Valuation of today’s internet firms : about Unicorns, non-financial value indicators and bubbles

(1)

MASTER THESIS UNIVERSITY OF AMSTERDAM (UVA) AMSTERDAM BUSINESS SCHOOL

Valuation of today’s

Internet firms

about Unicorns, non-financial value

indicators and bubbles

S.A. Tiemstra LL.M. Student number 10906541

8/30/2016

Supervisor: Dr. J.K. Martin

Abstract

This research finds that the value of users of Internet firms has changed significantly as from Q1 2011. This significant change is negative; the value of a user is declining over time. Furthermore, the research finds strong evidence that at least from Q1 2014 onwards - in line with current scientific literature- the amount of users impacts the market capitalization of Internet companies. The above impacts the equity value of publicly listed Internet companies. Therefore, the current valuation of privately held Internet companies may show signs of a bubble, although additional research is required to prove such.

(2)

2

1. Introduction

1.1 Introduction

Many (high) tech Internet startups currently receive astonishing valuation amounts. As per today, there are dozens of technology based startup companies who pass the one billion-dollar valuation threshold. In a TechCrunch blogpost (2013), Cowboy Ventures founder Aileen Lee introduced the term “Unicorn” to label hot technology based startups that pass the one billion-dollar valuation threshold. Although it is only 1% of all technology startups that will eventually reach and exceed that one billion-dollar threshold, the total group of companies falling within the category of Unicorns has grown significantly over the years.

Erdogan et al. (2016) and a study by CB Insights (2016) showed that as per the end of 2015, 146 private tech companies were valued more than double the number of 2014. Furthermore, there were 14 private companies classified as “Decacorns”, meaning their valuation exceeds the $10 billion threshold. These sky-high valuations are an interesting phenomenon. Many Internet companies that were listed in the past three to four years have shown poor performance. Lee and Ravichandran (2015) showed that the value of more than 40 percent of these Internet companies that went public since 2011 are flat or below their final private-market valuations. Parallels with the sky-high valuations for Internet companies in the 00’s come to life. Some research has been done in the field of the Internet bubble and the valuation methods used for public equity offering of Internet companies. It somehow seems that for both today’s Internet firms and the 00’s Internet firms, non-financial key business metrics such as subscribers, users, or customer basis (or a combination thereof) have value for investors. As the users and their activities within the environment created by the Internet firm are the “new currency” for investors.

Equity valuation of Internet firms stock shows similarities with any other (high-tech) new business start-ups. In absence of business track records, significant up-front capital investments for development of a critical mass of customers and for the architecture of their product is required, large expenditures on Sales & Marketing and on product development (R&D) results. Many of these early phase companies lack positive cash flow, profitability and sometimes even revenues. Traditional equity valuation methods such as DCF and P/E ratios will not work given these circumstances. This seems to have forced investors and analysts to focus on other non-financial measures of performance. This leads to a larger focus on the cash generating ability of intangible assets of such a firm.

(5)

5

Petterson et al. (2015), interviewed Venture Capitalist (“VC”) funds for a research report on Unicorns. They concluded that there are indications of a bubble in the valuation of Internet companies. They believe that investors overpay for equity in Internet companies and as a result their market caps are inflated. This bubble may be fueled by the large amounts of available cash within many corporations at the moment and an extreme low central bank and governmental backed low-interest environment.

It seems that these valuation concerns recently have reached the markets. The disastrous outcome of the Facebook IPO on May 18, 2012 is still fresh in the memories of investment banks. The market capitalization of Facebook peaked at an estimated $104bn and was therefore seen as one of the biggest IPO´s in the Internet industry history. Shortly before the stocks were listed, the introduction price per share was increased from $28-$35 per share to $35-$38 per share by the the underwriting banks, Morgan Stanley, JP Morgan, and Goldman Sachs. However directly after the stock were listed, the stock prices fell by more than 50%. As a result of the disastrous Facebook IPO, recent price per share on the IPO of Square, a Silicon Valley based payments company, was only $9 per share, $2 below the bottom range of the introduction price that was believed likely by the investment banks that took the company public. Shortly after introduction, the price bounced back by more than 50%. This suggest that either the bar for the introduction price was set deliberately low or the underwriting investments banks are not capable to get a good price for their customers´ IPO (Waters, 2015). Especially in the late-stage investments, asset managers mark down their stakes significantly. Slightly earlier in that same month and similar to the not so successful Square IPO, the $15bn stake that Fidelity Investments has in Snapchat was written down in by 25%. (Macmillian and Grind, 2015). In addition, Fidelity wrote down its stake in Dropbox, an online cloud storage company, by 31%. BlackRock did a similar write down of its stake in June 2015, where it wrote down 24%. (Waters, 2015). As per the first quarterly report of 2016, Baltimore based investment firm T. Rowe Price marked down most of its investments it holds in tech companies such as Uber and Dropbox. (Winkler, 2016).

The aim of this thesis is to research the development of this non-financial key business performance metrics over the years, measure the impact it has on market capitalization and try to give my initial thoughts on whether or not an equity pricing bubble exists in today’s Internet firms.

(6)

6

This thesis is structured in the following order.

Chapter 3 provides my theoretical framework. Definitions of Internet and Unicorn firms are explained and the characteristics of Internet firms are described. Furthermore, a summary description of equity valuation techniques and more specifically for early phase companies is provided and the focus of investors on non-financial key business performance metrics is further researched. Lastly, a short note on equity bubbles is made. The chapter is summarized in a summary paragraph.

Chapter 4 provides my practical research. The chapter starts with my hypothesis and the applied research methods. The chapter is continued with a description of the data selection process, by description of the availability and selection process of the data and contains statistical results and conclusions with regard to my research questions.

Chapter 5 concludes by connecting the practical findings of chapter 4 with the theoretical findings of chapter 3 and discusses shortcomings of the research and recommendations for future research.

The thesis is then finalized by a summary conclusion comprehending an answer to the research questions and a summary of the research results.

1.2 Why is this topic interesting for science?

The existence of “Unicorn” and “Decacorn” Internet firms is a recent phenomenon. Although there is academic proof that non-financial key business metrics such as the amount of users, members or subscribers have an influence on Internet firms stock value, not much research has been done or further scientific literature exists which researches the development of such non-financial key business metrics over the years and researches the impact of it on the market capitalization of today’s Internet firm’s.1 In contrast with today’s Internet firms valuation, more research has been done on the 00’s Internet bubble. There seems to be many parallels between the sky-high valuations of the 00’s Internet companies and today’s Internet companies.

For both today’s Internet firms and the 00’s Internet firms, non-financial key business metrics such as subscribers, users, or customer basis (or a combination thereof) have value for

1

It seems that this has been noted by MIT Sloan School of Management too, who recently started a research project on this topic; http://ide.mit.edu/research-projects/user-investment-and-firm-value-case-internet-firms.

(7)

7

investors as it is seen as a proxy for future revenue and cash flow. As the users and their activities within the environment created by the Internet company are the “new currency” for investors, it is interesting to learn how the value of such non-financial key business performance metrics has evolved over the years, and whether it showed a significant change. Furthermore, it would be interesting to research the impact of the non-financial key business performance metrics, such as user, on the market capitalization of an Internet firm, as research currently exists which proves that these non-financial key business performance metrics have an impact on stock prices.

Lastly, the findings may be used to draw an initial conclusion on whether or not today’s Internet firms show signs of overvaluation.

1.3 Research questions

This thesis researches the development of the value of non-financial key business performance metrics, such as users, and tries to find proof for the relationship between the amount of users and market capitalization. These findings will be used to give my initial thoughts on the existence of any equity pricing bubble in stock of today’s Internet companies.

The research questions and hypothesis which I would like to answer in this thesis are;

i. Did the value of users change significantly over time?

 Null-hypothesis (H0) is that there is no significant change in value per user over the years over the entire dataset;

 Alternative hypothesis (HA), is that there is a significant change in the value of users. This can either be;

 Positive: the value of a user is increasing over time.  Negative: the value of a user is decreasing over time

ii. Does the amount of users impact the market capitalization of Internet companies?

 Null-hypothesis (H0) is that the number of users does not have influence on the market value of the company;

(8)

8

 Alternative hypothesis (HA), is that the number of users has a significant influence on the market value of the company. This can either be;

 Positive: the higher the amount of users, the higher the market value of the company.

 Negative: the higher the amount of users, the lower the market value of the company.

1.4 Applied research methods

This research follows the characteristics of an empirical study. After a literature research as done in chapter 3, chapter 4 tries to statistically prove the hypothesis.

The first research question researches a single variable; the value of users. As the data is metric, the one sample t-test to compare the means should be performed in order to test whether the change is significant. The one sample t-test examines whether upon repeat of the test in the same population, we can expect that 95% of the intervals contains the mean. In the second research question, two scaled variables are compared; the number of users, and, market value. Both the correlation coefficient (using Pearson’s test) and (multiple) linear regression could be performed to test the relationship between the variables.

(9)

9

2. Theoretical framework

2.1 Definitions of Internet and Unicorn firms

According to Hand (2001) a firm is an Internet firm if at least 51 % of its total revenue originate from or are derived from internet (activities), and, that such firm cannot exist without internet itself. Internet firms are also known as dot-com companies, derived from the top-level domain “.com” and are companies that have most of their business activities on the Internet (Investopedia.com). The OECD (2012) does not provide a single definition of an Internet firm, but identifies several types of businesses that could be identified as Internet firms. These are companies that increasingly rely on the Internet for most of its functionalities, the providers of the Internet’s basic infrastructure and platforms by enabling communication and transactions between third parties, and, companies that provide traditional offline services migrated to an online platform. Unicorn firms, such as Facebook, LinkedIn, Uber, Palantir, etc. distribute new technologies and solutions (IT software and hardware) via the Internet and as such are dependent on the existence of the internet in line with the scope of the definition of an Internet firm.

The definition “Unicorn”, has been recently introduced on a TechCrunch blogpost, by Cowboy Ventures founder Aileen Lee (2013), to label the extreme high valuations that these type of Internet firms receive these days. Lee gathered and researched a publicly available data set from U.S. based tech startup companies starting from January 2003 and concluded that only 0.07% of all of these companies surpass the one billion-dollar valuation threshold. These companies were then labeled “Unicorns” as chances that an investor finds companies that will reach that valuation threshold are so rare, it resembles finding a Unicorn.

2.2 Characteristics of Internet firms; a comparison between the 00’s

and today

As today’s Internet firms and the 00’s Internet firms share many parallels, there are differences too. These differences have been subject of study by Erdogan et al. (2016), Cogman and Lau (2016) and the Economist magazine (2015). They found the following main differences between today’s Internet companies and the 00’s Internet companies:

(i) valuations exceeding the $1bn threshold;

(ii) concentration of investment within a small group; (iii) staying private (much) longer;

(10)

10

(iv) value creation takes place in pre-IPO period, and;

(v) business model aims for absolute market domination in their business category.

(i) valuations exceeding the $1bn threshold

00’s Internet companies backed by private investors, were all valued below the “Unicorn” threshold of $1bn. These days many privately held Internet startups are valued above this $1bn threshold, therefore making it a Unicorn. It seems that in the current market, private investors value pre IPO Unicorns higher than pre IPO Internet companies. 14 of the Unicorns classify as “Decacorns”, meaning their valuation exceeds the $10bn threshold. In addition, the pace at which these companies reach a Unicorn or Decacorn valuation is increasing. Figure 1 shows the increasing pace in which these companies reach their Unicorn or Decacorn valuation threshold. Furthermore, the amount of capital flowing into the private market is growing. In a two year period from 2013 to 2015, the amount of private capital invested in private companies almost tripled, from $26bn to $75bn, for all types of Series funding.

(11)

11 (ii) concentration of investment within a small group

The source of private funding is concentrated and is kept concentrated to exclusive and well connected groups of private investors who have privileged access to the startup market in Sillicon Valley. This is a huge difference as compared to Internet firms in the 00’s which aimed to go public as soon as possible. One of the side effects is that many startup companies are now “shielded” from public criticisms, therefore increasing chances that weak ideas survive for a prolonged time period.

Valuation of today’s non-listed Internet firms is hard as long as the shares are held in private hands. In general after completing a series of fund raising, the next series should imply a higher value of the companies’ stock. However, investors frequently demand special privileges to be attached to the shares they acquire. Examples of these are liquidation preferences (a guarantee that upon liquidation of the company their class of shares is first to receive any liquidation proceeds) and ratchets (additional shares that are allotted to the investor in case the IPO value of their investment is lower). As a result, these limitations should be taken into account, and add an additional layer of complexity, when assessing the value of this non-listed stock.

(iii) staying private (much) longer

Research showed furthermore that the time span in which today’s Internet companies are going public is changing. Ritter (2015) has shown that the average age of US technology companies that went public in 1999 was four years; this is in contrast with the findings for 2014 where the average age of the companies that went public changed to 11 years.

(iv) value creation takes place in pre-IPO period

Furthermore, Ritter (2015) found that the companies that went public in the period 2004 to 2015, only 6 out of 35 reached a pre IPO valuation above $10bn and the remained reached that valuation on average after eight years after IPO. It appears that these days privately held companies are able to raise capital in the private market at a larger scale than ever before. In a study by Brown and Wiles (2015) they introduce the term “Private IPO’s” or “PIPO” to define the deferring of an IPO event by allowing firms to raise capital while remaining private. The PIPO deals seem to provide the companies with the benefits of large-scale public funding, but without the cost attached to a public governance structure that listing on such public markets brings with it. The increased magnitude of such PIPO transactions has

(12)

12

several implications. One of these can be found in the time at which most of the companies’ value creation takes place. The “older” type of Internet companies like Amazon, Microsoft and Google, have shown a significant amount of value being created after they went public. For more recent IPO’s, like Twitter, most or even all of the value creation took place before its IPO. This is mainly due to the business approach of private equity investors: superior governance structures are imposed which reduce agency cost and bring a strict focus on free cash flow. Additionally, private equity investors aim to align ownership and incentive structure which push managers to make optimal decisions for the shareholders of the company.

In the PIPO transactions, today’s Internet firms seem to have access to a larger pool of private funding as compared to Internet companies in the 2000’s. Brown and Wiles (2015) suggested some driving forces behind the supply for PIPO funding:

(i) Low interest rates and the search for yield in the market;

(ii) Poor performing smaller company IPOs and PE governance, and;

(iii) Mainstream acceptance of PE investments and focus on Corporate Venture investments by corporations.

They furthermore suggested the following driving forces behind the demand for PIPO funding:

(i) Heavy, burdensome and costly compliance regulations for public companies introduced in amongst others SOX Act in 2002;

(ii) Limited analyst coverage and IPO activity for smaller sized listed companies, and; (iii) High costs for IPOs and risk.

(v) business model aims for absolute market domination in their business category

Another difference is that the business model of today’s Internet firms differ from that of the 00’s Internet companies. As I will point out in section 3.3.4, the business model of today’s Internet firms is aimed at gaining absolute domination in their business category (the “Winner-takes-it-all model”, or “Category King”). Such a dominating position should be captured by squeezing out any competitors against all cost. This behavior “produces a

positive feedback loop. More funding leads to a higher valuation, which generates more interest from the press, which makes it easier to attract and retain employees, which makes it possible to outperform rivals, which brings in more funding” (The Economist, 2015). The

(13)

13

research by Abeslon and Narasin (2015) showed that the larger the firm, the higher the trading multiple at time of their IPO and afterwards, and the better the stock performance.

Lastly, the current global dimension is different than in 2000. The recent dawn of China as a booming world economic power fueled innovation and growth, and as per today more than 3.2bn people are using Internet as compared to an approximate 400m as per 2000, with the Internet offering far more technical possibilities than ever before.

Lastly, the valuation of recently listed Internet firms in 2015 averaged around 20 times earnings, which is around 10% above the general market. This is a huge difference compared to the 00’s period where listed Internet companies were valued against price to earnings multiples around a factor 170.

2.3 About stock valuation and Initial Public Offerings (IPO’s)

2.3.1. Valuation methodology

Although the trend for today’s Internet firms is to remain private as long as possible, private equity investors have a limited investment horizon and their strategy is aimed to exit their investment eventually. IPO’s therefore remain one source of exit for private equity investors and are a good proxy for the market value of the equity of such company.

An IPO reflects the first time that stock of a company is sold on the public market. From the company’s perspective, it provides better access to capital and greater liquidity. From the investor(s) perspective it adds diversification to their investment portfolio and enables them to cash out, but also results in the equity holders being wider dispersed and control being less concentrated. The underwriting investment bank(s) work(s) together with the company to get to a price range which they believe is a reasonable valuation of the company’s stock (Berk and DeMarzo, 2014). It is important to realize that the initial offering price range (should) reflect the market’s perception of the future value and not the future itself. In theory, given the efficient markets theory as introduced by Fama, markets will always convert to equilibrium and therefore any mispricing of underwriting investment banks should be cancelled out (Fama, 1970).

Research has shown that firm characteristics, aggregate stock market returns and stock market volatility, influence the valuation methods used in an IPO process (Roosenboom, 2007).

(14)

14

Damodaran (2010), describes four main influencing factors that should be assessed In order to price a business:

1. the further amount of cash flow generated by investments of the company; 2. the value added by future growth;

3. the level of risk and cost of funding attached to the expected cash flow, and; 4. the terminal value of the company.

In most cases, the discounted free cash flow model including application of a discount rate (DFCF), the dividend discount model (DDM), or a multiple/comparables approach is used to value the stock. Although the valuation process for all types of businesses all follow one of these models or a combination hereof, the most important challenge is to assess which estimates should be used in the valuation model. This is depending on the growth phase a company is in.

As mentioned in section 3.2, many of today’s Internet firms share the characteristics of Start-up and Young Growth companies (“Early Phase companies”), as depicted in the below figure 2.

(15)

15

The figure shows that the amount of available information grows as the company ages. For Early Phase companies there is limited information on revenues, the operational history is limited, there are no or few comparable firms (especially of the businesses in new industries) and the source of value stems (predominantly) from future growth. It is therefore that especially for Early Phase companies, other measures of value such as rules of thumb or alternative non-financial information is used to fill the information gaps required to assess its value.

In their research, Bartov and Mohanram (2001) have performed a comparison between the valuation methods that were used for non-Internet and Internet companies at their initial public offering stage. Their main conclusion is that (i) the offering values of Internet companies and non-Internet companies differ significantly from each other; (ii) to reach to such a difference in offering value, underwriters use different value drivers to determine the price range and the final offering price at IPO; (iii) positive net income is positively valued for both non-Internet and Internet companies, but negative earnings do not have value except for internet companies; and (iv) they found that sales, sales growth and relative offering size are important value determinants for Internet firms.

IPO’s of European Internet companies in the period 1998-2000 was researched by, Knauff, Roosenboom and Van Der Goot (2003). Methods that were used to value these companies differ from the valuation methods used for traditional companies’ IPO’s. In contrast to valuation methods that are used for traditional companies’ IPO’s, the nature of the internet industry combined with lack of profitability and tangible assets pushed investors to look for other value drivers. These value drivers are for instance the number of visitors per website and the number of monthly subscriptions.

2.3.2. The focus on non-financial information for estimating the value of Early Phase companies

Hand (2001) measured the importance of three groups of factors in the pricing of U.S. Internet stocks: economic fundamentals, web traffic and supply and demand forces. He found and proved that (traditional) economic fundamentals such as forecasted one-year ahead earnings, current book equity and forecasted long-run growth in earnings explain stock prices of Internet companies. In another paper (Hand, 2003) he finds that basic accounting data are highly value-relevant in a non-linear manner. He finds that losses reflect immediate expensing of large investments into marketing and research and development intangible assets and not, poor operational performance. Core, Guay and Van Buskirk

(16)

16

(2001), focused on equity valuation of new economy stocks, using an industry wide sample set of data, not only including Internet firms. In line with Hand (2003), they find that traditional financial variables are still able to explain firm value during the period of their research (1995-2000), but that such valuations are subject to a greater variation to be explained by uncorrelated omitted factors.

To find such omitted factors, we need to search for other value drivers. The influence of non-financial information on stock prices has been previously researched for other industries. Amir and Lev (1996) examined financial and non-financial information for independent cellular telephone companies. Chandra, Procassini and Waymire (1997) researched the influence of announcements of book-to-bill ratios within the semi-conductor industry. Ittner and Larker (1997) researched the relation between customer satisfaction measures and both accounting numbers and market values. Deng, Lev and Narin (1999) researched usefulness of patent citations for predicting future market-to-book ratios and stock returns for high-tech firms.

Hand (2001) found that although both traditional industries and the Internet industry share the fact that economic fundamentals dominate their equities pricing, only stock prices of Internet companies are impacted by non-traditional value-drives such as web traffic -and more specifically the number of unique visitors to a firms’ website. He concludes that there is a reliable relation between the market value of an Internet firms’ equity and the amount of web traffic and supply and demand. As a result he concludes that this may be evidential for the fact that the pricing of Internet companies stock is less than fully rational. He furthermore found that public float, short interest and institutional ownership are also related with an Internet firms’ equity value. Rajgopal, Kotha and Venkatachalam (2000) underpin the conclusions found by Hand (2001); web traffic is an important non-financial indicator of the market values of Business to Consumer internet firms. They furthermore found that, in line with Trueman et al. (2000 and 2001), web traffic predicts level of sales up to two quarters ahead but the market does not value traffic merely because it predicts future sales though. Trueman, Wong and Zhang (2001), found that web usage data generally have significant incremental value for predicting the revenues of Internet companies. They furthermore found that gross profits are positively and significantly associated with stock prices of Internet companies. More importantly, they concluded that internet usage measures such as unique visitors and page views are found in most instances to provide incremental or even considerable explanatory power for the valuation of Internet companies’ stock, depending on the type of Internet business the company is active in. In addition to these findings, Demers and Lev (2000), found that three factors capture the most relevant dimension of website

(17)

17

performance, and, are value-relevant for the share prices of Internet companies: “reach” (ability to attract unique visitors to a website), “stickiness” (average time spent on the website) and “customer loyalty” (average number of visits per unique visitor). Equity valuation based on customers has been researched by Gupta, Lehmann and Stuart (2004). They underpin the above findings and show that valuation based on customers can indeed be a strong proxy for firm value. Their research furthermore suggests that customers are assets and that therefore customer-related expenditures should be treated as investments rather than expenses.

It is a fact that non-financial information plays a role to bridge information gaps for equity valuation for Early Phase companies. In line with this theory, the following two recent acquisitions can be mentioned which would have received other valuations if these companies were in another phase of their lifecycle and investors purely focused on traditional valuation methods:

 In February 2014, Facebook acquired messaging service WhatsApp for $19bn in cash and stock. In 2013, Whatsapp revenues were only around $10.2m, its net operating loss amounted to $138.1m and it had some 450m users. The valuation multiple of around 19 is based on projected sales of 1bn future users which would have to pay an annual yearly fee of around $1. This multiple is similar to those used for companies which developed life saving drugs and one of the highest ever paid in the market. Facebook recently announced that users of Whatsapp will not be charged a fee, but that its strategy is aimed to grow the pool of users and to capture them within the Whatsapp environment by way of adding payment services, online shopping services and providing games for instance. As such, Whatsapp aims to receive fees from the service providers each time a user is referred to their business and in this way tries to monetize its user base. It appears that the user base is more valuable to Facebook than financial figures such as revenues.

 On June 13, 2016, Microsoft announced that it reached agreement with LinkedIn to acquire 100% of its share capital for $196 per share, totaling to an all cash acquisition sum of $26.2 billion. Currently, LinkedIn has around 443 million users, which means that the value per user is $59.14. Microsoft did not announce which reasons are behind this recent acquisition. Gomes-Casseres (2016) points out that the acquisition could be initiated by a combination of a strategic remix and a value creation strategy. By such a strategy Microsoft acquires and further develops promised -yet risky- technologies or businesses, and sees how these develop over time. Mims

(18)

18

(2016) points out that the synergy between these companies could be found in the fact that both Microsoft and LinkedIn have an alike user base: (business) professionals. Again users seem to have value.

To conclude; there is strong academic evidence -and practical examples- that non-financial key business metrics such as amounts of users, customers, subscribers, etc. have an influence on the value of Internet companies.

2.3.3. About non-financial key business metrics

Since there is academic evidence that investors take into account non-financial key business metrics such as amounts of users, customers, subscribers, etc., it is important to look at the definition of these metrics. Surprisingly, there is no existing universal definition of what a “user”, “customer”, or “subscriber” is. There seems to be no formal accounting standard requirement (GAAP, IFRS or any other international recognized accounting standard) on how to report with regard to these non-financial business metrics. Furthermore, given the fact that these metrics are non-financial data, there are no accounting guidelines or regulations on how, and if, this information should be disclosed and there are no guidelines or regulations directing how consistency of such information is safeguarded. Given the earlier mentioned importance of non-financial key business metrics for investors, it is surprising that neither regulators nor investors press for clear regulation.

Although there is a lack of regulation, most of the Internet and Unicorn firms disclose information with regard to their non-financial key business performance metrics. Depending on the underlying business model of an Internet firm, the type of reported business performance metric changes. There are huge differences in consistency and quality of the provided non-financial key business performance metrics between companies. Social networks such as LinkedIn, Facebook, RenRen, Weibo most often use the term Monthly Active Users, or “MAU”. What such a definition comprises depends on the own discretion of the company. For instance Facebook defines MAU as:

“(…) a registered Facebook user who logged in and visited Facebook through our website or a mobile device, or took an action to share content or activity with his or her Facebook friends or connections via a third-party website that is integrated with Facebook, in the last 30 days as of the date of measurement (…)”.

(19)

19

Other social network companies may not have exact similar metrics, but all measurement techniques aim to capture the magnitude of usage of its products or services. Other Internet and Unicorn firms that are active as content providers or online marketplaces such as Netflix, Angie’s list and HubSpot report on a subscriber base, but try to capture similar characteristics as social media companies do: they all aim to capture the magnitude of usage of its products or services.

2.3.4. The winner takes it all – time to market cap

As non-financial key business performance metrics such as users, customers and subscribers are a very important value indicator, it is crucial that these customers keep consuming the companies’ services. Revenue generating capacity is dependent on either directly or indirectly the amount of users that a company has. This impacts the valuation of the company, and in connection therewith, also access to the capital markets since investors value those companies with a large customer base more than those with a small customer base. Any competition which might take a share of that user base is therefore a direct threat to the existence of the company.

Research by Peterson et al. (2015), from San Fransisco based Venture Capital and Advisory fund PlayBigger that focuses on improving today’s Internet startup firms with their business model to become a leader in their specific category they act in, has shown that the so called “time to market cap” for startups has increased significantly over the past years since the 00´s Internet IPO boom. The “time to market cap” is defined as the annual growth rate of a company’s market cap and therefore:

“provides a strategic lens into the velocity of new company, technology and category development. Time to market cap has emerged as the leading indicator that a company has the potential to become a Category King”.

Peterson et al. (2015), Play Bigger Research Uncovers 11 Super-Unicorns, para. 2 and 3. Retrieved from www.playbigger.com.

As shown earlier, startups that are capable of serving customers’ needs that they did not know they had by creating complete new categories of services and products can easily attract and retain a large and active user base as long as they are able to develop an ecosystem in which these users spend their time. Good examples are for instance Uber, Facebook and Linkedin. As these users become ‘captured’ in this ecosystem, these companies are able to dominate the market in that specific category of product or service

(20)

20

and will quickly become a category king. Category Kings are able to capture the majority of market share of their service or product offering. Peterson et al. (2015) and Harvard Business Review (2016) showed that 76% of a specific market niche is captured by those Category Kings.

2.4 Equity bubbles

Classical finance theories dictates that asset prices are determined by fundamentals only (expected future cash flows and risk premium for bearing risk) as introduced in the Capital Allocation Pricing Model (CAPM) by Markowitz and Sharpe (1964), the no-arbitrage pricing theory by Black, Scholes, Merton and Ross (1973) and the Efficient Markets hypothesis by Fama (1970). Under these assumptions economic agents are fully rational, markets are efficient (assets are always correctly priced) and managers make decisions in the best interest of shareholders. The more recently developed Behavioral finance theories, amongst others Kahneman and Tversky (1976), acknowledge that there are deviations from these efficient market theory assumptions: people can be irrational, assets can be mispriced (and bubbles can exist) and managers can make decisions which are contrary to shareholder interests.

Under standard believes and the standard concept of human behavior, investors beliefs are all rational and thus the same. As investors have differences in beliefs some may be more optimistic, others rational and some pessimistic which generates trade. In line with amongst others Morris (1996), Miller (1997), Scheinkman and Xiong (2003) and Hong, Scheinkman and Xion (2006), especially in markets where short sale options are limited and optimists form a large group, a price bubble may be easily formed. Bubbles are harmful since they lead to misallocation of investment, excess volatility and lead to a non-optimal distribution of welfare (Kyle, Laibson, Mollerstrom, 2011).

2.5 Summary

In the theoretical framework (section 3.1), the definitions for Internet and Unicorn firms have been defined. In section 3.2, I described the characteristics of the Internet firms in the 00’s and today’s firms. I concluded that today’s Internet firms share many similarities with the 00’s Internet firms, as they are both depending on the existence of the Internet itself. The main differences between 00’s Internet firms and today’s Internet firms are:

(21)

21

 Today’s internet firms are valued much higher than Internet firms before they went IPO, hence the term Unicorn firms was introduced to label these firms;

 Today’s internet firms are held in a privileged group of private investors for a longer time before they are publicly listed;

 Most of the value creation of today’s internet firms takes place in these private hands. After IPO of today’s Internet firms, research has shown that the valuation of listed firms show a multiple of around 20 times earnings, which is around 10% above the general market.

For valuation purposes, it seems that non-financial key business metrics impacts stock prices of Internet companies and thus that these non-financial key business performance metrics can be a strong value proxy. Given the fact that research on the development of this value indicator is limited, but that many of today’s Internet firms have reached Unicorn status – a private valuation exceeding USD 1bn- the relationship between non-financial key business metrics and firm value is an interesting topic for further investigation. This thesis researches the development of the value of such non-financial key business performance metrics, and tries to find a relationship between the amount of users and market capitalization. It furthermore tries to discover whether there might be signs of overvaluation for privately held Internet firms by using data from publicly available Internet firms as a proxy.

(22)

22

3. Practical research

3.1 Hypothesis

This research focuses further on the development of the value of non-financial key business performance metrics (users, subscribers or customers) over the years and tries to find a relation between these non-financial key business performance metrics and the market capitalization for today´s Internet companies.

Based on the previous chapter, I would like to answer the following research questions and hypothesis;

i. Did the value of users change significantly over time?

ii. Does the amount of users impact the market capitalization of Internet companies?

 Null-hypothesis (H0) is that the number of users does not have influence on the market value of the company;

 Alternative hypothesis (HA), is that the number of users has a significant influence on the market value of the company. This can either be;

 Positive: the higher the amount of users, the higher the market value of the company.

 Negative: the higher the amount of users, the lower the market value of the company.

(23)

23

3.2 Research methods

This research follows the characteristics of an empirical study. After a literature research as done in chapter 3, chapter 4 tries to statistically prove the hypothesis.

First research question

As we compare a single variable (namely the value of users) and our data is metric, the one sample t-test to compare the means should be performed to test whether the change is significant. The one sample t-test examines whether upon repeat of the test in the same population, we can expect that 95% of the intervals contains the mean. In order to test such we accept the null-hypothesis in case the chance of such situation occurs in more than 5% of the cases and reject our null-hypothesis in case the change is less than 5%. This means that we have to perform a one sample t-test on the means for the Dataset, using that respective mean as test value. Although the T-test is widely used, it is limited in its ability to proof what factors causes the significant change in value of users over the years.

Second research question

The second research question will be answered by using (multiple) linear regression. As we are comparing two scaled variables, namely the number of users and market value, both the correlation coefficient (using Pearson’s test) and a linear regression could be performed. The correlation coefficient indicates the level of linear correlation and direction of the relation between the variables. The correlation coefficient is always between -1 and +1, where -1 means that there is a perfect linear negative correlation (more of X, less of Y), 0 means that there is no linear relation and +1 means that there is a perfect positive linear relation (more of X, more of Y).

The linear regression analysis tries to predict outcomes, based on the linear relationship of the two variables. In general the regression equation could be written as:

(24)

24

Where:

 Y is the dependent variable (which is to be predicted);

 X is the independent variable (which is the predicting variable)  a is the regression-coefficient of X on Y

 b is the constant factor

In this thesis, Y reflects the market capitalization, X the number of users. Therefore the equation could be written as:

The regression-coefficient provides the increase in the dependent variable in case the independent variable is increased with 1. Furthermore, the regression-coefficient depends on the scale of measurement of the independent and dependent variable.

A multiple regression analysis could be used to include other factors that may have impact on the market capitalization. The factor time is another independent variable which can be used. To find the relationship between time, market capitalization and the number of users, we have to solve for:

The independent variables in this equation are the market capitalization and the date. The dependent variable is the market capitalization.

The (multiple) linear regression analysis that will be used to analyze the impact of the amount of users on the market capitalization of a company is limited by the omitted variables bias: there could be other influencing factors that impact the market capitalization which are not included in the analysis.

3.3 About the data, its availability and the selection process

To assess the hypotheses, datasets pertaining data about the amount of non-financial key performance metrics and market capitalization/ value information is required.

Market capitalization = a * Number of users + b

(25)

25

Obtaining reliable data from privately owned companies for non-financial key performance metrics and value is hard. Furthermore, privately held Internet firms do not have a market valuation and valuations on funding rounds leads to biased results in valuation. Therefore, this research only focuses on publicly listed Internet companies. As the U.S. market (NASDAQ, NYSE) are the leading markets for Internet stocks, the research will purely focus on companies that are listed on either of these markets.

For data selection criteria, the following selection criteria were used:

 Company that is part of the Technology sector and classified as subdivision Internet Content & Information Industry of the NASDAQ and NYSE;

 A current market capitalization exceeding $10 million;

 The Company discloses information on its non-financial key business performance metrics;

 Disclosure of such non-financial key business performance metrics is done for at least 4 consecutive quarters or if not disclosed on a quarterly basis, for at least 4 years.

The sector classification has been drawn from www.ycharts.com. Appendix A shows a list of companies that are part of the Technology sector. I have only selected the Content & Information industry, as the business models of companies active within this industry seems to be mostly driven by non-financial key business performance metrics such as users and subscribers. For information about market capitalization - amounts of outstanding shares and share prices - my sources of information were www.finance.yahoo.com and

www.ycharts.com.

Most, but not all listed Internet firms, disclose (parts) of their non-financial key business performance metrics. The financial key performance metrics are often disclosed as non-audited supplementary parts of the obligatory publicly accessible quarterly and annual company SEC-filings (10-K and 10-Q declarations). In some cases the data could only be collected on corporate websites - more specifically on the investor relations web pages- and on specific investor aimed documentation such as additional non-GAAP compliant and unaudited performance metric reports, press releases, earnings releases and transcripts of investor relations conference calls. For completeness sake, it should be noted that there is no formal accounting standard requirement (GAAP, IFRS or any other international recognized accounting standard) to report on these non-financial business metrics. Given the fact that these metrics are non-financial data, no formal accounting guidelines or regulation exist on disclosure and consistency of such data, as also discussed in section 3.3.3.

(26)

26

After obtaining the non-financial key business metrics, the amount of users is then divided by the market capitalization of the Internet or Unicorn companies, as per the end of the quarter. The market capitalization is calculated by the product of the total number of shares outstanding and the closing price. Given application of the efficient markets theorem, I assume markets incorporate expectations about the development of the user base in the share price directly. The result of the division is the value per user as per the end of the quarter.

3.4 Data - statistics

Many companies in the Content & Information industry are relatively new on the stock markets and only have data covering a short time frame. Additionally, a large part of these companies started disclosing this information only recently (2010 and onwards) or on an irregular basis. As a result, I have decided to compile the dataset based on quarterly data and limited the selection time horizon to 5 years. As many as possible companies which disclose their non-financial key business performance metrics were then selected.

Based on above mentioned selection criteria I have found 42 companies which disclose their non-financial key business performance metrics. Data was collected for a 5 year period starting in Q1 2011 up to and including Q1 2016. This resulted in a total of 574 data entries.

The dataset comprises the market capitalization (amount of outstanding shares times share price), the amount of users (based on information in the non-financial key performance metrics) and as output the value per user, as a division of the market capitalization and the users. The data is collected on a quarterly basis (if available) and otherwise yearly data is used.

Missing data issues

One of the main challenges within the dataset is the number of missing values. As elaborated earlier, the lack of regulation on disclosure of non-financial key business performance metrics results in inconsistencies with regard to the approach on disclosure and

Amount of users Value per user =

(27)

27

the timing thereof. Additionally, the dataset contains companies that have been founded recently and consequently prior data simply does not exist. This results in the data to be very inconsistent and without any further handling not to be useful for statistical analysis. In order to tackle this problem, the following steps were taken:

1. Excluding outliers and/or irregular reported values

In general, the value of users is roughly between USD 10 and USD 1,000. Any other values could be regarded as outliers as they have a relative large impact on the statistics. In addition irregular reported (only yearly instead of quarterly) values imply too many missing data points, these data points were excluded too. This approach results in the excluding of the in the Figure 3 showed companies.

Name Outlier Missing Value per user in USD

HubSpot (HUBS) x 91,149 NewRelic (NEWR) x 134,114 58.com (WUBA) x 7,772 YYinc (YY) x 24 ChinaCache (CCIH), x x 107,730 Limelight Network (LLNW) x 217,884 Phoenix New Media (FENG) x x 2

Figure 3 - Excluded data from data set and the reason of exclusion

2. Split the approach for further analysis based on information availability

After excluding outliers and/or irregular reported values, the dataset shows two main features; (1) almost all companies consistently disclose their non-financial key business performance metrics starting from Q1 2014, and (2) most of those who disclose prior to Q1 2014, disclose for (almost) the entire time period of measurement. As a result of this finding, further analysis of the dataset could be split in an analysis of the companies that disclose from Q1 2014 onwards “Dataset 1” and companies that disclose for the entire data series “Dataset 2”.

(28)

28 Dataset 1 (from Q1 2014)

The above mentioned approach results in a dataset comprising of 35 companies and 9 data points per company, totaling to 315 data entries to be available for analysis. There are only a few missing data points left.

Dataset 2 (from Q1 2011)

Only those companies who have missed a maximum of 5 data points, which is less than 20% of the total amount of data input per company, are included. This approach results in 19 companies and 21 data points per company, totaling to 399 data entries to be available for analysis.

This approach results in the datasets as depicted in Appendix B. It is aimed to find a trend (if any) within this dataset and then compare these two trends.

After split of the datasets, the datasets were uploaded in SPSS for further statistical analysis.

A missing value analysis was done. The two datasets now show little missing values as shown in depicted Figures 4 and 5; only 3.9% for Dataset 1, and 8.3% for Dataset 2 is missing.

(29)

29

Figure 5 - Missing data analysis for Dataset 2.

Without deepening too far into statistical theories, there is discussion on which percentage of missing data is acceptable for running a statistical analysis which is not biased by the missing data problem. In casa a dataset has missing values over 10%, Bennett (2001) stated that statistical analysis is biased. In this thesis, my assumption is that a maximum of 10% of missing data is therefore acceptable and consequently Dataset 1 and Dataset 2 are acceptable for further analysis.

In addition, the missing data mechanisms and the missing data patterns have an impact too. Therefore, in addition the randomness of missing data was analyzed by using the Little’s Missing Completely At Random test in SPPS. This test is useful to execute prior to imputing missing values, as it analyzes whether or not data is missing just randomly or whether there is any pattern in the missing data. Based on this analysis the data the data shows random signs of missing points.

As both Datasets pass this test too, we will now accept these datasets for further analysis without implying missing values.

For each company and their input value in the dataset (market capitalization, number of users and value of users) the level of normality of the distribution has been tested. As a rule of thumb the standardized skweness and kurtosis should be between -3 and +3. For Dataset 1 all input data except 3 cases out of the total of 102 variables are within these boundaries. For Dataset 2, 10 cases out of the total of 57 variables are within these boundaries. Although

(30)

30

it can be concluded that some of the data is not normally distributed, these findings are ignored for this thesis but should be further investigated in any additional research.

After these steps, a statistical analysis was done to calculate amongst others the means and standard deviations for each variable: market capitalization, the number of users and the value per user of both Datasets per period. This data will be used for answering my research questions.

3.4 Results and description of the results

3.4.1 Did the value of users change significantly over time?

Both Dataset 1 and Dataset 2 have been analyzed using descriptive statistics as shown in Appendix C. As the assumption for this further thesis is that the data is normally distributed, the mean is used for further analysis (and not the median). The Mean and Standard deviation of the number of users has been plotted against time.

Descriptive statistics for Dataset 1 (from Q1 2014)

The data show an almost consistently declining mean in the value per user from 253 in Q1 2014 to 150 in Q1 2016. In addition, the standard deviation of the value per user shows an almost consistent decline from 373 in Q1 2014, to 173 in Q1 2016. The declining standard deviation means that most of the companies in the Dataset 1 have a value per user that is converging to the mean over time. Figure 6 shows the development of the value per users in USD and standard deviation against time. A dotted line has been plotted to indicate the trends.

(31)

31

Figure 6 – Development of the value of users for Dataset 1.

The data show an overall declining mean in the value per user from 252 in Q1 2011 to 149 in Q1 2016, but this decline is not consistent. In addition, the standard deviation of the value per user has declined overall from 189 in Q1 2011, to 178 in Q1 2016, but shows an irregular pattern of dispersion showing a maximum standard deviation of 382 in Q1 2014. Figure 7 shows the development of the value per users in USD and standard deviation against time. A dotted line has been plotted to indicate the trends.

Q1 2014 Q2 2014 Q3 2014 Q4 2014 Q1 2015 Q2 2015 Q3 2015 Q4 2015 Q1 2016 Mean 253 223 213 191 203 204 181 181 150 Standard Deviation 373 290 271 249 249 256 204 204 173 125 175 225 275 325 375 Val u e p e r u ser in USD

Development of the value of users - starting Q1

2014

(32)

32

Figure 7 – Development of the value of users for Dataset 2.

Q1 2011 Q2 2011 Q3 2011 Q4 2011 Q1 2012 Q2 2012 Q3 2012 Q4 2012 Q1 2013 Q2 2013 Q3 2013 Q4 2013 Q1 2014 Q2 2014 Q3 2014 Q4 2014 Q1 2015 Q2 2015 Q3 2015 Q4 2015 Q1 2016 Mean 252 224 136 192 203 161 150 156 181 216 250 250 252 227 201 182 197 208 175 188 149 Standard Deviation 189 200 113 229 217 167 165 194 228 260 267 310 382 299 269 238 251 275 206 216 178 100 150 200 250 300 350 400 Val u e p e r u ser in USD

(33)

33

Significance analysis of Dataset 1 (from Q1 2014)

The data (Figure 8) show a significance of .000 (sig 2-tailed). This means roughly spoken that the chance that our null-hypothesis is correct equals .000%. As the chance is less than 5%, we can reject the null-hypothesis.

Figure 8 – Dataset 1 one-sample t-test results.

Significance analysis of Dataset 2 (from Q1 2011)

Dataset 2 (Figure 9) shows a significance of .001 (sig 2-tailed). This means roughly spoken that the chance that our null-hypothesis is correct equals .001%. As the chance is less than 5%, we can reject the null-hypothesis.

(34)

34

Conclusions for Dataset 1 and Dataset 2

Recall that my first research question was “Did the value of users change significantly

over time?” and that my hypothesis was:

As we can reject the null-hypothesis for both Dataset 1 and Dataset 2, we can accept the alternative hypothesis; there is a significant change in the value of users. The change is negative; the value of a user is declining over time. This is because the values we were assessing for (253,2 for Dataset 2 and 251.5 for Dataset 1) are not within the lower and upper boundaries of the 95% confidence intervals for the means as depicted in Figure 10 and Figure 11 below. Therefore, the null-hypothesis can be rejected for both Datasets. There is a significant change in the value of users over time. The change is negative; the value of a user is declining over time.

(35)

35

Figure 11 - 95% confidence intervals for the means of Dataset 1.

3.4.2 Does the amount of users impact the market capitalization of

Internet companies?

Both Dataset 1 and Dataset 2 have been analyzed using descriptive statistics as shown in appendix D and E. As the assumption for this further thesis is that the data is normally distributed, the mean is used for further analysis (and not the median).

The data show a consistently increasing mean in the number of users from 159 in Q1 2014 to 231 in Q1 2016. In addition, the standard deviation of the number of users shows a consistent increase from 333 in Q1 2014, to 417 in Q1 2016. The increasing standard deviation means that the number of users of most of the companies in Dataset 1 is wider dispersed over time as compared to the mean. Figure 12 shows the development of the number of users and standard deviation against time. A dotted line has been plotted to indicate the trends.

(36)

36

Figure 12 – Development of the number of users for Dataset 1.

With regard to the market capitalization, Dataset 1 shows an overall increasing mean in the market capitalization from 17,675 in Q1 2014 to 30,746 in Q1 2016. In addition, the market capitalization shows a consistent increase from 37,794 in Q1 2014, to 71,796 in Q1 2016. The increasing standard deviation means that the market capitalization of most of the companies in Dataset 1 is wider dispersed over time as compared to the mean. Figure 13 shows the development of the market capitalization and standard deviation against time. A dotted line has been plotted to indicate the trends.

Q1 2014 Q2 2014 Q3 2014 Q4 2014 Q1 2015 Q2 2015 Q3 2015 Q4 2015 Q1 2016 Mean 159 179 190 190 196 204 212 215 231 Standard Deviation 333 333 337 339 355 371 386 394 417 125 175 225 275 325 375 425 N u m b e r o f u ser s * 1 m ln

Development of the number of users - starting

Q1 2014

(37)

37

Figure 13 – Development of the market capitalization for Dataset 1.

The data show an overall declining mean in the number of users from 239 in Q1 2011 to 194 in Q1 2016. In addition, the standard deviation of the number of users shows an overall decrease from 567 in Q1 2011, to 397 in Q1 2016. The decreasing standard deviation means that the number of users of most of the companies in Dataset 2 is converging to the mean over time. Figure 14 shows the development of the number of users and standard deviation against time. A dotted line has been plotted to indicate the trends.

Q1 2014 Q2 2014 Q3 2014 Q4 2014 Q1 2015 Q2 2015 Q3 2015 Q4 2015 Q1 2016 Mean 17,675 20,019 28,009 28,068 28,142 28,891 25,030 30,343 30,746 Standard Deviation 37,794 40,416 56,949 61,644 59,456 61,004 55,190 67,096 71,796 17,500 27,500 37,500 47,500 57,500 67,500 M ar ke t cap itali zation * US D 1 m ln

(38)

38

Figure 14 – Development of the number of users for Dataset 2.

With regard to the market capitalization, Dataset 2 shows an overall increasing mean in the market capitalization from 11,098 in Q1 2011 to 35,725 in Q1 2016. In addition, the market capitalization shows an overall increase from 15,015 in Q1 2011, to 84,994 in Q1 2016. The increasing standard deviation means that the market capitalization of most of the companies in Dataset 2 is wider dispersed over time as compared to the mean. Figure 15 shows the development of the market capitalization and standard deviation against time. A dotted line has been plotted to indicate the trends.

Q1 2011 Q2 2011 Q3 2011 Q4 2011 Q1 2012 Q2 2012 Q3 2012 Q4 2012 Q1 2013 Q2 2013 Q3 2013 Q4 2013 Q1 2014 Q2 2014 Q3 2014 Q4 2014 Q1 2015 Q2 2015 Q3 2015 Q4 2015 Q1 2016 Mean 239 203 187 162 134 176 180 182 186 177 173 177 177 177 170 178 179 176 179 181 194 Standard Deviation 567 476 441 374 320 352 353 355 358 352 345 347 350 349 336 353 356 358 368 377 397 125 175 225 275 325 375 425 475 525 575 N u m b e r o f u se rs * 1 m ln

Valuation of today’s internet firms : about Unicorns, non-financial value indicators and bubbles