<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Rodrigo Marino on Medium]]></title>
        <description><![CDATA[Stories by Rodrigo Marino on Medium]]></description>
        <link>https://medium.com/@rodrigo.marino?source=rss-2b3565e4767------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*XBmd8HMeOhnRigBkE_6p4w@2x.jpeg</url>
            <title>Stories by Rodrigo Marino on Medium</title>
            <link>https://medium.com/@rodrigo.marino?source=rss-2b3565e4767------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 15:04:08 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@rodrigo.marino/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The impact of COVID 19 on the Venture Capital sector]]></title>
            <link>https://medium.datadriveninvestor.com/the-impact-of-covid-19-on-the-venture-capital-sector-85098c1de054?source=rss-2b3565e4767------2</link>
            <guid isPermaLink="false">https://medium.com/p/85098c1de054</guid>
            <category><![CDATA[great-depression-covid]]></category>
            <category><![CDATA[covid-and-venture-capital]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[investments-and-covid]]></category>
            <dc:creator><![CDATA[Rodrigo Marino]]></dc:creator>
            <pubDate>Mon, 27 Jul 2020 09:31:53 GMT</pubDate>
            <atom:updated>2020-07-27T16:25:12.182Z</atom:updated>
            <content:encoded><![CDATA[<h4>Venture Capital market analysis from the Great Recession to COVID19.</h4><p>When analyzing a startup, especially the ones at early-stage, investors focus on three main factors: team, market, and product. Times of crisis don’t make an extraordinary entrepreneur less capable, and a product does not lose its technical quality. Therefore, the main variable changing and consequently impacting the investments and valuations is the market.</p><p>The current scenario still presents a lot of uncertainty, markets that have always been large may decline drastically, while segments that until then were not relevant can gain strength. In this article, we will explore data from the Venture Capital market, in order to infer changes that are probable to happen.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*m304mFxzXUZgETOq" /><figcaption>Source: Unsplash <a href="https://unsplash.com/@frankbusch">@frankbusch</a> — A man in front of a screen with a red arrow</figcaption></figure><h3>1. Great Recession</h3><p>In this step, we will explore startups’ transactions¹ located in the US during the Great Recession period.</p><h4>1.1. Number of Transactions and the Total Amount Invested Between 2006 and 2011</h4><p>The 2008 crisis did not impact the number of Seed transactions, which grew significantly in the period. The other rounds decreased in 2008 and 2009, starting to regain growth only in 2010.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/888/1*Fcqs8hXnc_4YvoZ70QwZDQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/879/1*rTAdLXKcN-gMCIGwES0YYA.png" /><figcaption>Left: Number of transactions per year given the classification of the round. Right: Total amount invested, in USD billions, per year, given the rating of the round. (Startups based in the USA). Source: Crunchbase.</figcaption></figure><p>The goal of a startup is to go public or be acquired by a large company for a good valuation. In this way, the changes generated by a crisis impact the publicly-traded and decrease the chances, especially for late-stage startups, of a successful liquidity event. In this context, I believe that Seed investments become more attractive since the long investment horizon collaborates with the expectation of improvement in market conditions.</p><p><a href="https://www.datadriveninvestor.com/2020/05/17/how-to-truly-measure-a-venture-capitalists-performance/">How To Truly Measure A Venture Capitalist&#39;s Performance? | Data Driven Investor</a></p><p>When looking at the volume invested, we realize that the representativeness of the Seed rounds, despite having grown in the middle of a crisis, is still very small. Thus, the growth of Seed investments can be understood as an alternative for Venture Capital funds, to continue the search for opportunities, discarding the need for a capital call.</p><p><strong>All rounds starting from Series A suffered a strong reduction</strong>, <strong>mainly noticed in 2009</strong>, since the first half of 2008 was able to guarantee good investments for the year. It is also possible to note that even in 2011, the volume invested in certain rounds did not reach the level observed in 2007.</p><h4>1.2. Rounds Size Between 2006 and 2011</h4><p>The median size of the round fell in 2009, regardless of the classification of the transaction, in which Seed investments showed the biggest drop. For the early-stage and late-stage, the interquartile range has changed little, but the acute drop in the number of transactions has resulted in a decrease in the total amount invested in the period.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/867/1*ZtqOr4MomnZ-y1Rx_MWn3w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/867/1*XKTEiYyFp0KYphlSVQjrCw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/851/1*CLuLsmKGIo4UqnttdKpu3Q.png" /><figcaption>Boxplot⁷ — Annual distribution of the rounds size in USD given the classification. Early-Stage: Series A and Series B. Late-Stage: Series C + (Startups located in the USA). Source: Crunchbase.</figcaption></figure><p>We saw that, in general<strong>, the crisis reduced both the number of transactions and the size of the rounds.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/1*atQe1aDb94Se1eoQtiEE7w.png" /><figcaption>Representativeness of rounds of a certain size (range in USD) in relation to the total number of transactions in a given year. (Startups located in the USA). Source: Crunchbase.</figcaption></figure><p>In the graph below, transactions were segmented into some groups, based on the total amount invested. Investments of up to USD 2M gained importance after 2008, surpassing 50% in 2010 and 2011. This increase is the result of the bigger number of Seed transactions and size drop of Series A rounds.</p><p>The sample used does not include valuation data, so it’s impossible to identify the exact number of <a href="https://www.investopedia.com/terms/d/downround.asp">down rounds</a>. However, we managed to get a close idea by looking at the number of startups that raised new rounds with a smaller size than on previous rounds. An infrequent situation, which most often occurs when the startup is in need of cash and cannot find investors to support a bigger transaction. The difficulty in raising capital results in valuations less than or equal to the last round</p><p>This adverse scenario results in companies looking for <a href="https://microventures.com/startup-financing-understanding-bridge-rounds">bridge rounds</a>, when investors from the last round make a new contribution, typically at the same valuation, with the purpose of giving some oxygen to the startup’s cash flow and avoiding a <a href="https://www.sciencedirect.com/topics/social-sciences/write-off">write-off</a>. When it doesn’t happen, the startup is obliged to accept smaller investments at a lower evaluation. Furthermore, many don’t get to obtain new contributions, even with unfavorable terms, being then forced to cut their costs and, in many cases, leading to the death of the company.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/880/1*-Y-DzA3OQxgEaziR68j74g.png" /><figcaption>Y-axis: Shows the percentage of startups that raised smaller size rounds in comparison to the previous (SeriesA to SeriesE). Axis X: Indicates the relationship between the years of the last two transactions of the startup. Ex: The first bar indicates that, for startups that raised subsequent rounds in 2006 and 2007, 22% raised, in 2007, a smaller round in comparison to 2006 round. (Startups based in the USA). Source: Crunchbase.</figcaption></figure><p>Thus, a round size reduction, although not being conclusive, may indicate that the value of the startup was less than or equal to that of the last investment round.</p><p>The chart on the side indicates the percentage of rounds, in which the total amount collected was less than the last transaction. We can notice that the worst scenario occurred for startups that raised subsequent rounds in 2007 and 2009<strong>. In this group, 53% of startups raised, in 2009, a lower value than what was raised in 2007.</strong></p><h4>1.3. Valuations Between 2006 and 2011</h4><p>The value of a <em>startup</em> is a combination of objective and subjective factors, but the general dynamic is summarized as follows: the <em>startup</em> decides the amount of capital needed to achieve its short and medium-term plans, the <strong>valuation ends up being a consequence of round size and </strong><a href="https://blog.adioma.com/how-startup-valuation-works-infographic/"><strong>expected dilution</strong></a>. By accepting a certain amount, the investor believes that the company will be able to deliver enough progress to raise on a new round (or be acquired) at a higher valuation than the current one.</p><p>In times of crisis, liquidity events decrease and publicly traded companies are priced at lower levels. Thus, late-stage investors adjust the valuations in a manner consistent with what the financial market is paying for public listed companies. The other investors, seeking to preserve their return, also reduce the valuation they are willing to pay</p><p>In practice, the process of adjusting return expectations is slow and often only affects late-stage rounds. In this way, the <strong>Venture Capital industry becomes less susceptible to fluctuations in the capital market</strong>, but deep crises such as the Great Recession and COVID-19 tend to affect this market as well.</p><p>In adverse times, many LPs start to avoid illiquid and high-risk investments, while others see good opportunities in other assets, such as the stock market. In the crisis, it is difficult to see new Venture Capital funds, and the already capitalized ones invest more cautiously. This results in entrepreneurs with difficulty creating competition for the investment round, which is one of the main reasons for the increase in valuations.</p><p>As the valuation information is not available on Crunchbase used the estimates made by Pitch Book² to visualize what happened during the 2008 crisis. <strong>A drop is noticeable, especially in the early-stage and late-stage rounds, in which the full recovery occurred only in 2011</strong>. The valuation of Seed transactions, despite having decreased in 2009, was already fluctuating before the crisis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/657/1*7YNfS3GbvQMji04vLQFVKQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/588/1*mqFEV8KjvzDsOSYrqN22bw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/597/1*D4n3V4r0qJwy4iiuWNZc9g.png" /><figcaption>The figure above indicates the distribution of the Pre-Money Valuation in USD, given the year of the transaction, with the upper line representing the 75th percentile, the center line indicating the median, and the lower line the 25th percentile. The graph on the left is for Seed rounds, the center graph is for Series A and Series B rounds, and the right graph for Series C + rounds. (Startups located in the USA). <br> Source: Pitch Book Report.</figcaption></figure><p>Liquidity events are the main way for a Venture Capitalist to manage expectations. <strong>The number of IPOs fell from 89 in 2007 to 13 in 2008</strong>, and remained below the observed point before the crisis, even in 2011. The year 2009 suggests a strong recovery in the valuation of IPOs, however, with only 11 transactions, the median it’s not enough to infer an improvement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oFwTnQzlL5ZsVPqZbA78Ig.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ft2Rg0LNaz3DxT94s7MV_g.png" /><figcaption>Line graph with two Y-axes (left): Shows the annual evolution of the number of IPOs and the median of Post-Money Valuation (in USD) at the time of the IPO for companies located in USA which had received investments from Venture Capital. Line graph with two Y-axes (right): Shows the annual evolution of the number of acquisitions and the median of Post-Money Valuation (in USD) at the time of transaction, for companies located in USA which had received investments from Venture Capital. Source: Pitch Book Report.</figcaption></figure><p>The number of acquisitions fell slightly in 2008 and 2009, while the median of valuations fell by half, suggesting that capitalized companies took advantage of the crisis, to acquire for lower prices startups facing cash flow difficulties. These acquisitions usually focus on technology and the team (acqui-hire), with the goal of accelerating the digital transformation process of buyers. It is interesting to note that, unlike IPOs, acquisitions showed a strong recovery in 2010.</p><h3>2. The US Venture Capital Market After 2010</h3><p>In this step, we will explore ³ startups’ transactions located in the USA between 2011 and 2019.</p><h4>2.1. Number of Transactions and the Total Amount Invested Between 2011 and 2019</h4><p>After 2010, the asset investment class experienced a period of bull market. In the case of Venture Capital, this period was marked by growth, both in the number of transactions and in the total amount invested.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LH_bUuUWVVufQkr2b6jTYA.png" /><figcaption>Indicates the number of transactions in a given year. The colors indicate the total amount invested in USD. (Startups located in USA) Source: Crunchbase.</figcaption></figure><p>Late-stage rounds (Series C +) are the main cause of the increase in the total amount invested, since few transactions move a lot of capital. This phenomenon indicates that <strong>many startups are postponing a possible IPO, choosing to raise large rounds in the private market</strong>.</p><p>There are some reasons why a company remains in the private market: 1) High cost of an IPO. 2) Possibility to choose investors and keep the <a href="https://www.investopedia.com/terms/c/capitalization-table.asp">cap table</a> under control. 3) It is simpler to make major changes in the company. 4) The expectation of a more prosperous capital market in the coming years.</p><p>These reasons, combined with the capital availability led by late-stage Venture Capital funds, contributed to the increase in the volume invested in the period.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/887/1*B9yBbPKh9WwCNRhcB1VCFg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/874/1*zrwAu3lq5oNSZDhvQEWjRw.png" /><figcaption>Left: Number of transactions per year, given the classification of the round. Right: Total amount invested, in USD, per year, given the rating of the round. (Startups located in USA). Source: Crunchbase</figcaption></figure><p>The Seed rounds, despite being less representative in relation to the total capital invested, differed in the number of transactions. One of the reasons is that new Venture Capital managers typically start investing in Seed transactions, as these investments require less capital. Upon reaching satisfactory returns, new funds are launched with a focus on the other stages.</p><p><strong>Another way of looking at the boom in the number of Seed transactions is through the drop in the cost of creating a startup</strong>⁴. Horizontal computing and the popularization of open-source software reduced costs that went from USD 5M in 1999 to USD 500k in 2005. Cloud computing, led by AWS, enabled this cost to reach USD 50K in 2010. Test an idea effectively got easier and easier, encouraging the growth of Seed rounds.</p><p>The number of Series A and Series B transactions hasn’t kept pace with Seed’s growth, indicating that many startups have not advanced to the other stages of the investment cycle, reducing the expected return on a portfolio. This scenario made many early-stage investors choose to invest in a little more mature startups, even if they had to pay higher valuations, contributing to the reversal of the upward trend in the Seed rounds after 2015.</p><h4>2.2 Rounds Size Between 2011 and 2019</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AbDCyY39P-t6y0CQRSbe7g.png" /><figcaption>Representativeness of rounds of a certain size (range in USD) accordingly to the amount of transactions in a given year. (Startups located in USA). <br> Source: Crunchbase.</figcaption></figure><p>We noticed that after the Great Recession, rounds below USD 2M gained representativeness, a trend that continued up until 2013. The representativeness of rounds above USD 50M gained strength as of 2014, in line with the idea that <em>startups</em> are opting to stay in the private market for a longer period.</p><p>The figure below helps us to visualize the upward trend both in the median of the rounds and in the interquartile range. The increase in variability is the result of factors ranging from how capital intensive a given business model is, to macroeconomic motivations, such as low-interest rates<strong>. </strong>The increase in the rounds size is a strong indication of bigger valuations, since entrepreneurs would hardly accept aggressive dilutions in periods of heated market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/852/1*TiHelvWlhwl547UhTUvLCQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/856/1*21yGX3ygCDW8s48cAlj76w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/867/1*g6N2A-JI4vowscb5_FsPSQ.png" /><figcaption>Boxplot⁷ — Annual round size distribution, in USD, given the classification. Early-Stage: Series A and Series B. Late-Stage: Series C + (Startups located in USA). Source: Crunchbase.</figcaption></figure><h4>2.3. Valuations Between 2011 and 2019</h4><p>As startups become successful, a success that translates into new rounds, acquisitions, and <em>IPOs</em>, optimistic expectations regarding the market gain strength. This scenario contributes to the valuations growth at all stages, Pitch Book⁵ estimates help us to visualize this trend.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/653/1*YBtH-wyDTMVnafeuQ1jcsg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/593/1*obzM94uECsQ8MyszuAq9UQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/599/1*V5mJfxRmA-jhwNwN4JSskw.png" /><figcaption>The figure above indicates the distribution of Pre-Money Valuation, in USD, given the year of the transaction, with the upper line representing the 75th percentile, the center line indicating the median and the lower line the 25th percentile. The chart on the left represents the Seed rounds, the center chart the Series A and Series B rounds and the right chart the Series C + rounds. (Startups located in USA). Source: Pitch Book Report.</figcaption></figure><p>In recent years, the<em> </em>Venture Capital industry has received criticism for “insane valuations”, often suggesting a bubble. We saw that the valuations in fact managed to increase, but the rate was not much different from what was observed in the US capital market. The median Pre-Money valuation in <em>late</em>-<em>stage</em> rounds grew 2.5x between 2010 and 2019, while the Nasdaq index grew 3.7x in the same period. However, if we look at the average valuations, there was a growth of about 5.7 times, which indicates the existence of outliers in the Venture Capital industry.</p><p>Often, extreme cases such as WeWork are used as a <em>proxy</em> for the industry’s behavior, which generates a myopic view of the sector. Many criticisms point out that the growth in valuations is the result of an extremely heated market, since such an increase is not justified by metrics, such as revenue and profitability. This situation is related to the strong influence of expectations on the evaluation of startups.</p><p>Liquidity events are crucial to align expectations, after all, they are the ones that guarantee a return. In the graphs below, we can see an increase in the Post-Money valuation of IPOs and acquisitions, however, less consistent than the observed <strong>growth in the valuations of the <em>Venture</em> <em>Capital</em> rounds</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZNcuS3lczK2f7NwCzOm1Kg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lEY8LSQgy5nAFl4o3G7IZw.png" /><figcaption>Line graph with two Y-axes (left): Shows the annual evolution in number of IPOs and the median of Post-Money Valuation (in USD) at the time of the IPO, for companies located in USA which had received investments from Venture Capital. Line graph with two Y-axes (right): Shows the annual evolution in number of acquisitions and the median of Post-Money Valuation (in USD) at the time of the transaction, for companies located in USA which had received investments from Venture Capital. Source: Pitch Book Report.</figcaption></figure><p>It is difficult to conclude whether the industry is experiencing a bubble or not, but historical data shows that Venture Capital returns are not the best⁶. In 2017, the median returns for the last 10 years of the VC industry in the U.S. were 1.6 percentage points below the return generated by Nasdaq in the same period. This indicates that Venture Capital is not a good investment in terms of risk and return and, considering illiquidity, it gets even worse. So what justifies the global growth of the industry?</p><p>Although <em>Venture</em> <em>Capital</em>, on average, is not a good asset investment class, the best investors make extraordinary gains and motivate the entire sector. Thus, the feeling of a bubble, generated by high valuations may indicate that the market is in fact moving in this direction, or simply open the reality of an industry that, historically, does not generate satisfactory returns.</p><h3>3. The First Quarters of 2020</h3><p>Global investments in <em>Venture</em> <em>Capital</em> showed a slight drop in the first quarter of 2020, varying by region. While in China the reduction was strong, in the USA there was a slight increase.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/915/1*PEBdaxO0tD9rzTxzMXjBcw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/926/1*k2NB0oFCSrXFT6DJAzOvEg.png" /><figcaption>Left: Comparison of the total amount invested in Venture Capital between the first quarter of 2019 and 2020, given the investment stage. Direct: Comparison of the total amount invested in Venture Capital between the last quarter of 2019 and the first quarter of 2020, given the stage of the investment. Source: Crunchbase</figcaption></figure><p>The second quarter of 2020 positively contradicted expectations by showing a slight growth compared to the first quarter. Even so, there was a decrease in relation to the same period in 2019. This decrease was weaker than the observed in the quarters that followed the worsening of the 2008 crisis, suggesting that the current crisis will have a simple impact on investments in Venture Capital.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/920/1*MrcZ60CIv4kBAiX-rdlBzQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/921/1*HDAtKm9XQng1_-gT51IE_g.png" /><figcaption>Left: Comparison of the total amount invested in Venture Capital between the second quarter of 2019 and 2020, given the investment stage. Direct: Comparison of the total amount invested in Venture Capital between the first quarter of 2020 and the second quarter of 2020, given the stage of the investment. Source: Crunchbase</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/785/1*CgEO4mlKLAOF1116JECrqA.png" /><figcaption>Comparison of the total amount invested in Venture Capital and the number of transactions between 2019 and 2020. Source: Crunchbase</figcaption></figure><p>Regarding exits by M&amp;A, the first quarter of 2020 was not impacted by the crisis, marked by growth in the value of the movement for acquisitions. In the second quarter of 2020, it is possible to notice a negative impact, explained by the total trading volume and the number of records, both in the YoY and QoQ comparisons.</p><h4>3.3. Capital market in 2020</h4><p>The stock market reflects feelings and expectations in real-time, with which we can visualize the impact of the crisis on investments. Thus, we are able to see which segments will be harmed and which will stand out in the current scenario.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/792/1*s0xcF-H2PsRRnUKrlvAPQA.png" /><figcaption>Comparison of different capital market indices return, in the USA, between 01/02/2020 and 05/28/2020. Nasdaq (IXIC), S&amp;P Technology (XLK), S&amp;P Health Care (XLV), S&amp;P 500 (GSPC), S&amp;P Real Estate (XLRE), Dow Jones (DJI), S&amp;P Industrial (XLI). Source: Yahoo Finance.</figcaption></figure><p>The technology companies that make up the S&amp;P 500 performed above average, generating a return of 13.5% between 2020/01/02/ and 2020/07/06. In addition, Nasdaq’s return outperformed the S&amp;P 500 by 17.3 p.p. and Dow Jones by 24.1 p.p. in the analyzed period.</p><p>In times of crisis, most sectors are expected to retract, however, some are more resilient and end up performing better. <strong>In this case, many technology companies tend to benefit from the conditions imposed by social distance</strong>.</p><p>The chart below helps us to visualize the distribution of returns on certain assets, in 2020, based on the first day of the year. The first chart includes companies that received a Venture Capital contribution and made an IPO in 2019. The other two charts refer to the companies that make up the Emerging Cloud Index and Dow Jones.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/964/1*YDz9lQ6MbfzNmv6oH5jRBQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/976/1*Ia6ZdC9JGCv9xh9NQduntA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*mpUMg-qcBxr5cJxiTSAJZg.png" /><figcaption>The graphs show the distribution of returns in relation to the first day of 2020, for a given group of assets, with the upper line representing the 75th percentile, the center line indicating the median and the lower line the 25th percentile. The 2019 IPO group includes companies based in USA which had received input from the Venture Capital industry and held an IPO in 2019. The Emerging Cloud Index group includes the companies that make up the Emerging Cloud Index of Nasdaq and Dow Jones and the companies that make up the Dow Jones index. Source: Crunchbase, Yahoo Finance.</figcaption></figure><p>Cloud solutions have benefited from the forced digitization process, generated by social distance. The <a href="https://www.bvp.com/bvp-nasdaq-emerging-cloud-index">Emerging Cloud Index</a> includes companies that offer this type of solution, such as Zoom, Paypal, Shopify, and had one of the best performances since the beginning of the pandemic crisis. The companies in the third quartile showed a return greater than 80%, and even the first quartile showed a return above 15% in the period.</p><p>The companies in the 2019 IPO group are a good proxy for startups<em>,</em> since they represent the latest investments in the Venture Capital industry that have managed a public offering. But, as this group also includes companies from sectors that were severely affected by the crisis, a worse return is expected than the Emerging Cloud Index. In addition, companies in the 2019 IPO group are at greater risk than the technology giants, as they are less consolidated in the market.</p><p>The median return of companies in the 2019 IPO group was 13%, lower than the 42% of the Emerging Cloud Index, and also lower than the return of the bigger technology companies, such as Amazon and Google. However, it was significantly higher than the return of the Dow Jones, whose median was -13% in the period. Scenario reinforces the thesis that <strong>technology companies are more resilient to the current crisis</strong>.</p><h3>4. Final Comments</h3><p>I am publishing this article seven months after the confirmation of the first cases of COVID-19 in China. Since then, we have seen the scenario worsen around the world, contaminating more than twelve million people and causing more than five hundred thousand deaths. Today, 2020/8/7, the scenario is still very uncertain, we are not sure when the vaccine will be available, how many lives will be lost, and how long the economy will take to recover.</p><blockquote>I am prepared for the worst, but hope for the best. <em>(Benjamin Disraeli)</em></blockquote><p>In the face of so much uncertainty, the recommendations for <strong>startups to secure cash flow</strong> and extend the <a href="https://www.davemanuel.com/investor-dictionary/cash-runway/#:~:text=%22Cash%20runway%22%20refers%20to%20the,until%20the%20middle%20of%202018.">runway</a> are the most plausible. Although social distance imposed by the pandemic is more damaging to tractional businesses, most startups will suffer from reduced demand generated by unemployment and recession.</p><p>We saw that, during the 2008 crisis, the total amount invested was below the observed level before the crisis, even in 2009 and 2010. At that time, the terms were less favorable for entrepreneurs, evidenced by the increase in the number of down rounds and the decrease in the size and valuations of the investments. In the first crisis after the Great Recession, the Venture Capital industry will feel the <strong>impact of a long cycle prioritizing growth over profitability</strong>.</p><p>Until then, the 2020 data does not show a strong reduction in investments, even for late-stage startups. Likewise, the stock market is heading for a rapid recovery, in which technology companies are showing the best performance. If this increase is long-lasting, the Venture Capital market tends to benefit from liquidity, despite past crises suggesting that <strong>investments should decline in the coming months</strong>.</p><p>The pandemic exposed up inefficiencies, accelerated the digitization process, and transformed consumption habits. Some changes are temporary, while others are structural. In the post-COVID19 world, many business models will cease to exist, but there <strong>will be plenty of opportunities for a new wave of entrepreneurs willing to solve the problems that traditional players have failed</strong>. On the investor side, structural changes generate unique opportunities for returns.</p><h4>References and Methodology:</h4><p>[1][3] We used information from 57,814 transactions published in the USA-based <em>startups</em> <a href="https://www.crunchbase.com/home">Crunchbase</a> between 2006 and 2019. Transactions classified between <em>Seed and Series J</em> were included, whose total amount invested was disclosed.</p><p>[2][5] <a href="https://pitchbook.com/news/reports/q1-2020-pitchbook-nvca-venture-monitor">PitchBook Analyst Note: COVID-19’s Influence on the US VC Market</a></p><p>[4] <a href="https://www.slideshare.net/msuster/seed-stage-2019-final">What’s Happening to Seed stage Venture Capital? — UP Front Ventures</a></p><p>[6] <a href="https://a16z.com/book/secrets-of-sand-hill-road/">Secrets of Sand Hill Road, Chapter 2: So Really, What is Venture Capital? </a><a href="https://www.cambridgeassociates.com/wp-content/uploads/2018/05/WEB-2017-Q4-USVC-Benchmark-Book.pdf">US Venture Capital Index and Selected Benchmark Statistics</a>, <a href="https://www.globenewswire.com/news-release/2018/10/15/1621269/0/en/US-Venture-Capital-Achieved-Respectable-Double-Digit-Return-in-2017.html">US Venture Capital Achieved Respectable Double-Digit Return in 2017</a>, <a href="https://www.insper.edu.br/wp-content/uploads/2018/09/Performance-private-equity-venture-capital-industry-Brazil.pdf">PERFORMANCE OF THE PRIVATE EQUITY AND VENTURE CAPITAL INDUSTRY IN BRAZIL</a></p><p>[7] Outliers have been excluded from Boxplot graphics to ensure better viewing.</p><p>Other relevant materials on the subject:</p><ul><li><a href="https://medium.com/sequoia-capital/coronavirus-the-black-swan-of-2020-7c72bdeb9753">Coronavirus: The Black Swan of 2020</a></li><li><a href="https://news.crunchbase.com/news/massive-slowdown-in-2020-vc-funding-hasnt-happened-yet/">Massive Slowdown In 2020 VC Funding Hasn’t Happened … Yet</a></li></ul><p><strong>Gain Access to Expert View — </strong><a href="https://datadriveninvestor.com/ddi-intel"><strong>Subscribe to DDI Intel</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=85098c1de054" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/the-impact-of-covid-19-on-the-venture-capital-sector-85098c1de054">The impact of COVID 19 on the Venture Capital sector</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Data Science in Venture Capital]]></title>
            <link>https://medium.com/data-science/data-science-in-venture-capital-8c13ec0c8458?source=rss-2b3565e4767------2</link>
            <guid isPermaLink="false">https://medium.com/p/8c13ec0c8458</guid>
            <category><![CDATA[towards-data-science]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[data-driven]]></category>
            <category><![CDATA[co-investment]]></category>
            <category><![CDATA[network-analysis]]></category>
            <dc:creator><![CDATA[Rodrigo Marino]]></dc:creator>
            <pubDate>Wed, 04 Mar 2020 22:29:34 GMT</pubDate>
            <atom:updated>2020-06-14T14:12:11.926Z</atom:updated>
            <content:encoded><![CDATA[<h3>Network Charts and Venture Capital</h3><h4>How to use network charts to monitor co-investments among venture capital funds</h4><h4>Introduction</h4><p>Venture Capitalists strive to invest in incipient companies that have the potential to lead the market in the future. Despite investing in tech companies, the daily activities in a VC fund use to be too manual and based on feelings and impressions instead of data.</p><p>Of course, most part of the time, the data is not available and the decisions must be guided by subjective aspects. In this case, the experience and the fund’s network matter a lot. Sometimes, the data is available and VCs can use it to guide their decisions.</p><p>In this article, I will show how to visualize the connections among the venture capital funds based on their co-investments. To do so, we will resort to the network charts.</p><figure><img alt="Image source: NetworkX: Network Analysis with Python, Petko Georgiev -Computer Laboratory, University of Cambridge" src="https://cdn-images-1.medium.com/max/1024/1*wHNJVGV6MiPYr3OZdSYHUw.png" /><figcaption>Network Chart — Examples</figcaption></figure><h4>Why network charts?</h4><p>Our network chart will be useful to a VC fund that wants to understand their competitor’s interactions, to a Fund of Funds that intends to create a diversified portfolio and even to a startup that is planning its roadshow. The network chart is also a good tool to precede a detailed cluster analysis.</p><p>Here we will focus on the algorithm and the interpretation will be detailed on a second article. We will use Python and the library Networkx, which is very flexible and fits our needs. In our undirected network chart, each node represents a VC Fund and the edge thickness indicates the number of co-investments between two funds.</p><p>Edge colors:</p><ul><li>Dashed Black: 1 co-investment</li><li>Blue: 2 co-investments</li><li>Green: 3 co-investments</li><li>Red: 4 co-investments</li><li>Purple: 5 co-investments</li><li>Pink: 7 co-investments</li><li>Orange 16: co-investments</li></ul><p>You may be wondering why the number of co-investments jumps from 7 to 16 (max value). That’s a special case of two government-backed funds.</p><p>The figure below shows a portion of our network chart. We can see that a few funds have many co-investments and that there are three major groups based on the co-investment criteria.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/879/1*GjjEgIfNH7Ko-N9Bquyjvw.png" /><figcaption>Network chart — output</figcaption></figure><p>Before going ahead, let’s understand the source of our data!</p><h4>The data</h4><p>The data came only from Crunchbase. In real-life, we collect data from different sources, such as VCs funds website, press and public documents. That’s important because a relevant percentage of the investments are not on the Crunchbase’s platform.</p><p>In this article, I focused only on Brazillian funds and startups, but it can be easily replicated to any other country. l also didn’t consider the number of investments or portfolio companies per fund, but it could be well represented in the node size.</p><h4>Data cleaning</h4><p>Let’s code!</p><p>Now I will show how to organize the data according to the output from Crunchbase.</p><p>First of all, let’s import all the libraries.</p><pre>import matplotlib.pyplot as plt<br>import networkx as nx<br>import pandas as pd<br>import numpy as np<br>from itertools import combinations</pre><p>We need to read two datasets. The df_investors includes the name of all institutional venture capital funds headquartered in Brazil with at least one investment. The df_startups includes the name of Brazilian startups with a total funding amount higher than USD 300k, as well as the name of their investors.</p><pre># Investors dataset</pre><pre>df_investors=pd.read_csv(&quot;Investors.csv&quot;)<br>print(len(df_investors))</pre><pre># Startups dataset</pre><pre>df_startups=pd.read_csv(&quot;companies.csv&quot;)<br>print(len(df_startups))</pre><p>Let’s check how the datasets look like.</p><pre>df_investors.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/809/1*bfyIfU84BRYomrT7-HgAbg.png" /></figure><pre>df_startups.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/783/1*2Ft1T3b78YL9QJsHYsxAgw.png" /></figure><p>The column “Investors” in the df_startups dataset includes all the investors. We will filter to get only the Brazilian investors.</p><pre># each investors in one row <br># we will use .split() and .stack()</pre><pre>df_startups = (pd.DataFrame(df_startups.Investors.str.split(&#39;,&#39;).tolist(), index=df_startups.Organization)<br>               .stack().reset_index([0,&quot;Organization&quot;]))</pre><pre>df_startups.rename({0: &#39;Investors&#39;}, axis=1, inplace=True)<br>df_startups[&quot;Investors&quot;] = df_startups[&quot;Investors&quot;].str.lstrip()</pre><pre># merging the datasets to get only the brazillian investors</pre><pre>df=pd.merge(df_startups, df_investors, how=&quot;outer&quot;)</pre><pre>df=df[df[&quot;Location&quot;]==&quot;Brazil&quot;].dropna().drop([&quot;Location&quot;],axis=1).reset_index(drop=True)</pre><pre>df.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/767/1*3btPfJwJon7zTYPEwkMGLA.png" /></figure><p>There are more efficient ways to get the outcome above, but merging the datasets is very simple to implement when we have different sources of data.</p><p>Now we will exclude the startups with only one investor because they are not relevant to our analysis.</p><pre># we only want startups with more than one investor</pre><pre># df_aux with the number of investors per startup<br>df_aux=pd.DataFrame(df[&quot;Organization&quot;].value_counts().reset_index())<br>df_aux.rename({&quot;Organization&quot;: &#39;Number_of_investors&#39;,&quot;index&quot;:&quot;Organization&quot;}, axis=1, inplace=True)</pre><pre># merging the datasets <br>df=pd.merge(df, df_aux, how=&quot;outer&quot;)</pre><pre># filtering the dataset to get only the startups with more than one investors <br>df=(df[df[&quot;Number_of_investors&quot;]&gt;=2].sort_values(by=[&#39;Organization&#39;,&#39;Investors&#39;])).drop([&quot;Number_of_investors&quot;],axis=1)</pre><pre>df.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/936/1*GAXqTP2met6fNd85tX97pA.png" /></figure><p>Now we have 207 records that represent the number of investments that resulted in at least one co-investment. The real number larger than this, but due to the source of our data and the criteria adopted, that’s our number.</p><p>Before going ahead, let’s create a list of all the possible combinations of investors. It will be useful soon.</p><pre># unique investors name<br>comb=df[&quot;Investors&quot;].sort_values().unique()</pre><pre># list with all the combinations of investors<br>cc = pd.Series(list((combinations(comb,2))))</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/993/1*7_RIOvjPgD694EanRRut8A.png" /><figcaption>print(comb)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/993/1*g_t46O1yIa_SWXWi9NNqmw.png" /><figcaption>cc.head()</figcaption></figure><p>Now, we will create a dictionary in which each key will represent a venture capital fund and the values will be the startups backed by this fund.</p><pre># creating a dictionary <br># each key represents one investors</pre><pre>co_investments = { i : None for i in comb }</pre><pre>for investors in df[&quot;Investors&quot;]:<br>    co_investments[investors]=set()</pre><pre># each value represents the investments     <br>for startups,investors in zip(df[&quot;Organization&quot;],df[&quot;Investors&quot;]):<br>    co_investments[investors].add(startups</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/993/1*lqARedu65givlm8FGaqILQ.png" /></figure><p>Now, we will create a new data frame, which will be responsible to store the count of the co-investments among the venture capital funds. The “<em>Co-investments</em>” column is the 2-length combination of investors and the “<em># Co-investments</em>” shows the number of co-investments for each combination. It’s common to have more than two investors per startup, we are considering it.</p><p><em>Thanks to </em><a href="https://medium.com/u/5904750d29e0"><em>Daniel Gribel</em></a><em>, who helped me to correct and improve this step!</em></p><pre>#let&#39;s create a new df, which will be in charge to store the countage of the co-investments</pre><pre># new data frame <br>data=pd.DataFrame(data = {&#39;Co-investment&#39;: cc.apply(lambda x: str(x)[1:-1].replace(&quot;&#39;&quot;,&quot;&quot;))})<br>data=pd.concat([data, data[&quot;Co-investment&quot;].str.split(&#39;, &#39;, expand=True)], axis=1).rename({0:&#39;Fund1&#39;, 1:&#39;Fund2&#39;}, axis=1)</pre><pre>#let&#39;s count the intersection among the investments<br>count=[]<br>for fund1,fund2 in zip(data[&quot;Fund1&quot;],data[&quot;Fund2&quot;]):<br>    count.append(len(co_investments[fund1].intersection(co_investments[fund2])))<br> <br>data[&quot;# Co-investments&quot;]=count</pre><pre># filtering to get only the pairs with co-investments <br>data=data[data[&quot;# Co-investments&quot;]!=0]</pre><pre>data.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/921/1*rKC_pFR6xtRsbp0J_VcTPw.png" /></figure><p>The data frame above will work as input for the network chart.</p><h4>Plotting the chart</h4><p>The Networkx uses Dijkstra’s algorithm to find the shortest path in a weighted network. Let’s start by creating an undirected Graph.</p><pre>G = plt.figure(figsize=(30,15))<br>G = nx.Graph()</pre><p>Now we will read the data frame df_network_1 to add the weighted edges in the graph. We also will style our chart, adding different colors and thicknesses to the edges according to the number of co-investments.</p><pre># adding the edges with the information from df_network_1<br>    for i,j,w in zip(data[&quot;Fund1&quot;],data[&quot;Fund2&quot;],data[&quot;# Co-investments&quot;]):<br>        G.add_edge(i,j,weight=w)</pre><pre># weights represents the number of co-investments <br>    # the max values is 16 co-investments <br>    s1 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 1]<br>    s2 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 2]<br>    s3 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 3]<br>    s4 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 4]<br>    s5 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 5]<br>    s6 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 7]<br>    s7 = [(u, v) for (u, v, d) in G.edges(data=True) if d[&#39;weight&#39;] == 16]</pre><pre>pos = nx.spring_layout(G)  # positions for all nodes</pre><pre># nodes<br>    nx.draw_networkx_nodes(G, pos, node_size=50, node_color=&quot;gray&quot;)</pre><pre># edges<br>    nx.draw_networkx_edges(G, pos, edgelist=s2,<br>                           width=2, alpha=1, edge_color=&#39;blue&#39;)</pre><pre>nx.draw_networkx_edges(G, pos, edgelist=s1,<br>                           width=1, alpha=1, edge_color=&#39;black&#39;,style=&quot;dashed&quot;)</pre><pre>nx.draw_networkx_edges(G, pos, edgelist=s3,<br>                           width=3, alpha=1, edge_color=&#39;green&#39;)</pre><pre>nx.draw_networkx_edges(G, pos, edgelist=s4,<br>                           width=4, alpha=1, edge_color=&#39;red&#39;)</pre><pre>nx.draw_networkx_edges(G, pos, edgelist=s5,<br>                           width=5, alpha=1, edge_color=&#39;purple&#39;)<br>    <br>    nx.draw_networkx_edges(G, pos, edgelist=s6,<br>                           width=7, alpha=1, edge_color=&#39;pink&#39;)</pre><pre>nx.draw_networkx_edges(G, pos, edgelist=s7,<br>                           width=16, alpha=1, edge_color=&#39;orange&#39;)</pre><pre>plt.axis(&#39;on&#39;)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uMoTXGbEPEF3V4KzzkN9YA.png" /><figcaption>Network chart: At least one co-investment</figcaption></figure><p>The figure above is our final result! I hid the names for better visualization, but they are useful in a detailed analysis. We can notice that some funds have a strong connection, meaning they have many co-investments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UFv6fLJXzh52R-Drzk0bPA.png" /></figure><p>I highlighted some groups in the chart. Group 1 includes some of the first Brazillian Venture Capital Funds who raised capital from the government to foster the ecosystem. Group 2 includes the most active funds, those with a major number of investments, exits, and unicorns. Group 3 includes private VCs and government-backed VCs, being connected by investment in agribusiness startups.</p><p>The figure below is output from <a href="https://flourish.studio/">Flourish</a>, a platform that allows us to turn spreadsheets into interactive online charts without coding.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/936/1*NbtAPOLADZROcTqQ4D8gUg.png" /><figcaption>Network chart — Flourish</figcaption></figure><h4>Final comments</h4><p>The chart above is a proxy of how the Brazillian Venture Capital funds are connected, to obtain a conclusive outcome we need to search for alternatives sources of data besides Crunchbase (funds’ website is a good fount!). Networks charts across time is also a good way to visualize how new funds are getting into the market.</p><p>The network chart for monitoring co-investments is only one example of how a VC can use data in its daily activities. Since I started working on a data-driven VC fund I have come across several ways to use public and proprietary data to enrich investment decisions.</p><p><em>If you find any mistakes, please don’t hesitate to contact me! I started to surf the data science world recently and in spite of loving it, I am an amateur.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8c13ec0c8458" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/data-science-in-venture-capital-8c13ec0c8458">Data Science in Venture Capital</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What’s Tidy Data?]]></title>
            <link>https://medium.com/data-science/whats-tidy-data-how-to-organize-messy-datasets-in-python-with-melt-and-pivotable-functions-5d52daa996c9?source=rss-2b3565e4767------2</link>
            <guid isPermaLink="false">https://medium.com/p/5d52daa996c9</guid>
            <category><![CDATA[data-cleaning]]></category>
            <category><![CDATA[how-to-use-pivot-table]]></category>
            <category><![CDATA[melt]]></category>
            <category><![CDATA[tidy-data]]></category>
            <category><![CDATA[pandas]]></category>
            <dc:creator><![CDATA[Rodrigo Marino]]></dc:creator>
            <pubDate>Wed, 15 Jan 2020 23:24:33 GMT</pubDate>
            <atom:updated>2020-01-18T14:25:50.846Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to organize messy datasets in Python with Melt and Pivotable functions</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7jjzhy4KknPz9hJVnC_w7w.png" /><figcaption>Source: R for Data Science (Hadley Wickham &amp; Garrett Grolemund)</figcaption></figure><p>Data scientists spend about 80% of their time cleaning and organizing the data. Tidy Data is a way of structuring datasets to facilitate analysis.</p><p>In 2014, Hadley Wickham published an awesome <a href="https://vita.had.co.nz/papers/tidy-data.pdf">paper named Tidy Data</a>, that describes the process of tidying a dataset in R. My goal with this article is to summarize these steps and show the code in Python.</p><p>In <strong>tidy data</strong>:</p><ol><li>Each variable must have its own column.</li><li>Each observation must have its own row.</li><li>Each type of observational unit forms a table.</li></ol><p>Messy data is any other arrangement of the data.</p><h3>Messy Data</h3><p>There are 5 examples of messy data we will explore here:</p><ul><li>Column headers are values, not variable names.</li><li>Multiple variables are stored in one column.</li><li>Variables are stored in both rows and columns.</li><li>Multiple types of observational units are stored in the same table.</li><li>A single observational unit is stored in multiple tables.</li></ul><p>Of course, there are more types of messiness that are not shown above, but they can be tidied in a similar way.</p><h4>Column headers are values, not variable names.</h4><p>For this example, we will use the dataset “relinc.csv” which explores the relationship between income and religion. Note, despite being messy, this arrangement can be useful in some cases, so we will learning how to tidy and untidy it.</p><pre>import pandas as pd</pre><pre>df_relinc=pd.read_csv(&quot;relinc.csv&quot;)<br>df_relinc.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EEOa37nAPNW9sQFP0MUlWw.png" /></figure><p>There are three variables in this dataset: religion, income, and frequency. The column headers are values, not variable names, so we need to turn the variables in the columns (income) into rows. We will use Panda’s function <strong>Melt.</strong></p><pre># Applying melt (to a long format)</pre><pre>df_relinc=df_relinc.melt(id_vars=[&quot;religion&quot;],var_name=[&quot;income&quot;],value_name=&quot;frequency&quot;)</pre><pre>df_relinc.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lK-ZSuZ0t4fGV6VHsr9Fpw.png" /></figure><p>The output above is our tidy version of the dataset.</p><p>To return the dataset to a wide format we will use Panda’s function <strong>pivot_table</strong>.</p><pre># Applying pivot_table (to a wide format)</pre><pre>df_relinc=(df_relinc.pivot_table(index = &quot;religion&quot;, columns = &quot;income&quot;, values = &quot;frequency&quot;)<br>   .reset_index() <br>   .rename_axis(None, axis = 1))</pre><pre>df_relinc.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K_YiJqZfUUQO4vT0TwMd8g.png" /></figure><h4>Multiple variables stored in one column.</h4><p>Now we will explore the tuberculosis dataset from the World Health Organisation. The records show the count of tuberculosis cases by country, year and demographic group.</p><p>The demographic groups are broken down by sex (m, f) and age (0–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65+, unknown).</p><pre>df_tb=pd.read_csv(&#39;tb.csv&#39;)<br>df_tb.columns</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Px5g3Mg_dby-aRnYAxbV8Q.png" /></figure><pre>df_tb.tail()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9P45U4mC3EERt374AVjhCQ.png" /></figure><p>The columns contain sex and age values. Note the mixture of 0s and missing values (NaN). This is due to the data collection process and the distinction is important for this dataset.</p><p>First, we will gather up the non-variable columns, by moving the age range and sex to a single column. To do so, we will use the <strong>Melt.</strong></p><pre># Applying melt (to a long format)</pre><pre>df_tb=df_tb.melt(id_vars=[&quot;iso2&quot;,&quot;year&quot;],var_name=[&quot;demographic&quot;],value_name=&quot;cases&quot;)</pre><pre>df_tb.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g2Nuhppw7zKlJXJPSEx9fw.png" /></figure><p>Now we need to split the column demographic to get two columns for the variables sex and age.</p><pre># Creating new columns for sex and age</pre><pre>df_tb=(df_tb.assign(</pre><pre>sex = lambda x: x.demographic.str[0].astype(str),<br>age = lambda x: x.demographic.str[1:].astype(str))<br>      .drop(&quot;demographic&quot;,axis=1))</pre><pre>df_tb.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5kmayl7gerJCiZYjbJ8Cgw.png" /></figure><p>Now, each observation has its own row and each variable has its own column. We just tidied our dataset! Before go ahead, let’s clean the data.</p><pre># Styling the dataset</pre><pre>df_tb.update(pd.DataFrame({&quot;age&quot;:[age[:2]+&#39;-&#39;+age[2:] if len(age) == 4 else (age) for age in df_tb[&quot;age&quot;]]}))</pre><pre>df_tb=(df_tb.replace(to_replace =[&quot;m&quot;,&quot;f&quot;,&quot;014&quot;,&quot;65&quot;,&quot;u&quot;],value =[&quot;Male&quot;,&quot;Female&quot;,&quot;0-14&quot;,&quot;65+&quot;,&quot;unknown&quot;])<br>            .dropna())</pre><pre>df_tb.sample(10)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1s_RfdsiIxmckGRAO5F89Q.png" /></figure><h4>Variables are stored in both rows and columns.</h4><p>We will use the data from the Global Historical Climatology Network that represents the daily weather station (MX17004) in Mexico for five months in 2010.</p><p>It has variables in individual columns (id, year, month), spread across columns (day, d1–d31) and across rows (tmin, tmax) (minimum and maximum temperature). Months with fewer than 31 days have structural missing values for the last day(s) of the month. The columns d9 to d31 have been omitted for better visualization.</p><pre>import datetime</pre><pre>df_weather = pd.read_csv(&#39;weather-raw.csv&#39;)</pre><pre>df_weather.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*og3a7tzqW78COCXf4lcMxA.png" /></figure><p>As seen above, the dataset is messy. Variables are stored in both rows (tmin, tmax) and columns (days). Let’s start by working on d1, d2, d3… columns.</p><p>We will apply <strong>melt </strong>to create a row for each record for the day variable.</p><pre># Applying melt (to a long format)</pre><pre>df_weather=df_weather.melt(id_vars=[&quot;id&quot;,&quot;year&quot;,&quot;month&quot;,&quot;element&quot;],var_name=[&quot;day&quot;],value_name=&quot;temp&quot;)<br>df_weather.update(pd.DataFrame({&quot;day&quot;:[day[1:] for day in df_weather[&quot;day&quot;]]}))</pre><pre>df_weather.sample(5)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s0WHQ__JoyBAtLmBpMYi-A.png" /></figure><p>Now, we will use <strong>pivot_table</strong> function to create new columns for the tmin and tmax, once they are variables.</p><pre># applying pivot_table to create columns for tmin and tmax</pre><pre>df_weather=(df_weather.pivot_table(index = [&quot;year&quot;,&quot;month&quot;,&quot;day&quot;,&quot;id&quot;], columns = &quot;element&quot;, values = &quot;temp&quot;)<br>       .reset_index().rename_axis(None, axis = 1))</pre><pre>df_weather</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hsk6JP-uYb2ol7Ylr4iuhg.png" /></figure><p>The dataset looks better but we still need to improve it. Let’s create a column for the dates and group it.</p><pre># Creating a date column</pre><pre>df_weather=(df_weather.assign(date = lambda x: x.year.astype(&quot;str&quot;) +&quot;/&quot;+ x.month.astype(&quot;str&quot;).str.zfill(2) +&quot;/&quot;+ x.day.astype(&quot;str&quot;).str.zfill(2))<br>            .drop([&quot;year&quot;, &quot;month&quot;,&quot;day&quot;],axis=1))<br>df_weather[&#39;date&#39;] =  pd.to_datetime(df_weather[&#39;date&#39;], format=&#39;%Y/%m/%d&#39;)</pre><pre># Grouping by date</pre><pre>df_weather=(df_weather.filter([&quot;date&quot;, &quot;tmax&quot;,&quot;tmin&quot;]).groupby(&quot;date&quot;).aggregate(&quot;mean&quot;).reset_index())</pre><pre>df_weather</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vnX7rWTS5CAuseEQm30yew.png" /></figure><p>We finally tidied our dataset.</p><h4>Multiple types of observational units are stored in the same table.</h4><p>The dataset shows the Billboard top hits for 2000. This dataset records the date a song first entered the Billboard Top 100. It has variables for artist, track, date entered, date peaked, genre, time, rank and week.</p><pre>import pandas as pd<br>import re<br>import numpy as np<br>import datetime</pre><pre>df_bill = pd.read_csv(&#39;billboard.csv&#39;,header=0,encoding = &#39;unicode_escape&#39;)</pre><pre>df_bill.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J2FVVptdEBcMWzDGkTLxww.png" /></figure><p>The rank in each week after it enters the top 100 is recorded in 76 columns, x1st.week to w76rd.week. If a song is in the Top 100 for less than 76 weeks the remaining columns are filled with NaN.</p><pre>df_bill.columns</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qwTlwH-jiE17-eH6e2IrkQ.png" /></figure><p>This dataset contains observations on two types of observational units: the song and its rank in each week. As a consequence of it, the artist and time are repeated for every song in each week. Before break the Billboard dataset into two, we need to tidy it. Let’s start by gathering all the week columns.</p><pre># Applying melt (to a long format)<br>df_bill=(df_bill.melt(id_vars=[&quot;year&quot;,&quot;artist.inverted&quot;,&quot;track&quot;,&quot;genre&quot;,&quot;date.entered&quot;,&quot;date.peaked&quot;,&quot;time&quot;],var_name=[&quot;week&quot;],value_name=&quot;rank&quot;))</pre><pre># Week to number <br>df_bill.update(pd.DataFrame({&quot;week&quot;: np.ravel([list(map(int, re.findall(r&#39;\d+&#39;, i))) for i in df_bill[&quot;week&quot;]])}))</pre><pre>df_bill.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zCHcjCYwg7RqHKYMnO4fsA.png" /></figure><p>It looks better! Now we have a column for the variable week. By the way, we can use the information from the date entered and the week to create a new column, which will be the date column.</p><pre># creating a date column from date.entered and week</pre><pre>df_bill[&#39;date.entered&#39;] =  pd.to_datetime(df_bill[&#39;date.entered&#39;], format=&#39;%Y/%m/%d&#39;)</pre><pre>df_bill=(df_bill.assign(date= [df_bill[&#39;date.entered&#39;][i]+datetime.timedelta(weeks = df_bill[&quot;week&quot;][i]-1) for i in range(len(df_bill[&quot;week&quot;]))])<br>         .drop([&#39;date.entered&#39;,&#39;date.peaked&#39;,&#39;week&#39;], axis=1)<br>         .sort_values(&#39;artist.inverted&#39;, ascending=True)<br>         .reset_index(drop=True))</pre><pre>df_bill.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pn29xYo7JvFssqP6IgDj7Q.png" /></figure><p>Now, we will create an id from the track. Each song must have a unique id number. To do so, we will use Panda’s function <strong>factorize.</strong></p><pre># creating an id column from track</pre><pre>labels,unique=pd.factorize(list(df_bill[&quot;track&quot;]))<br>df_bill[&quot;id&quot;]=labels+1<br>df_bill.reset_index(drop=True)<br>df_bill.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B2e3f55MLvLn6luq0pwN7A.png" /></figure><p>Finally, we will break our dataset into two datasets: the song dataset and the rank dataset.</p><pre># creating a new dataframe for rank</pre><pre>df_rank=df_bill.filter([&quot;id&quot;, &quot;date&quot;, &quot;rank&quot;]).dropna()<br>df_rank=df_rank.sort_values(by=[&#39;id&#39;,&#39;date&#39;]).reset_index(drop=True)</pre><pre># creating a new dataframe for song</pre><pre>df_song=df_bill.filter([&quot;id&quot;, &quot;artist.inverted&quot;, &quot;track&quot;,&quot;time&quot;])<br>df_song=df_song.drop_duplicates(&#39;id&#39;).reset_index(drop=True)</pre><pre>df_rank.head(10)<br>df_song.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L_SMPrntfcxDGOLjc98JlA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TPtiuxEAGBKExgt3DQf0sA.png" /></figure><p>We just tackled the problem of multiple types of observational units stored in the same table!</p><h4>A single observational unit is stored in multiple tables.</h4><p>This problem uses to be easy to fix. We basically need to read the tables, to add a new column that records the original file name and finally combine all tables into a single one.</p><pre>import pandas as pd</pre><pre>df_baby14 = pd.read_csv(&quot;2014-baby-names-illinois.csv&quot;)<br>df_baby15 = pd.read_csv(&quot;2015-baby-names-illinois.csv&quot;)</pre><pre>df_baby14.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zcNPh9b-DXqY1d6NX5yiFw.png" /></figure><pre>df_baby15.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FHbNj3hX3QDoHRa8JK9lOQ.png" /></figure><p>Let’s create a column year in each dataset according to the file name. Finally, we will apply Panda’s <strong>concat</strong> function to concatenate the data frames.</p><pre># Creating a column for the year<br>df_baby14[&quot;year&quot;]=&quot;2014&quot;<br>df_baby15[&quot;year&quot;]=&quot;2015&quot;</pre><pre># Concatenating the datasets<br>df_baby = pd.concat([df_baby14, df_baby15]).sort_values(by=[&#39;rank&#39;])</pre><pre>(df_baby.set_index(&#39;rank&#39;, inplace=True))</pre><pre>df_baby.head()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ch5PiLbRHim-RDaqF8vQRw.png" /></figure><h3>Final Comments</h3><p>The goal of this article was to explain the concept of Tidy Data, by covering the five most commons types of messy data, as well as how to organize and clean these datasets in Python.</p><p>If you find any mistakes, please don’t hesitate to contact me! I started to surf the data science world recently and in spite of loving it, I am a dummy.</p><p>My code and the datasets are available on Github: <a href="https://github.com/rodrigomf5/Tidydata">https://github.com/rodrigomf5/Tidydata</a>. Also, here is the Hadley Github with much more information about Tidy Data: <a href="https://github.com/hadley">https://github.com/hadley</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5d52daa996c9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/whats-tidy-data-how-to-organize-messy-datasets-in-python-with-melt-and-pivotable-functions-5d52daa996c9">What’s Tidy Data?</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quanto as maiores startups brasileiras captaram ao longo do tempo?]]></title>
            <link>https://medium.com/@rodrigo.marino/quanto-as-maiores-startups-brasileiras-captaram-ao-longo-do-tempo-b218f393b738?source=rss-2b3565e4767------2</link>
            <guid isPermaLink="false">https://medium.com/p/b218f393b738</guid>
            <category><![CDATA[brazil-startup]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[startups-in-brazil]]></category>
            <category><![CDATA[fundraising]]></category>
            <category><![CDATA[unicornio]]></category>
            <dc:creator><![CDATA[Rodrigo Marino]]></dc:creator>
            <pubDate>Fri, 03 Jan 2020 14:09:04 GMT</pubDate>
            <atom:updated>2020-01-03T14:09:04.781Z</atom:updated>
            <content:encoded><![CDATA[<p>O ano de 2019 foi bastante positivo para o ecossistema de Venture Capital, uma vez que o Brasil ganhou mais cinco unicórnos (startups com <em>valuation</em> de pelo menos USD 1 bilhão), sendo eles: Loggi, Gympass, QuintoAndar, Ebanx e WildLife</p><p>O objetivo desse artigo é ilustrar a evolução do valor captado em cada rodada de investimento das maiores startups brasileiras até dezembro de 2019. Para isso, utilizei dados do <a href="https://www.crunchbase.com/">Crunchbase</a>. Para a criação dos gráficos considerei somente as startups que levantaram um total superior à USD 75 milhões em rodadas investimento. Esse corte buscou selecionar as startups que estão mais próximas (ou que já se tornaram) unicórnios, preservando a visualização gráfica.</p><p>Considerei somente rodadas de <a href="https://blog.ycombinator.com/how-to-raise-a-seed-round/"><em>equity</em></a> ou <a href="https://blog.ycombinator.com/how-to-raise-a-seed-round/"><em>convertible</em></a><em>,</em> logo as rodadas de <a href="http://www.fundingpost.com/glossary/debt-financing.asp"><em>debt financing</em></a> e <a href="https://en.wikipedia.org/wiki/Private_equity_secondary_market"><em>secondary market</em></a> do Nubank não foram incluídas. Uma parte relevante do valor investido nas rodadas <em>late-stage</em> não resulta em dinheiro no caixa das startups, pois essa parcela pode ser destinada à fundos <em>early-stage</em> que estão realizando a saída ou aos próprios empreendedores. Com a base de dados utilizada não foi possível realizar a distinção.</p><p>Você sentirá falta de algumas startups! Infelizmente não são todas as grandes startups que possuem a série histórica de investimentos no Crunchbase, como é o caso da <a href="https://braziljournal.com/breaking-gympass-vale-mais-de-us-1-bi-com-rodada-do-softbank">Gympass</a> que já captou mais de USD 300 milhões e da <a href="https://www.reuters.com/article/us-softbank-latam-vtex/softbank-leads-140-million-funding-round-for-brazil-start-up-vtex-idUSKBN1XW2AK">VTEX</a> que recentemente concluiu uma rodada de USD 140 milhões. Além disso, não considerei rodadas compostas exclusivamente por anjos, assim as rodadas <a href="https://www.techrepublic.com/article/glossary-startup-and-venture-capital-terms-you-should-know/"><em>seed</em></a> contaram com a participação de pelo menos um fundo de investimento.</p><p>Não incluí nos gráficos abaixo a <a href="https://braziljournal.com/wildlife-o-unicornio-global-brasileiro-ja-vale-us-13-bi">WildLife</a>, que foi avaliada em USD 1.3 bn com o investimento <em>SeriesA</em> de USD 60 milhões. A VivaReal, que apesar de ter captado USD 74.8 milhões, concluiu sua última rodada em 2014 e por esse motivo não foi incluída. Também adicionei startups que já realizaram IPO, como a Stone, PagSeguro e Arco Educação.</p><p>Os gráficos vão só até dezembro de 2019, logo já estão desatualizados. A <a href="https://epocanegocios.globo.com/Empresa/noticia/2020/01/exclusivo-apos-novo-aporte-loft-se-torna-11-unicornio-brasileiro-e-o-mais-rapido-da-historia-do-pais.html">Loft acabou de captar USD 175 milhões</a>, se tornando o primeiro unicórnio de 2020!</p><h3>Vamos aos gráficos!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hB3oNuTMQ6c6gqQoiAlu3g.png" /><figcaption>Investimento acumulado em USD ao longo do tempo por startup. Fonte: Crunchbase</figcaption></figure><p>A 99 Táxis e a Yellow passaram por um processo de <a href="https://endeavor.org.br/desenvolvimento-pessoal/25-termos-que-todo-empreendedor-precisa-saber/">M&amp;A</a>, logo elas não realizarão novas rodadas de investimentos com fundos de Venture Capital. A 99 Táxis foi <a href="https://valor.globo.com/empresas/noticia/2018/01/03/didi-compra-a-99-que-vale-us-1-bi.ghtml">adquirida pela chinesa Didi Chuxing em janeiro de 2018</a>. Já a <a href="https://techcrunch.com/2019/01/30/electric-scooter-startup-grin-merges-with-brazil-based-yellow/">Yellow realizou uma fusão com a Grin </a>Scooters em janeiro de 2019.</p><p>O Nubank e o IFood captaram muito mais do que as outras startups. Para facilitar a visualização reproduzi o gráfico acima excluindo as duas startups.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OzfBKHvBhIQexCmnrkkqzw.png" /><figcaption>Investimento acumulado em USD ao longo do tempo por startup (Excluindo IFood e Nubank). Fonte: Crunchbase</figcaption></figure><p>No vídeo abaixo é possível visualizar a evolução mensal do investimento acumulado por startup.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fplayer.vimeo.com%2Fvideo%2F382351307%3Fapp_id%3D122963&amp;dntp=1&amp;url=https%3A%2F%2Fvimeo.com%2F382351307&amp;image=https%3A%2F%2Fi.vimeocdn.com%2Fvideo%2F843434480_1280.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=vimeo" width="1808" height="1080" frameborder="0" scrolling="no"><a href="https://medium.com/media/8dce8b3394d8856f2a00b1cc25ce79d7/href">https://medium.com/media/8dce8b3394d8856f2a00b1cc25ce79d7/href</a></iframe><h3>Primeira Rodada de Investimento:</h3><p>Analisando somente o tempo até a primeira captação percebe-se que as startups costumam esperar entre 2 e 20 meses. A MadeiraMadeira é a startup que levou mais tempo para concluir a primeira rodada de investimento, mas também é a startup mais antiga. Possivelmente, nesse período o ecossistema ainda era pouco desenvolvido e as startups precisavam de métricas financeiras mais tangíveis para realizar a primeira rodada de investimento.</p><figure><img alt="Tempo em meses até rodada de investimento, desde a fundação da startup. Fonte: Crunchbase, LinkedIn." src="https://cdn-images-1.medium.com/max/1024/1*bXJX0f9yA3ToS9C7lGrQag.png" /><figcaption>Tempo em meses até rodada de investimento, desde a fundação da startup. Fonte: Crunchbase, LinkedIn.</figcaption></figure><p>Provavelmente os fundadores do QuintoAndar e do Neon conseguiram captar ainda no “Power Point”, ou seja, realizaram a primeira rodada de investimento antes mesmo de fundarem oficialmente a empresa.</p><h3>Loft e Yellow começam na frente!</h3><p>As startups Loft e Yellow são outliers em relação ao valor captado na primeira rodada e também são as duas startups mais novas. Fundadas somente em 2018, ambas levantaram bastante capital logo na primeira rodada. Provavelmente o principal motivo é que tanto a Loft como a Yellow foram criadas por empreendedores que já fundaram startups bem-sucedidas, situação que ajuda a reduzir boa parte da incerteza dos fundos Venture Capital na hora de realizar um investimento.</p><figure><img alt="Boxplot. Gráfico de dispersão com a idade aproximada e o valor captado na primeira rodada de investimentos." src="https://cdn-images-1.medium.com/max/1024/1*bgfV8SA0npuQTBdtTox8bQ.png" /><figcaption>Boxplot. Gráfico de dispersão com a idade aproximada e o valor captado na primeira rodada de investimentos.</figcaption></figure><h3>Tamanho das Rodadas por Startup:</h3><p>O gráfico abaixo mostra a quebra do valor total investido de acordo com a classificação da rodada, sendo que o tamanho do investimento por rodada costuma variar bastante entre as startups. Além disso, as startups QuintoAndar, Creditas, Loggi e MadeiraMadeira receberam investimentos do <a href="https://www.reuters.com/article/us-softbank-group-latin-america/softbank-in-talks-to-invest-in-latam-venture-capital-funds-sources-idUSKCN1VV1UL">Softbank</a> entre 2018 e 2019, situação que resultou no aumento do volume das rodadas que contaram com a participação do fundo.</p><p>Em termos de rodada, o IFood é a startup mais madura. Sua última rodada aconteceu em novembro de 2018, em que a startup levantou <a href="https://www.naspers.com/news/ifood-funding-to-total-unprecedented-usd$500m">USD 500 M em um <em>SeriesG</em></a>. Porém, em termos de valor total captado a liderança é ocupada pelo Nubank. Também é possível perceber a diferença entre a evolução das rodadas, enquanto o Nubank aumentou o valor captado por rodada, o IFood captou menos na rodada <em>SeriesF</em> do que na <em>SeriesE</em>.</p><p>Algumas startups não realizaram a rodada <em>seed. </em>Isso pode indicar que a primeira rodada<em> </em>foi feita por investidores anjos e não foi divulgada no Crunchbase ou apresentou um tamanho típico de <em>SeriesA. </em>Para elaborar os gráficos abaixo, utilizei a classificação do disponível no Crunchbase.</p><p>Em novembro de 2019 o Neon realizou o maior <em>SeriesB</em> do Brasil, em que a startup<a href="https://braziljournal.com/general-atlantic-e-votorantim-investem-r-400-milhoes-no-neon"> levantou USD 95 milhões do Banco Votorantim e da General Atlantic.</a> O maior <em>SeriesA </em>foi da WildLife, no valor de USD 60 milhões. Apesar da rodada ter transformado a WildLife em um unicórnio, não incluí no gráfico abaixo pelo fato de ter sido a única rodada de investimentos da startup.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BjwU99BoXjrYYxqFApY_nQ.png" /><figcaption>Quebra do valor captado por rodada para cada startup. Fonte: Crunchbase</figcaption></figure><p>O gráfico abaixo indica que o tempo entre as rodadas, assim como o valor, costuma variar bastante entre as startups. Também é possível perceber que para a maioria das startups a rodada SeriesA costuma ter um dos maiores intervalos.</p><figure><img alt="Quebra do tempo entre as rodadas por startup. Fonte: Crunchbase." src="https://cdn-images-1.medium.com/max/1024/1*lUcB8dCXjl7zqYjFQbCDFA.png" /><figcaption>Quebra do tempo entre as rodadas por startup. Fonte: Crunchbase.</figcaption></figure><p>É notável o excelente desempenho do Nubank em termos de fundraising. A startup conseguiu captar quantias crescentes rodada após rodada em um curto intervalo de tempo. A startup também apresenta o maior valuation entre os unicórnios, <a href="https://techcrunch.com/2019/07/26/fintech-decacorn-nubank-raises-400m-led-by-tcv/">valendo mais de USD 10 bn </a>de acordo com a última rodada de investimento.</p><p>Se você chegou até aqui, deve estar sentindo falta dos valuations das startups ao longo do tempo! Infelizmente as startups costumam divulgar os valuations somente após se tornarem unicórnios, dificultando uma análise mais aprofundada. Aproveitando o tema valuation, segue o <a href="https://blog.adioma.com/how-startup-valuation-works-infographic/">link</a> para um excelente artigo que explica de forma clara o racional por trás do valor de uma startup.</p><p>O valor captado por rodada, assim como tempo entre as rodadas, varia muito de startup para startup. Variáveis como o setor, o momento do mercado, a tração da startup, a tragetória dos empreendedores etc. influenciam o processo de captação. Espero que os gráficos acima, apesar de não incluirem uma parte das grandes startups, tenham ilustrado o papel o do capital de risco no crescimento exponcial dessas empresas.</p><p>Por fim, os unicórnios já não são raros no Brasil! Na<a href="https://indicatorcapital.com/"> IndicatorCapital </a>monitoramos ativamente o ecossistema, mapeando investimentos, tecnologias e tendências. O ano de 2019 foi recordista não só em número de investimentos, mas principalmente no valor total captado (com uma participação significativa das startups acima) indicando o desenvolvimento do ecossistema.</p><p>Espero ter que atualizar os gráficos em breve!</p><p>O que acharam?</p><p><em>Se encontrarem algum erro, sentirem falta de alguma startup ou tiverem sugestões, é só me falar!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b218f393b738" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>