Wednesday, December 31, 2014

Hortonworks goes public , sticker HDP (Hadoop)

A few weeks ago, I wrote an entry trying to answer this question Why Forbes tweets old news on Cloudera's valuation?
Why did Forbes tweet eight month old news?  (that Cloudera is valued at $4.1 billion) Here are some speculations
  1. Cloudera valuation is not proven by market quotes an IPO did not take place yet.
  2. Cloudera competitor  - Hortonworks Inc   - valued at 1,8 billions -  is rumored that filed to go Public in November 2014.
  3. Another smaller competitor, MapR also announced its intention to go public next year in 2015 (which is only 10 days from now)
  4. All these companies bank on Hadoop. Hadoop, which like Spark (promoted by Databricks) are not user friendly for non-developers. Sure they are excellent tools for big data, see Hadoop 101 However being hard to use is an issue.
    Hortonworks people celebrating their IPO
     Hortonworks (HDP it's their symbol) opened at $16.00  on December 12 and today January 31, 2014 at 3:40 p.m. EST the quote is $27.55, The total valuation is $1,14 billion much lower than Hortonworks  earlier statement that the company is valued at $1.8 M. HDP lost money, lot's of money, $87 million with sales of only $33 M for 9 month in their fiscal year.

    Andrew Burst from Gigaom, writes:
    In the run-up to the IPO, some criticism of it emerged, based (understandably) on the company’s startup status and current lack of profitability. Revenues may be substantial, but if costs exceed Hortonworks, you can’t blame people for being skeptical around its shares.
    Suffice it to say, there are plenty of companies in the data sphere whose business model seems to be all about the exit. These companies are less built to be profitable than to be bought by profitable companies that don’t feel like building competing products themselves. That premise feels non-compliant with the laws of business gravity I learned over my own career, but it does at least follow a certain logic. And Hortonworks’ pre-profit IPO doesn’t seem any more deviant.

    My take


    Hortonworks was always the underdog behind Cloudera, a company that claims revenues and a number of customers  three times as big as HDP.

    The word of mouth  on the Valley says Cloudera reputation is that of "nice people to do business with" as the public image, while HDP appeared as uncouth obsessed runner-ups .

    Cloudera Halloween party.
    Subjective gossip aside, among Hortonworks investors, Larry Fink from BlackRock stands out. His firm  has almost four trillion dollars under management. Wall Street is not run by Institutions, it is run by individual Kings who create companies unable to exist or remain successful without them at the helm.

    I don't know his exact role in HDP IPOs - time will tell - but he definitely played a role. His influence outsmarted Cloudera, who now have to catch up.

    HDP (and New Relic) IPOs unforeseen success s show how thirsty the investors are for high tech stocks. The recent cyber-attacks are a blessing for the high tech bonanzas. I doubt 99% of the buyers of HDP stock know what Hadoop does.

    Ironically, the name Hadoop was coined by Doug Cutting, now Cloudera Chief Architect and formerly at Yahoo!. Hadoop was his son toy elephant. Yet HortonWorks stole the sicker name: HDP

    Are we creating a Hadoop bubble?

    By the time three or four companies will go public, built on services for HADOOP open source software , what is going to happen with the share price?

    Hadoop is just a product to build solutions, What customers need are solutions. IMHO the winner will be the the company which offers solutions anyone can understand. Cloudera offerings on their web site are for geeks, not for investors and not for mainstream customers. See for example their product comparison

    This offer of tangible solutions is the best anti-bubble vaccination. Read Note 1 on this blog entry

    Saturday, December 27, 2014

    Selling performance computing. The Gospel 2015 edition

    On August 27, 2013, I wrote How to sell performance computing in 2013 . Here is an update for 2015.

    How to sell performance solutions 2015

    1 Make User Experience a priority
    2 Listen
    3 Know whom you want to please
    4 Be different: resist the hype temptations
    5 Offer the best there is in technology, wherever you find it
    6 Offer Open Ended Performance  Solutions, not products (see Note 1)
    7 Generate and collect "Aha" testimonials
    8 Seed for follow up business
    Demographics, Behavior and Goals for, Ahrono Consulting Services  Prospects

    How to Start a Startup at Stanford University 

    Fall 2014 Stanford offered the CS183B  class on line that  " is designed to be a sort of one-class business course for people who want to start startups.

    The successful founders must become the best sales people

    Here is a quote from Tyler Bosmeny talk at Stanford
    I'm the CEO of Clever.  I was about to start at a hedge fund, but at the last second a friend of mine roped me into joining his startup to do sales, which I knew nothing about. I had to figure it out on the fly. I spent a couple of years there figuring out sales for this very early stage company. When it came time to start Clever, I started Clever with two co-founders who were very technical and very product oriented. We wanted to build this product for schools and I thought that experience would have no relevancy whatsoever. It turns out that what I picked up while doing sales at this previous job has been a huge part of what’s made Clever grow so quickly today.
    A quick background on Clever: we build software for schools.

     The Bosmeny-zation of sales for 2015 statup

    I had no option but to invent this word, - inspired by Tyler Bosmeny - because starting from 2015 the definition of sales in high tech and this zero-to-one world described by Peter Thiel. The Bosmeny vision on sales is in great part an inspiration from Y Combinator's  (YC) way to do business
    1. What I've learned is that when it comes to "hiring the sales people," as a founder, the reality is that it's you.
    2. As a founder you have some unique advantages that make it possible for you to be really, really good at sales
      1. You know the product
      2. You know the industry
    3. Attend conferences as a speaker
    4. Use personal networks
    5. Do cold emails
    6. Use only one takeaway, fitting the prospect. Two may be too much
    7. Talk maximum 30% of time, listen minimum 70% of the time
    8. Don't accept free trials. Just offer a 30-60 day opt out clause in a yearly contract
    9. Don't accept contract conditional to an one off request for a feature; but offer to include this feature if more customers ask for it
    10. Elephants: startups who want to reach 100 million in sales by selling a $100,000 per year to 1,000 customers
    11. Rabbits: startups who want to reach 100 million in sales by selling a $1,000 per year to 100,000 customers 
    12. You can't spend the same amount of time and people resources (per prospect) selling to Elephants versus Rabbits 
    13. Your goal should be to get people to a yes or no as quickly as you can
      1. Use red lining, meaning send an agreement for their lawyers to look at.

    Note (1) 12/30/2014

    Offer Open Ended Performance  Solutions, not products

    In performance computing  the definition of a product is usually a building block for many solutions. Look at https://www.palantir.com/products/ (the Big Data company). I guess there are a few cases where a customer just buys a product

    Using these products, Palantir  sells mostly solutions, https://www.palantir.com/solutions/  They reached $ 1 billion in sales this way. Some customers may buy the products, but unless they are highly sophisticated users - which the vast majority of prospects aren't - they need to buy consulting services. A complete proposal will include everything needed for buyer to accomplish what she really needs, and not what we think she must have.

    Thursday, December 25, 2014

    Paul Graham's question and Kabbalah

    Paul Graham asks in his December 2014 essay: " "Can you protect yourself against obsolete beliefs?" 

    His answer is to look for hints. If an idea that looked preposterous before, it makes sense now, we have a change and we have an opportunity. He borrows the words from Linda Rottenberg's book - Crazy is a Compliment - to describe how they select the funding for new startups in Y Combinator. Rather that debate over and over over new idea, as we do on twitter most of of the time, we should bet on this idea.

    In Kabbalah, there are interpretations that mirror the idea of Paul Graham. Here is quote from the kabbalist Rav Yitzchak Ginsburgh. Rav Ginsburg is an University of Chicago graduate in mathematics and philosophy and now is a highly respected teacher of Kabbalah in Jerusalem.
    The beginning of turning the world upside down begins with turning my own world upside down and revealing the yechidah in my psyche. "World" in Hebrew means "concealment." The people walking in darkness saw a great light. People itself עם also means darkness. Each one of us is in darkness and we are walking in darkness, and we have to turn upside down and shine a great light. The light that comes out of darkness is infinitely much stronger and more essential than direct light. That is why the Almighty created the world in a manner of "first darkness and then light," so that the great light would shine. Thank God, as it were, we have a lot of darkness, and now is the time for the great light to shine through this darkness. this truth that is in each one of us is what needs to shine through.
    So the one kabbalistic answer to the question: "Can you protect yourself against obsolete beliefs?"  is: "Yes by turning the visible world  - which hides the real world - upside down. Then we turn ourselves upside down and see the light within the darkness."

    Saturday, December 20, 2014

    Why Forbes tweets old news on Cloudera's valuation?

    Forbes magazine twitted yesterday December 19, 2014

    On April 8, 2014 I published on this blog the same thing

    What is this?

    Why did Forbes tweet eight month old news? Here are some speculations
    1. Cloudera valuation is not proven by market quotes an IPO did not take place yet.
    2. Cloudera competitor  - Hortonworks Inc   - valued at 1,8 billions -  is rumored that filed to go Public in November 2014. 
    3. Another smaller competitor, MapR also announced its intention to go public next year in 2015 (which is only 10 days from now)
    All these companies bank on Hadoop. Hadoop, which like Spark (promoted by Databricks) are not user friendly for non-developers. Sure they are excellent tools for big data, see Hadoop 101 However being hard to use is an issue.

    How much room we have on market for three varieties of Hadoop and Spark ? 

    Loosing money

    Quoting from WSJ
    Hortonworks’ IPO filing shows both the opportunity and the challenges. The company’s revenue has more than doubled in the past year, while operating expenses and losses also have roughly doubled.
    For the nine months ended Sept. 30, revenue was nearly $33.4 million, compared with nearly $16 million for that same period in 2013. But losses were also substantial and growing—$86.7 million in the first nine months compared with $48.4 million in 2013.
    Eric Baldeschwieler, a Hortonworks co-founder who was the company’s CTO until he departed last year, acknowledged it was still early days. However, signs in the market were positive, he said. “We’ve seen companies move from pedestrian use cases at small scale to exciting use cases at a large scale,” he said.
    Cloudera CEO Tom Reilly said in an interview that his company wasn’t yet ready for an IPO, although its revenues and customer numbers were double those of Hortonworks, which said it had just under 300 customers, including partners. Cloudera was also losing money, but at a lower rate, he said. “We would like to be a public company, and we will do it on our time,” he said.
    Unlike Hortonworks, which received a sizable portion of its revenue from its partnership with Microsoft, Cloudera’s revenue came entirely from paying clients, Riley said.
    Not making a profit now is acceptable if long term the expectations of future profits will transform in reality. Twitter did not make a profit yet.

    But what Hadoop companies lack,  is the built-in ability top go viral like LinkedIn, Twitter, Facebook to attract customers fast.

    The puzzle

    How can a company with 300 customers be valued at 1.8M? Can a company with 1,000 customers be valued at $4.1 million? Who am I to judge, other than being a small independent blogger.

    Like the child from the tale The Emperor's New Clothes I see that Hortonworks value is $ 6 million per customer and Cloudera assuming 1,000 customers has a value of $4.1 million per customer

    By comparison Twitter valuation of $24 billion for 0.3 billion users, represent $60 per user. This means Horton valuation per user / customer is 100,000x that of Twitter.

    One may argue that a customer and a user are two different things. But for Hadoop companies, what it counts is  who pays the bills.

    I have seen many wonders in the Valley, and I sure I will see another one. But the key of the success is to make the customers addicted to Hadoop. This means  ease of use

    Thursday, December 18, 2014

    Marketing predictions versus reality in 2014

    All Twitter, the unofficial resource,  published this infographic:
    From Media Bistro

    Take aways

    There are five predictions covered, According to the Infographic, two came out true, two came out false and one had mixed results. Mixed results is a de-facto false, as no one can make investment decisions based on something vague

    If we look carefully, the true predictions are not big data related. They are the results surveys carried out, with responses Yes or No

    • Did you integrate email with social media? Yes? No?
    • Did you leverage data from social media to gain insights to your customers? Yes? No?
    The so called "true" predictions are in fact built-in in social media and everyone, not only marketers, answer yes to these questions. For example  on LinkedIn I send InMails, invitations, requests for references , etc to network efficiently. I engage with people after reading their tweets, google and bing them. And so does everybody else I know.

    Big Data predictions?

    Analytics is  the process of collecting, processing and analyzing data to generate insights that inform fact-based decision-making. 

    One type of forecasting are election polls. We refined those to the point that a few hundred, maybe one thousand interviews can predict election results in a nation of a few hundred million people population. But this is not big data, as Peter Theil says the most valuable companies in the future won’t ask what problems can be solved with computers alone. Instead, they’ll ask: how can computers help humans solve hard problems?

    Not all polling professional firms are equal, although they use similar algorithms. Some are better, because the people in those companies are more skilled at interpretation data. We used to call this statistics. Now we may call it Big Data, as we expect some voice to come out of a machine generated melange and tell us what to do.

    In this context, 100% of all predictions made using algorithms for marketing trend predictions in 2014 are inconclusive and do not inspire confidence. (See Larry Fink, CEO of Blackstone  dilemma

    Fortune cookies

    I invited my son to a fast food Chinese restaurant , 

    My fortune cookie said: You will soon be honored by someone you respect.
    My son cookie said: Before a evening of romance, don't forget to turn off the cell phone

    Both tickets are prophetic and applicable to the entire population of the world. The rule is: the vaguer the prediction, the better chance has to occur. 


    Monday, December 15, 2014

    Docker revolution and Joyent cool idea

    Summary

    Docker is making waves and its adoption is viral because it is easy to use and its' portability. Docker is the simplest and most elegant tool for applications deployment invented so far.

    However as everyone attempts to make Docker work on all environment possible, there is a new, labyrinthine muddy  eco-system which obliterates Docker extraordinary user experience. 

    User experience is the single most important element of success.

    What looks promising is that a reputable player in cloud computing, like Joyent, who declared to create a single place for container deployment and restore the original ease of use Docker's deployments plus unparalleled performance 

    Why Docker 's viral spread?

    My post Docker, a New Business Star from January 2014 anticipated the creation of the new buzz word that is all over the media. The blog article is included on Docker's relevant new and article listed on their web site

    How did Docker become a "zero-to-one" solution? I quote from original post above:
    Originally designed as a platform as a service (PaaS), Docker showed promise for its flexible capabilities in providing developers with a service that supported multiple programming languages. But the competition from companies like Heroku and VMware’s Cloud Foundry made for a challenging market, further exacerbated by the lack of a widespread market acceptance for the benefits that PaaS providers offered.
    This above was a crowded "one-to-n" with lot of competition
    But developers did need a way to move their code to cloud services in a lightweight way without the tax of heavy virtual machines that were difficult to move and required a degree of manual integration. The problem stemmed from the virtualization technology itself, which sits below the operating system. It virtualizes the server, not the app. And because of that, the operating system has to move in order to run the app wherever it might be transported. Once delivered, it has to be booted up and configured to run with the database and the rest of the stack that it depends on.
    With Docker, the container sits on top of the operating system. The only thing that moves is the code. The developer does not have to boot and config. Instead, the container syncs with the cloud service.
    The brilliant idea they got rid of the Hypervisor. They also made it easy to use, which is huge

    The Docker ecosystem

    Now that everyone wants to use Docker, we have an entire, mind-blogging  eco-system
    Click on  http://www.mindmeister.com/389671722/docker-ecosystem 
    Looking at all this, the simplicity of Docker goes away. This is typical in high tech. Someone invents something simple. Then we manage to make something very powerful, but very complex and unmanageable by mainstreams users

    Translating complexity for 95% of the users

    I am fortunate to work and learn from some of the best high performance engineers around. I asked them y to look at this demo. They are hard to please. They detect immediately what is so-and so.

    I pointed them this demo on you tube, Docker Clustering on Mesos with Marathon I am not embedding it here as takes 40 minutes to watch. The demo is 80% command line writing. The audience was as delighted  as a music connoisseur listen to a piano concerto by Arthur Rubinstein.

    I think I understood part of the demo, by my expert friends were delighted:.
    "that's bloody amazing!  I knew Mesos worked, I had no idea it could be done that easily.  If I was going to deploy Docker in a production environment, that's how I'd do it."
    "though running docker with mesos is nice.  Google is pushing their own solution: https://cloud.google.com/container-engine/ which uses the open source Kubernetes: https://github.com/GoogleCloudPlatform/kubernetes Anyways, good looking proof of concept"

    Joyent

    Joyent is one of the nicest cloud company I know. I always admired them. Many great people left the company. Other good people joined the company. CTO Bryan Cantrill comes from Sun Micro where he was a star distinguished engineer. 

    Joyent supported for many years its own implementation of a containers on its cloud service but reading from this source on December 3, 2014:
    Joyent CTO Bryan Cantrill says that with Docker containers emerging as a de facto standard, Joyent sees a significant opportunity to not only host application development projects based on Docker containers, but also on actual production applications.
     Rather than relying on hypervisors that introduce both additional compute overhead and network virtualization complexity, Cantrill says Joyent is betting that developers that build applications using Docker containers would much rather attain higher levels of performance using bare-metal servers rather than relying on hypervisors to access virtual machines.

    What Joyent says tacitly is this: "Ok guys, true, you can run Docker anywhere, but if you run them on  Joyent cloud you get a performance and security that you don't get anywhere else on the planet, because we are unique."

    If this message gets across, Joyent will want new users who never before used Docker and Joyent together. Who are those users, apart from elite developers? How can they be wow-ed? What  should a user do after staring the beautiful home page? How they can be addicted to Joyent's container solutions? What exactly should a user do, step by step to get hooked first time? How we discover when users feel pleasure using Joyent solution? User sentiment is very important

    There is a great opportunity for Joyent to create a new monopoly, based on innovation, a good monopoly as Peter Thiel says.

    Peter Theil is one of the two seed stage investors in Joyent. And on October 31, they got a series E investment of $15M showing the confidence the investors have in their new focus on containers.

    Saturday, December 06, 2014

    Big Data Definitions

    Big Data definition 

    From Roberto Zicari talk at Stanford University .
    Large amounts of different types of data produced with high velocity from a high number of  various types of sources. These variable and real time data sets require new tools and methods, such as powerful processors, software  and algorithms

    Big Data McKinsey Global Institute definition

    Big Data refers to data sets whose size is beyond the ability of typical database software tools to capture, store, manage and analyze.

    Big Data from Peter Theil's , Zero-to-One book

    Today’s companies have an insatiable appetite for data, mistakenly believing that more data always creates more value. But big data is usually dumb data. Computers can find patterns that elude humans, but they don’t know how to compare patterns from different sources or how to interpret complex behaviors. Actionable insights can only come from a human analyst (or the kind of generalized artificial intelligence that exists only in science fiction).
    We have let ourselves become enchanted by big data only because we exoticize technology. We’re impressed with small feats accomplished by computers alone, but we ignore big achievements from complementarity because the human contribution makes them less uncanny. Watson, Deep Blue, and ever-better machine learning algorithms are cool. But the most valuable companies in the future won’t ask what problems can be solved with computers alone. Instead, they’ll ask: how can computers help humans solve hard problems?
    Thiel, Peter; Masters, Blake (2014-09-16). Zero to One: Notes on Startups, or How to Build the Future (pp. 149-151). Crown Publishing Group. Kindle Edition. 

    Data Science Ontology


    From  Sean McClure  Data Scientist and author of the blog ThoughtWorks.
     You can see the an expandable ontology visualization.  Check it out here: http://www.datascienceontology.com.

    Quote:
    Although by no means exhaustive of every idea, the hope is that there is sufficient depth and detail to help guide and teach newcomers as they try to understand the tools and approaches of data science. Since Wikipedia represents a curated collection of knowledge from a variety of experts it made sense to connect wikis to the low-level concepts to assist those interested in learning more. I would encourage those interested in data science to use the ontology as a starting point, and dig deeper on your own as you learn how each of the concepts fit into the bigger picture. Every concept is a result of decades of work in a particular area, each made to study phenomena from a different angle and to address challenges in a variety of domains. 


    Blog Archive

    About Me

    My photo

    AI and ML for Conversational Economy