whole marketing game and doing digital marketing and have a company, up to at least 70% of the budget is going towards Facebook for sure. You cannot do without Facebook because it’s very, very good at targeting as you get really good results. It can be expensive but if you optimize it effectively, you get some really great results.”

• "Facebook is just getting started. It is phenomenally effective."

Facebook is bringing new inventory online that reduces impacts from ad load limits in the News Feed-  

• Facebook does not segment Instagram revenue but digital ad agencies indicate clients have high interest in Instagram and are achieving positive results. I expect this inventory will be increasingly valuable, particularly if marketers allocate incremental digital budget dollars to Instagram instead of substituting Facebook budgets. Instagram had a 75% y/y growth rate in MAU in Q1 2017.  As of March 2017, ads are now available in the wildly successful Instagram Stories (Snapchat clone) with 250m DAU.

• Instant Articles allows publishers to deliver original content inside Facebook that improves load times and provides inventory for native ads in line with content. Publishers have been lukewarm about the results but Facebook is now testing even subscription services with revenue sharing that appears favorable to publishers.

• The newer Canvas and Collections ad types allow advertisers to build small microsites inside an ad to increase engagement with multiple products and services.

• Facebook Live content grew by 4X in Q1 2017 with 20% of all Facebook videos being live. Ad breaks can be inserted into this content similar to the way ads are inserted into longer form YouTube videos.

• Facebook Groups are largely ad-free but are likely to begin to include ads soon. The average Facebook user is a member of 30 groups with 2-3 of these groups which are considered “meaningful” to that person. In June 2017, Facebook announced that about 150m users were in meaningful groups with a goal of getting 1b people to join meaningful groups.

• Facebook Search is actively in development and will likely eventually offer a Google Adwords style bidding models with the exception that result sets will be delivered with social context (i.e. AAA Plumbing is liked by 3 friends of mine who can provide trusted referral information).

• Messenger inventory is also coming online and didn’t go global until July 2017.

Rapid increases in business penetration and lower organic reach are additional drivers of price increases –

To date, about 5m businesses advertise (vs. 3m in April 2016 and 4m in Dec 2016) on Facebook while 70m businesses manage their business pages on Facebook and freely benefit from the viral nature of the network when happy consumers “like” products/services. When businesses produce content on Facebook, a subset of the people that “like” that business will see the content in their News Feed. Such views are considered “organic” views and businesses consider it a form of free advertising. The rate at which the News Feed is delivering organic posts from businesses continues to fall as the amount of branded content has increased and the Facebook algorithm needs to be more and more selective of what it chooses to display in the News Feed. Typical brands have seen their organic reach decline dramatically over the years with some brands declaring that organic reach is all but dead. Marketers now characterize Facebook and Instagram as a “pay-to-play” mediums where businesses need to pay Facebook to promote their own content in the News Feed to increase reach. As the number of advertisers increase, the auction model leads to price increases. This is analogous to what has been occurring at Google over the last few years with organic search results being replaced by ads, directly benefiting ad clicks volume.

Software tools are improving and the CMO role is becoming more and more data driven–

The last 5 years have witnessed considerable advancements in advertising technology (adtech) and marketing technology (martech). The 5,000 adtech and martech firms (as depicted in the LUMAscape) is easily discounted as some sort of venture “bubble” but my view is that these tools are wringing inefficiency out of a notoriously inefficient ad industry that is still wedded to relationship-based (e.g. sports tickets, steaks, martinis) ad agencies that have been overcharging for decades for advertising that does not work very well. The whining from public ad agency heads about the dominance of Facebook and Google has reached a crescendo for a reason.

Relative to digital products like Google Adwords or even standard TV campaigns, social media marketing is more complex with numerous layers of user engagement and more types of creative content. New software tools are empowering firms to measure and analyze the abundance of data in social networks to even include how users respond to campaigns. Many firms are still building out their internal digital teams and data warehouses to make this type of analysis possible.

The upshot of these technologies is that marketing and advertising investments can be flexed according to efficacy. Fully measured campaigns can be thought of as variable costs rather than fixed costs. I have repeatedly heard leading CMOs highlighting the importance of their relationship to CFOs. Increasingly these data are enabling real-time marketing strategies via marketing automation technology. Programmatic ad buying continues to gain traction as marketers seek customers across a variety of digital platforms. Pepijn Rijvers, CMO of Booking.com, is the largest single buyer of keywords on Google Adwords. Rijvers recently said: “I think regarding that capacity to capture and analyze data for the marketing industry, there is still an enormous opportunity ahead of us. I think that in the 10 years to come we will see quite a lot of disruption and innovation in the marketing science arena particularly”. I think CMO/CTO skill sets are still playing catch-up to digital marketing while undergraduate and even graduate programs struggle to deliver marketers with sufficient data management and statistical analysis skills. These are factors that have caused digital ad spend growth to lag the attention shift to mobile devices in the last several years. The CEO of the media agency WPP characterizes the gap by citing that agencies only invest 6% of campaign money in digital even though digital is 25% of consumer’s time.

Consumer “attention” to mobile is increasing faster than ad spend –

Attention is the finite currency for marketers. To state the obvious, mobile is taking attention share because it is “mobile” – people have more time to consume media wherever they are since mobile devices work better on even faster networks. Mobile devices take attention share away from T.V., books, interpersonal interactions, etc. If 10 hours per day are sleeping and other non-discretionary activities, there are 14 hours of available attention per day. In the U.S., average daily time on mobile devices has increased from 93 minutes in 2011 to 300 minutes in 2016.

While mobile ads were virtually non-existent in 2011, today U.S. mobile ad spending was $46b or only 23% of all advertising spend according to Comscore, much less than the 36% (5 out 14 hours) of available attention spent with mobile devices. This gap will eventually close and could even overshoot if digital marketers can achieve higher returns on ad spend compared to TV or other mediums that more difficult to measure.

Capital Allocation & Management History

I have a bias that favors founder led businesses and Zuckerberg continues to lead the strategic direction of the company and is focused on responsibly growing the size and quality of the Facebook community. Thefacebook.com was founded in Zuckerberg’s undergraduate dorm room in 2004 to provide simple connections between Harvard students to share basic interests and course lists. It was quickly expanded to other colleges and eventually opened to the public in 2009. Zuckerberg had several opportunities to cash out but he thought he was “not going to have a better idea in his life so he thought he should stick with this one”. Microsoft offered $15b in 2007 and he declined the offer. He would have netted $4b at the age of 22. The disinterest in personal wealth is consistent with his recent pledge to give away 99% of his shares eventually which are valued today at $71b.

From inception, the historical focus of the business has not been revenue generation. As Mark Zuckerberg described to investors in the S-1, “Simply put: we don’t build services to make money; we make money to build better services.” The primary focus has been to build a platform that is reliable, sticky and enjoys network effects. For a young founder of a fast growing business, Zuckerberg is unusually long-term oriented. Zuckerberg played a key role in isolating the Facebook community from bad, untargeted advertising which would have generated near term revenue but also would have hampered the development of the social network and may have made it more susceptible to competition. This ethos persists today. From the moment the application was launched in 2004, the primary use of cash raised from early round investors was to invest in software and server architecture to deliver a responsive website with few outages – issues that had limited competing social networks at the time. Facebook has not suffered usability issues and has likely over-invested in its server and data center capabilities. Recently, even Google agreed to adopt Facebook’s OpenCompute project as a useful open-source data center architecture.

Other than the acquisition of FriendFeed in 2009 ($50m), Facebook grew organically through 2011 where invested capital of $3.2b was used to build the network to 845m monthly active users (MAU) at a cost of $3.77 per MAU. In 2011 average revenue per user (ARPU) was $3.15 per year which increased to $15.98 in 2016. Thus, the early investments in building the network have been shown to a prudent use of capital. The increase in ARPU has been almost entirely organic and is a key valuation driver going forward.

Aside from 3 smaller technology acquisitions in 2012-2013, Facebook’s most important acquisitions have been Instagram, WhatsApp and Oculus Rift. Instagram was acquired for $715m ($300m cash/12m class B shares/11m unvested class B to vest over 3 years). Although Instagram revenue is undisclosed, Street estimates are for $4b revenue in 2017. This is amazing growth considering Instagram only launched ads in 2014 and highlights the ability of the ad platform to cross sell other products. There are many well designed ads in Instagram and agencies report that advertisers are happy with their results thus far.

The most controversial acquisition to date was Oculus Rift in July 2014 for $1.9b (78% in stock). This virtual reality (VR) hardware product does not appear to complement the existing business model. Facebook believes that VR is a significant new social networking and gaming platform. VR developers using Rift in the past year have expressed considerable excitement about the platform. I can’t value Rift at this time and Facebook has no experience with successful hardware distribution.

Also in 2014, Facebook acquired messaging platform WhatsApp for $17.2b (73% in stock - 178m shares). Since the WhatsApp acquisition, its number of active users has more than tripled from 450m to 1.3b. WhatsApp was charging $1 per year after providing the first year for free. They pivoted from this pay model and the platform is now free for users. In general, communication services such as Messenger and WhatsApp will be more difficult to monetize from an advertising standpoint. They are likely to charge businesses fees for using the platforms for marketing and customer communication. These products are potentially large call options.

Other smaller acquisitions were completed for $485m in 2014. This included LiveRail, which was an ad exchange platform (supply side platform; SSP) that focuses on mobile video and allows publishers to expose their inventory to automated demand-side platforms (DSP) who then bid for that inventory. Facebook acquired LiveRail to improve LiveRail targeting capability with Facebook data. LiveRail as since been shut down.

There were no material acquisitions in 2015, 2016 or 2017 YTD.

Zuckerberg was mentored by Don Graham of The Washington Post. Graham almost invested in 2005 but Accel Ventures came in with a higher offer. Graham was a Director from 2009-2015 when he was required to roll off due to reaching 70 years of age. Although Zuckerberg has 60.1% of voting control, the Board is high quality: Sheryl Sandberg (COO), Marc Andreesen, Erskine Bowles, Susan Desmond-Hellmann (CEO Gates Foundation), Reed Hastings (Netflix), Jan Koum (CEO WhatsApp), and Peter Thiel. The key executive team all has 5-7 years tenure in their respective roles.

Competition

In the U.S., eMarketer forecasts that Google and Facebook, taken together, will garner 63.1% of the incremental dollars invested in digitals ads in the U.S. in 2017. Current and forecast market shares are shown below:

Digital advertising can generally be divided between display (47% of digital) and search (46% of digital). Display is forecasted to grow over the next several years, largely at the expense of TV budgets, due in part to cord-cutting and various OTT content services.

Digital Display –  Global display market share is Facebook (25%), Google (13%; net of TAC) and Alibaba (9%). Other than native display ads on Facebook and Google properties, display markets are highly complicated by programmatic buying and various types of negotiated deals - all of which are complicated by ad fraud claims.

Digital Search – Marketers note that most visits to Facebook are not because people are in search of something and thus the buying intent signal is low. Google Adwords has about 59% market share in paid global search according to eMarketer. Google’s brand in search is very strong and advertisers are comfortable with Adwords because Google Analytics provides full click attribution to evaluate the ROI. In most of its markets, it is the dominant search platform and many firms are essentially required to both optimize their organic search rank and improve visibility via AdWords. For general purpose search on the desktop in the U.S., the key players are Google (64% market share), Bing (22% market share), Yahoo (12% share), Ask Network (1.7%), AOL (1%). Globally, mobile search market shares are Google (80%), Bing (10%) and Yahoo (8.4%).

With 55% of product search starting on Amazon.com and 50-60% of incremental e-commerce dollars going to Amazon, retailers are doing everything possible to avoid losing sales to Amazon and/or listing their wares on Amazon.com and paying the 15% listing fees charged by Amazon. Amazon is also in the early stages of developing its own advertising products based on its customer’s search and purchase history. These ads are now being sold to people who are already selling inside Amazon but can also be accessed by non-endemic advertisers who would not sell on Amazon necessarily (i.e. financial products, etc.).

I discuss the general trends here in the context of the U.S., the largest advertising market globally and the one most relevant to Facebook today.    

TV – While TV currently has the largest share of ad spend today, it has now been eclipsed by digital ads in 2017. A key signpost for the industry was the August 2015 announcement by Disney that ESPN subscriptions had fallen. Online and mobile video viewing continues to disrupt linear TV. Video viewing behavior is changing very quickly. Time spent in the U.S. on mobile consumption of video has tripled in the last 3 years. Facebook video traffic doubled in 2015 to 8b video views per day which is close to 100m hours of video being watched daily. There are 500m people watching video content on Facebook daily. Small businesses on Facebook uploaded 1.5m videos in December 2015.

I attended a speech by Linda Yakerino (Head of Ad Sales, NBC Universal) where she made bold statements about two sacred cows in TV advertising. First, she discredited Nielsen ratings systems because they have failed to properly measure the total audience across YouTube and mobile devices. Second, she admitted the quality of their content was continuing to decline and the prices they were charging to advertisers was increasing. I also note that in November 2015, NBC merged the linear TV and digital ad sales teams. NBC made $200m equity investments each in Vox and BuzzFeed in 2015 which highlights their need to migrate client ad dollars to audiences on digital platforms. Both Vox and BuzzFeed have significant paid presences on Facebook.

In summary, I am very challenged to quantify how quickly brands will allocate advertising dollars between the traditional broadcast / pay TV value chains and the online/mobile worlds. The inertia for agencies and brands to continue to advertise in TV is strong. However, Facebook Live, a new live broadcasting tool, has had early success and may provide Facebook a reason to secure digital rights to certain content on their own.

Print - Newspapers / Magazine – The secular decline in print is well documented. Newspapers are increasingly challenged to move subscribers to digital editions and only the largest newspapers (i.e. NYT, WSJ, and FT) have been successful in this regard.  Magazine publisher revenue of $28.3b fell by 3.7% last year, down 43% from $49.3b in 2007. Although magazine formats should be better at exploiting long tail interest groups, the critical mass requirements of printing and distribution costs make the medium very challenging. Brands are seeking Instagram to deliver compelling and memorable images to consumers.

Radio – The wide availability of digital storage devices and internet access to streaming services like Pandora have led to the decline in radio advertising. Radio budgets will continue to be a source of funds for other advertising channels.

Directories -  Like Google Adwords, directories have inherent value since the user is seeking them with clear intent to make a purchase decision.

Valuation

My estimates are inclusive of Facebook and Instagram but don’t take into consideration monetization strategies for Messenger, WhatsApp, Oculus Rift or other adjacent advertising opportunities. Key revenue drivers in my model are monthly average users (MAU) and average revenue per user (ARPU). In this case, ARPU is an output measure of factors that drive advertiser spend.

Global Facebook MAU of 1.8b grew 17% y/y in 2016. (+5%, U.S./Canada; 8% Europe; +25% APAC; +19% ROW).  Brazil, India, Indonesia, and the U.S. have been core markets with high adoption for many years although the company no longer provides data for these countries. To start, I reference global census benchmarks that predict global population will increase from 7.4b in 2015 to 8.2b in 2025. I apply census benchmarks to the population growth in each reportable segment (range is 0.8% - 1.0%) and internet user growth (range is 1% to 6%) to model Facebook penetration rates as follows over the next 10 years. These rates could be meaningfully different according to internet penetration rates which are sensitive to technological advances and economic factors specific to each region.

ARPU is a function of multiple variables; 1) ad bid competition determined by the number of businesses that seek to advertise; 2) the targeting ability of Facebook software to deliver high ROI ads; 3) the amount of time that users spend on Facebook; 4) type of content that advertisers choose with video typically being the highest value. Each of these factors has been trending positively for the social network which accounts for the trailing 4-year ARPU CAGR of 32%. This trend has continued with ARPU growing 26% in 2017 YTD. For 2017-2026, I model a potentially conservative 6.2% CAGR for ARPU.

Higher ARPU for U.S. & Canada is due to a variety of factors. 1) Per capita ad spend in the U.S. is the one of the highest globally at $567; 2) Users in the U.S. spend an average of 50 minutes per day on Facebook (vs. 20 minutes for other markets). The more time in the app, the more ads that will be viewed; 3) Data privacy laws in Europe limit some targeting strategies.

Estimating future operating margins is challenging in view of the potential leverage in the business model. An impressive reported operating margin of 40% in 2016 coupled with 460bp of operating margin expansion in 1H 2017 suggest “mature” margins of 55% are probable in a few years and seem like a good starting point for further sensitivity analysis. The cost of revenue line is largely operating data centers and storing information with 950bp improvements in gross margin from 2011 to 2016. I model mature gross margins increasing from 86% in 2016 to 90% in 2019. Moore’s law provides a natural tailwind to data oriented businesses that don’t have to pay for content.

Full year operating expense guidance came in at the low end in 2016. This appears to be repeating again in 2017 with the range narrowed from +40-50% down to +40-45% as of Q2 2017. As the reported operating expense growth was only 36% in 1H 2017 implying even the low end of guidance will be easy to achieve for the full year. R&D headcount is about 34% of total headcount. Engineer hiring practices adhere to strict meritocracy and require excellent cultural fit such that accelerating hiring practices to keep pace with revenue growth is difficult. I estimate R&D headcount increased 34% compared to 54% revenue growth in 2016. Reducing R&D to 16% of revenue in 2025 still implies R&D dollars that are 4.5X larger than those spent today. Were this increased R&D budgeted allocated to software engineer headcount with an average salary of $252k (2X the Glassdoor.com reported average), this would still imply an additional 68,000 engineers in 2025, an outcome which is unlikely. However, I still assume heavy investment in software or related technologies to include Rift and other internet access technologies (akin to the Other Bets segment at Google). Sales & marketing is modeled in line with its historic run-rate at 13% on the premise that many businesses need to be instructed about the benefits of advertising. I model G&A leverage from 6.5% today to 6% by 2025.

In 2016 Facebook revenue of $28b is 4.4% market share of $622b in global ad spend. I estimate revenue of $156b in 2026 or 17% market share of $904b global ad spend. This assumes global ad spend has grown 3.8%/year in line with GDP growth in the prior 5 decades. However, digital ad spend could continue to grow at even higher rates (>15% forecast growth rate today) depending on future purchasing behaviors that fundamentally transform retailer P&Ls by substituting retail lease expense (i.e. footfall) with digital ad expense (i.e. “clickfall”).

Base case metrics are shown below.  

Base case (70%) – I estimate 2018 revenue of $55b, operating margin of 52.5%, FCF of $18.3b based on declining capital expenditures given reduced data storage costs. With no additional acquisitions, a 10-year DCF (10% WAAC) suggests fair value of $225ps. Today, FB trades at 24X/17X/13X  EV/EBIT for 2017/ 2018/2019, respectively. These are very attractive multiples given the quality of the business.

I forecast 2017 revenue $1b above Street consensus in 2017 based on muted ad load impacts in 2H 2017. My estimate of 2019 revenue ($70.7b) and EBIT ($38.3) are 12% and 26%, respectively, above Street consensus. Given the size of the market cap and the visibility into the Facebook business model and advertising trends, I think this constitutes a meaningful variant view.

I assume MAU growth rates of 17%, 14% and 12% for 2017, 2018, and 2019 respectively. I assume ARPU growth rates of 25%, 18% and 14%, for 2017, 2018, and 2019 globally. ARPU growth rates are especially reliant on U.S./Canada growing ARPU at 37% in 2017 as this represents 50% of revenue mix by geography. As ARPU is truly an output measure of the value of a user to advertisers, this growth trajectory assumes that Facebook increases its market share of global ad spend from 4.4% in 2016 to 10.1% in 2019. This base case assumes EBIT margins of 52% in 2017 expanding to 55% by 2021. This translates to a 23% IRR at a 22X EV/EBIT exit multiple in 2019.

Bull case (15%) – Same assumptions as base case however with 10% higher ARPU growth rates for each geographic segment in 2018 and 2019 relative to the base case. For example, the growth rate in the U.S. / Canada is at 25% in the base case and would be increased to 35% in 2018. A 10% higher ARPU growth rate is also estimated for 2019. In this scenario, ad targeting would be delivering even higher ROAS than forecast. This implies 11.8% market share for global ad spend in 2019. A 10-year DCF (10% WAAC) suggests fair value of $260ps and translates to a 30% IRR at a 22X EV/EBIT exit multiple in 2019.

Bear case (15%) - Same assumptions as base case however with 10% lower ARPU growth rates for each geographic segment in 2018 and 2019 relative to the base case. For example, the growth rate in the U.S. / Canada is at 25% in the base case and would be decrease to 15%. This implies 8.6% market share for global ad spend in 2019. In this scenario, ad targeting would be delivering lower ROAS than forecast. A 10-year DCF (10% WAAC) suggests fair value of $192ps and translates to a 8% IRR at a 18X EV/EBIT exit multiple in 2019. The bear case for FB does not imply much impairment risk.

Risks

Tax rate increase -  The Double Irish phases out in 2020 and Facebook will likely need to find a new tax domicile to achieve 20% tax rates. I model 26% tax rates in out years but this may be too low.

User reaction to changes in Facebook products, policies and ad serving have been loud -  The same network that makes Facebook powerful can also turn users against it. Beacon was one such product that displayed purchase history on an automatic opt-in basis. It was too invasive, people were angry and Facebook was slow to disable it and admit the mistake. A Board member suggested that Zuckerberg hire a veteran COO as a sounding board which led to hiring Sheryl Sandberg who built Google’s self-service ad business. Zuckerberg eventually led an extended dialog with users to develop the “Facebook Principles” and “Statement of Rights and Responsibilities” as a democratic process to define the community. Zuckerberg having voting control and 20% ownership reduces this risk.

Slower development of programmatic advertising and reporting tools may delay wide scale adoption for brand marketers-  Although ad tech is developing quickly, it is extremely complicated and it is not clear if marketers will implement solutions that can consistently give comfort to clients regarding digital efficacy. Agency clients and brands need to commit resources to digital ad creative which is fundamentally different from TV ad creative. Agencies are rarely accountable for efficacy and ROI so the existence of superior advertising channels does not necessarily lead to adoption.

Global internet user growth may be constrained by governments– Of late, both India and Egypt have pushed back against the Facebook Free Basics program which seeks to provide free or low cost internet in developing countries where internet adoption lags. Governments are concerned Facebook won’t meet net neutrality standards or they fear citizen uprising organized via Facebook. India only has 350-400m internet users or about 33% of the population. However, efforts are in place by private firms to offer inexpensive smartphones and 2G connections that may bring adoption to 500m by 2017. Internet access is a clear gating factor for social networking growth.

Facebook penetration in teenagers is declining given increasing social networking options -  eMarketer expects the number of teen (12-17 yo) Facebook users in the U.S. and U.K. has peaked. This is largely due to migration to SnapChat and Instagram. The impacts to Facebook could be largely offset if Instagram’s network effects are strong and if monetization levels are similar to Facebook.

Zuckerberg is an unproven capital allocator- The business generates cash today and will generate much more in the future. It is difficult to know how the cash will be redeployed and/or returned to shareholders. Zuckerberg is also 33 years old, didn’t finish college, never worked at another company and has not experienced normal business setbacks as would be typical for a CEO of a firm this large. However, there is some evidence the risk of bad capital allocation is diminished. First, Zuckerberg was very impressed with Don Graham’s focus on maintaining control over the business so he could force long term strategies and investments. He tried really hard to get Don on the board of the company in its formative days. I know that several of the free call options (Messenger, WhatsApp, publisher network, etc.) have been receiving substantial internal investments primarily to maximize value to the user. These projects do not appear to be without merit but with an opportunity cost that is hard to quantify. The company is now publicly saying they want to make phone numbers useless and drive a multitude of transactions services through Messenger (e.g. Uber, airline tickets and customer service). There are several more risky projects such as artificial intelligence, VR (Oculus Rift) and aircraft projects to deliver internet to remote parts of Africa. From the standpoint of acquisitions, I note they have had ample resources for many years and have done only a few deals since 2014. Instagram was obviously very good. Rift seems like a low cost way to gain a leadership position in a totally new market that could be very innovative as VR and AR use cases increase. The WhatsApp deal was a mostly stock deal and dilutive of 7% which appears very expensive. However, WhatsApp has more than doubled its users to 1b since the acquisition and there are examples of other messaging services like WeChat that have been very successful in driving new interactions between consumers and businesses in China.

Catalyst

I believe that management’s persistent commentary about headwinds from News Feed ad load limits reaching a peak in Q2 2017 has been more about managing the stock price and will be a non-event for Q3 2017 with Facebook delivering a 2H 2017 revenue growth rate of 45% vs. Street consensus forecast of 38%.