<?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 Fredrik Rönnlund on Medium]]></title>
        <description><![CDATA[Stories by Fredrik Rönnlund on Medium]]></description>
        <link>https://medium.com/@fredrikrnnlund?source=rss-3132bad6b2c1------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*x4u8pbcDN10gKcTrJVGDLA.jpeg</url>
            <title>Stories by Fredrik Rönnlund on Medium</title>
            <link>https://medium.com/@fredrikrnnlund?source=rss-3132bad6b2c1------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:57:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@fredrikrnnlund/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[Machine Learning in the cloud vs on-premises]]></title>
            <link>https://medium.com/data-science/machine-learning-in-the-cloud-vs-on-premises-452c9c34e6fd?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/452c9c34e6fd</guid>
            <category><![CDATA[edge-computing]]></category>
            <category><![CDATA[on-premise]]></category>
            <category><![CDATA[hardware]]></category>
            <category><![CDATA[cloud]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Mon, 10 Feb 2020 14:18:20 GMT</pubDate>
            <atom:updated>2020-02-10T14:18:20.812Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KFMF1kqcY3J1EBr_" /></figure><h3>The cloud is just somebody else’s computer</h3><p><em>It’s a running joke among developers that the cloud is just a word for somebody else’s computer. But the fact remains, that by leveraging the cloud you can reap benefits that you couldn’t achieve with your on-premises server farm.</em></p><p>The hyper cloud providers (AWS, Azure, GCP) are able to offer a smaller total cost of ownership while delivering superior features from scalability to security. It doesn’t make financial sense to build everything in-house when you can get it off the shelf for only the time you need it. The cloud vendors are constantly innovating with solutions such as servers that only cost for the time they are used, as opposed to the time they are up and waiting for requests. They are also able to attract talent specialized in e.g. scalability and security in ways that would be impossible for every other vendor on their own.</p><h3>Cloud and Machine Learning</h3><p>Most machine learning experimentation starts from understanding your data on your laptop and doesn’t require that much computation power. But very quickly you will run into the need more than your local CPU can provide you with. The cloud is by far the more scalable place to do machine learning. You’ll get access to the latest GPUs or, even TPUs that you wouldn’t be able to afford and maintain on your own.</p><p>However, on-premises server farms form a valid alternative when your needs fulfill some of the following criteria:</p><ol><li>You need calculation capacity 24/7 (e.g. due to large teams or large models)</li><li>You have sensitive data that cannot leave your data centers for compliance or other reasons</li></ol><p>In many cases, a hybrid solution might also be viable, where the preliminary model testing is done on-premises and the production-ready models are developed on more powerful cloud machines or by trying out hundreds of different models before inference. In most cases and with the right tools, the cloud is the fastest way to do both prototyping as well as production-ready machine learning.</p><p>The three main reasons for doing machine learning in the cloud:</p><ol><li>Flexible resource usage to combat spiky hardware resource needs</li><li>Access to the newest hardware at the click of a button</li><li>De-coupled architecture by not being bound to specific hardware</li></ol><h3>Best practices for ML cloud implementations</h3><p>Having worked with hundreds of companies doing machine learning, we’ve seen the benefits of both on-premises hardware as well as cloud computation. The key takeaways are the following:</p><ul><li>There are differences between cloud providers, especially for enterprises looking for data privacy — find the best one for you</li><li>Ensure your data is in the same cloud data center as the computation you’re going to use is. Training models on large data sets become slow if your data needs to be transferred from e.g. an on-premises data storage across the world for calculation. Store your data in S3 if you’re using AWS EC2 to minimize latency and so forth.</li><li>For structured data (tabular data in a database or Spark cluster), use a pre-processing step that first dumps the data into flat files. Files offer a persistency layer for database queries and thus improve version control and re-usability when you’re training the models. Follow the golden principle of “extract once, train indefinitely”.</li><li>Hybrid solutions are a best practice: If you already have on-premises hardware, regardless if it’s one Tesla V100 or a server-farm of Titan RTXs, use them. Transitioning to the cloud doesn’t happen overnight, and that is fine.</li><li>Get a machine learning platform that abstracts the hardware away from your data scientists.</li><li>Let your data scientists concentrate on data science — not DevOps or MLOps, regardless of which cloud or on-premises hardware they use.</li></ul><p><em>Originally published at </em><a href="https://blog.valohai.com/machine-learning-in-the-cloud-vs-on-premises"><em>https://blog.valohai.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=452c9c34e6fd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/machine-learning-in-the-cloud-vs-on-premises-452c9c34e6fd">Machine Learning in the cloud vs on-premises</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[Three ways to categorize machine learning platforms]]></title>
            <link>https://medium.com/data-science/three-ways-to-categorize-machine-learning-platforms-924cf1588163?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/924cf1588163</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[platform]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Wed, 05 Feb 2020 14:39:01 GMT</pubDate>
            <atom:updated>2020-02-05T14:39:01.326Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zWmJCePlO_VivVJr" /><figcaption>Photo courtesy of <a href="https://unsplash.com/@image_conscious">https://unsplash.com/@image_conscious</a></figcaption></figure><p>Machine learning (ML) platforms take many forms and usually solve only one or a few parts of the ML problem space. So how do you make sense of the different platforms that all call themselves ML platforms?</p><p>Machine learning platforms take many forms, from labeling of data to visualizing data and from training models to monitoring deployed models. All of these different factors, and more, constitute what people call a machine learning platform, so it is your responsibility as a technology evaluator to understand the core concepts around machine learning so that you know what kind of platform best suits your needs. This blogpost will introduce three ways of looking at ML platforms.</p><h3>What are machine learning platforms?</h3><p>Machine learning (ML) platforms are services or tools that allow you to automate or outsource parts of your data science work. The way they do that can, however, be very different. A library, such as MLFlow ( <a href="https://mlflow.org/),">https://mlflow.org/),</a> is a platform, in the same way, that an analytics platform H2O ( <a href="https://www.h2o.ai/)">https://www.h2o.ai/)</a> is a platform. With a few minutes of reading about the two products, you will realize that they solve completely different problems with completely different approaches. How can then a layman compare these two platforms?</p><p>First, let’s define the standard problems in machine learning. On a higher level, machine learning can be divided into three parts:</p><p>Let’s look at each one of these and see what they entail.</p><h3>Data management</h3><p>Data on its part can be divided into tabular data (e.g. databases with customer information) and unstructured data (e.g. images of our product in different scenarios). Data management in machine learning has to do with issues such as collection, preprocessing (ETL), labeling, annotating, and exploring data. Each of these challenges requires different tools.</p><h3>Model development</h3><p>Model training on its part can be divided into feature extraction and training. Training is also different depending on if you’re building AutoML solutions, using pre-existing models, building traditional machine learning models (like decision trees) or deep learning models on large unstructured data. For each case, there are different needs for infrastructure, frameworks and collaboration tools.</p><h3>Prediction serving</h3><p>The main two categories for prediction serving are the way it is deployed, either as a part of a software or as an external access point that can be accessed from serving. Models can also be served for batch inference (when the need for predictions is sporadic) or live inference (when the need is constant). Other issues to consider in prediction serving are AB testing of models, canary serving of new models, rollback of models, model staleness and more.</p><h3>Summary</h3><p>As you can see from the main three machine learning pipeline categories, the needs and tools supporting these needs vary greatly. Further dimensions in the whole toolchain come from the machine learning team’s background. Data scientists with a background in software engineering tend to value tools that allow them to develop models in an IDE whereas recent students of analytics and data science are more accustomed to interactive web interfaces, such as Jupyter notebooks. Also, your company’s internal toolchain from the cloud provider you use or on-premises GPU clusters you might have has an impact on what tools can be used.</p><p>It’s thus fair to assume that the best tool for you is different from that of your competitors.</p><p><em>Originally published at </em><a href="https://blog.valohai.com/three-ways-to-categorize-machine-learning-platforms"><em>https://blog.valohai.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=924cf1588163" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/three-ways-to-categorize-machine-learning-platforms-924cf1588163">Three ways to categorize machine learning platforms</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[The Marketing Moneyball]]></title>
            <link>https://medium.com/swlh/the-marketing-moneyball-a4532e7b94c7?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/a4532e7b94c7</guid>
            <category><![CDATA[sports]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[moneyball]]></category>
            <category><![CDATA[go-to-market]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Wed, 15 Jan 2020 20:31:24 GMT</pubDate>
            <atom:updated>2020-01-15T20:31:24.522Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YWP_BAb8vwCSSBn0Chiilw.jpeg" /></figure><p>Marketing and professional basketball have one thing in common; the importance of <a href="https://en.wikipedia.org/wiki/Moneyball">moneyball</a> – a metrics based approach to finding what works and boosting it. In basketball this is the 3-point shot. The probability of the 3-pointer multiplied by 3 points is larger than the probability of 2-pointers multiplied by 2 points. Thus everyone in the NBA today plays the moneyball — trying to score 3 pointers. In marketing, we also need to find our moneyball metric. And, just like in basketball, it is the rules of the game that dictate your moneyball. Optimizing 3-pointers in hockey wouldn’t make much sense.</p><p>The marketing gurus out there will tell you to focus on the funnel, your marketing automation tools, channels, organizations, PR and whatever their gospel of the day is. But before you go out on a whack and blindly copy someone, get to know your game. Your game is going to be different based on your buyers, your geography, your customers’ company size, the competition and more. These are the rules of your particular game.</p><p>But that’s not all. Your marketing moneyball consists of two things: <em>frameworks</em> that help you concentrate on the metrics and your ability to enjoy the game so you can make the numbers. Just like basketball, 110m hurdles, swimming or snowboarding – it’s not enough to do mechanical work. Ivan Drago didn’t with Rocky, because he was too much of a machine.</p><h3><strong>Frameworks to drive your moneyball</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xCDCu1FDRwkf3ox6J7749Q.jpeg" /></figure><p>I used to compete in 110m hurdles and our training was always divided into three parts: 1) endurance 2) strength and 3) technique. These were our frameworks for practice – the holy trinity for getting better. Our moneyball for the little Ivan Dragos inside of us.</p><p>In marketing, the frameworks we work with are different. First of all, in marketing our Olympic race is at the same time our practice run. We compete way more often than NBA players play a game or I raced in the 90s. But we train just as often – in marketing, every training session is a competition for the buyers. There is no sandbox for marketing, everything is always live, which is a benefit because we get immediate feedback on how we did.</p><p>Instead of walking over thousands of hurdles you’re trying out hundreds of ad copies. Instead of deadlifts and squats, you’re trying variations of webinar blueprints. Instead of VO2 max interval runs, you’re changing conversion points for marketing qualified leads.</p><p>Frameworks do one thing and one thing only, they simplify the world for you so you can focus. And focus is everything – but you have to know what to focus on.</p><blockquote>Sadly it feels like most marketing gurus are preaching the use of 3-pointer moneyball shots with cricket balls for professional swimmers. You have to know your framework, not just blindly copy something that works in a different market, product, competitive landscape and more.</blockquote><p>When you’ve done your job for long enough, you implicitly know your rules and how to follow them. And more importantly, you teach yourself to not even think about them. In both cases, the probability of success also increases if you have a mentor. A mentor will help you realize you’re practicing the wrong thing, or even worse, playing the wrong game.</p><h3>Freedom to make art</h3><p>When getting into your starting blocks for my hurdle dash you can’t be thinking about endurance, strength or technique anymore. It has to all be in muscle memory. Frameworks only help us practice, but when we’re competing – we have to get into the zone and enjoy it.</p><blockquote>It’s an oxymoron, but unless you force yourself to enjoy what you’re doing, you won’t succeed at it.</blockquote><p>When you’re in the zone, you don’t compete against anyone else and the entire event is like a form of meditation. The same is true for figure skating, golf, orienteering, and every other professional sport. If you start looking to the sides or worrying about if your leg will hold, you’ve already lost. You just have to <em>surrender yourself to the zone</em>.</p><p>Both running hurdles, as well as marketing, are a form of art. They support each other: <em>our frameworks give us the confidence to make art. </em>We know what to do but to get to the goal we need to get into our zen state where we create art.</p><p>Every situation is new and the parameters are different; from the direction of the wind and our competitors next to us and from our google search rank to the visitors at a trade show. In order to perform in the moment, we need to let our subconscious brain do its work.</p><h3>TL;DR</h3><p>There is much to be said about marketing frameworks and best practices. But marketing at its essence boils down to these two things:</p><ul><li>You have to build your muscle memory. The frameworks are there to help you know what to do. Nobody learns to run hurdles at an Olympic level without a training program. The same is also true for marketing professionals, we need to find our training program. And that’s why we need coaches.</li><li>Every situation is new and you can’t think of the frameworks while performing. You need to love what you’re doing and make art. Take what the wind or your opponent throws at you and enjoy it. Mechanical processes and frameworks won’t get you past the first hurdle.</li></ul><blockquote>What got you to the starting blocks won’t get you to the finish line. Frameworks give you the confidence to start, but trust and enjoying the moment gets you the gold medal.</blockquote><p>That’s the marketing moneyball.</p><p>Fredrik Rönnlund is the founder of DigitalBooker, builder of Vaadin’s marketing team and the Marketing Lizard at Valohai. Follow him on <a href="https://twitter.com/freducom">Twitter</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a4532e7b94c7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/the-marketing-moneyball-a4532e7b94c7">The Marketing Moneyball💰🏀</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your startup is complex algebra, but not imaginary!]]></title>
            <link>https://medium.com/swlh/your-startup-is-complex-algebra-but-not-imaginary-67d06852fb1d?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/67d06852fb1d</guid>
            <category><![CDATA[lessons-learned]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[focus]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Thu, 21 Nov 2019 10:49:08 GMT</pubDate>
            <atom:updated>2019-11-21T14:42:07.157Z</atom:updated>
            <content:encoded><![CDATA[<h3>Your startup is complex algebra, but not imaginary!😎</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_X7vDz0ns9h9P4FpUP4oMw.jpeg" /></figure><p>If you haven’t studied math after high school, chances are you’ve forgotten most <a href="https://en.wikipedia.org/wiki/Complex_number">complex algebra</a>. Chances also are you’ve never thought of your own startup in the form of complex numbers, i.e. in the imaginary field.</p><p>Complex algebra is the part of mathematics that involves numbers that aren’t real (oh, they’re real but not real-numbers). Numbers like 5+3i, where the real number is 5 and the imaginary number is 3i. The letter i here stands for a number that solves the equation <em>x</em>² = -1. A number that no real number has a solution to.</p><p>Imaginary numbers have real applications in life! They let us fold proteins, calculate trajectories for rockets or plan fluid dynamics in exhaust pipes. They’re used all over in applied science. They, however, have no place in running your startup! And as you’ll see in the picture below, most likely you’ll still end up somewhere on the plane.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/644/1*O2L__dZRbJbrJy5P_TgegQ.png" /><figcaption>The cartesian field with a real and imaginary axis. Where on the axis are you?</figcaption></figure><p>In startups, we tend to pursue the imaginary axis. The axis of quick returns, fluffy promises, and imaginary partnerships. How many times have you attended a conference, because somebody invited you or told you it would be cool? How many times have you spent days collaborating with another company, without knowing what’s at the end of the rainbow? How many times have you implemented a feature because it’s cool? These are all imaginary axis issues.</p><p>At the same time, your startup has been wrestling with real problems, in the real field. All while you’ve been distracted to in your imagination. Your startup’s been struggling to develop the right feature. It’s been struggling to find the right marketing channels. It’s been struggling to negotiate a deal with a new customer. Spending time in the imaginary field didn’t advance these issues. And time is all you have. The money you borrowed from your investors is just borrowed time at the end of the day.</p><p>In the case of startups, the cartesian plane isn’t your friend! Big companies can afford to spend time in any dimension they like — but you can’t.</p><p>Your startup consists of only two pieces: Your product engine and your Go-to-Market (GtM) engine. These two are in the real plane, on the real axis of your cartesian plane. You can’t afford to wander around aimlessly in the cartesian complex plane.</p><p>The next time you do something, ask yourself: Where in the cartesian field are you and where do you wanna be?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67d06852fb1d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/your-startup-is-complex-algebra-but-not-imaginary-67d06852fb1d">Your startup is complex algebra, but not imaginary!</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine learning is a zero-sum game!]]></title>
            <link>https://medium.com/data-science/machine-learning-is-a-zero-sum-game-7498614c9ac3?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/7498614c9ac3</guid>
            <category><![CDATA[enterprise-technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Fri, 15 Nov 2019 11:38:32 GMT</pubDate>
            <atom:updated>2019-11-16T14:50:06.152Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zrajrqgEFz-vwAOY" /></figure><h4>ML Is Unlike Industrialization, Electricity and IT</h4><p>I predict that only the companies that invest in machine learning today will exist 10 years from now. The ones that look to the sidelines will be eaten by their competition.</p><p>The machine learning (ML) revolution is unlike any of previous revolutions; the industrial revolution, electricity or the IT revolution.</p><p>The epoch we call the industrialization, evolved around moving away from manual production methods to mechanical automation. The second industrial revolution took place some hundred years later when we started to get more reliable energy sources through electricity from dams. The third large revolution was through automation in the form of the IT boom from the 1990s all the way to today.</p><p>During these three previous revolutions, companies that didn’t adapt to industrialization, electricity or IT eventually died out. But the ones that did, didn’t get an infinite competitive advantage. The revolutionary steps were huge, but they didn’t result in permanent competitive advantages. Everyone benefited and everyone could start from scratch and fairly quickly catch up. You could e.g. today start a shoe manufacturer and the old players wouldn’t have a competitive advantage because of having used electricity for a hundred years.</p><h4>Early movers in ML get an unfair advantage</h4><blockquote><em>Machine learning is different from previous revolutions because it relies so heavily on historic data you’ve collected.</em></blockquote><p>The ones with the most data will be able to create an unfair competitive advantage! Imagine for instance a manufacturing company that can source their raw material 5% better, reduce loss by 2%, increase energy efficiency by 10%, cut down human errors by 7% and so on. The slimmer the margins of the business, the larger the possibilities of ML in that business. And the more they can optimize due to historic data, the harder it is for an underdog to catch up.</p><p>And the companies that start collecting data earlier and using it to drive further improvement in their ML models will get a permanent unfair advantage. It’s impossible to compete with fewer data and it’s impossible to catch up or start from scratch. Data isn’t the new gold, data is your kryptonite — it will make it impossible for your competitors to get near to you.</p><p>This is especially true in B2B where customer data depends on agreements with other businesses. And already a small competitive advantage will help you win customers that help you grow your data and further improve your ML models.</p><h4>Enterprises have an edge, but only if they wake up</h4><p>The enterprise giants of the world have an early mover’s advantage, in the way that they already have the most data about their customers. When the first giant in a business vertical wakes up to the possibilities of ML, the rest of the giants will be eaten quickly and much faster than e.g. in the past 20 years through IT. And the enterprises left sleeping won’t get a second chance.</p><p>It’s time to wake up. Most of your and my work will eventually be displaced and there is a possibility that governments will regulate data for this reason, but governments are slow and the ones that move now will at least have a 20+ year advantage over the rest. It is a zero-sum game: Do you want to be a zero or a one?</p><p>This article was inspired by a discussion on the Andreessen Horowitz podcast. For the original source and the entire discussion listen to it at <a href="https://podcasts.apple.com/fi/podcast/a16z-podcast/id842818711?i=1000454683165">Apple Podcasts</a>.</p><p><em>Originally published at </em><a href="https://blog.valohai.com/blog-machine-learning-is-a-zero-sum-game"><em>https://blog.valohai.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7498614c9ac3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/machine-learning-is-a-zero-sum-game-7498614c9ac3">Machine learning is a zero-sum game!</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[The Journey from Deep Learning Experimentation to Production-Ready Model Building ]]></title>
            <link>https://medium.com/data-science/the-journey-from-deep-learning-experimentation-to-production-ready-model-building-d9fe294d6be6?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/d9fe294d6be6</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Mon, 03 Dec 2018 16:10:38 GMT</pubDate>
            <atom:updated>2018-12-03T16:23:18.756Z</atom:updated>
            <content:encoded><![CDATA[<h4>Key differences between small-scale and über-scale deep learning</h4><p>Since the rise of the deep learning revolution, springboarded by the <a href="https://www.cs.toronto.edu/~kriz/imagenet_classification_with_deep_convolutional.pdf">Krizhevsky et al. 2012 ImageNet</a> victory, people have thought that data, processing power, and data scientists were the three key ingredients to building AI solutions. The companies with the largest datasets, the most GPUs to train neural networks on, and the smartest data scientists were going to dominate forever.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fpKy_QNXVN1YwIPSYmKtug.jpeg" /></figure><p>This, however, was only part of the truth.</p><p>While it’s true that companies with the most data are able to build better predictive models, the improvements to model quality are not linearly proportional regarding dataset size. At the same time, most companies today either have Big Data™ or are built with data gathering at their core — so one could argue that while more data is a competitive advantage it’s not as big an advantage anymore. The challenges with data are more in having meaningful data lakes with labeled and structured data than the sheer amount of it. There was a correlation between better AI solutions and data, but not necessarily causality.</p><p>Equally, while it’s true that the need for processing power rises in proportion to the amount of data a model is trained on, virtually every company today has access to practically unlimited processing power. Powerful on-premises server farms and big cloud operators offer everyone access to thousands of GPUs over the web. The challenge with processing power is more about making use of these resources in an efficient way than having access to them. Just like airlines optimize the time planes are in the air, effective data science is about optimizing how cloud GPUs are used.</p><p>While there is a shortage of data scientists, shown by their rising salaries (<a href="https://www.glassdoor.com/Salaries/san-jose-data-scientist-salary-SRCH_IL.0,8_IM761_KO9,23.htm">30% higher salaries than their Software Engineer counterparts</a>) the need for algorithmic development is not as imperative as cobbling together models built on pre-researched best practices. AI Expert, Author, and Venture Capitalist <a href="https://www.youtube.com/watch?v=ajGgd9Ld-Wc">Kai Fu Lee</a> called this shift in competence “a shift from thinkers to tinkerers”. We’ve moved along from research to engineering when it comes to AI, and that requires a different set of skills.</p><p>Through this transformation in data, processing power and competence, deep learning has in the past five years matured from the question “how can it be applied?” to the more down-to-earth question “how can we scale to production quickly?”. Building production scale solutions quickly require a new set of tools than those required by research or exploration.</p><p>Let’s have a look at what that means in practice.</p><h3>AI Tools and Frameworks to the rescue! 🚀</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t3A-s2eExektSb4S48r0FQ.jpeg" /></figure><p>The combination of AI hype and the need for more skilled people has drawn in people from different fields into data science. Software engineers, mathematicians, and statisticians all have different backgrounds and different ways of working. Software engineers are probably the only ones that have worked in large groups together under tight deadlines.</p><blockquote>Engineers are by definition the tinkerers of solutions proposed by the thinkers, whereas mathematicians and analysts are more one-person jobs.</blockquote><p>But software engineers have not always been working together either. In the early 1990s and earlier, software development tended to be a one-man job where hero coders hacked together solutions that nobody else understood or could collaborate on. There was no version control that supported real collaboration (files on your desktop, anyone?), not to mention unit testing (println() is testing right?), continuous integration, cloud computing or Scrum (are UML diagrams and use-case specifications still a thing?). The methodologies have been iteratively developed over the past 30 years to accommodate the needs of accelerating software development and highly efficient teams.</p><p><em>However, we’re still lacking these tools in the field of data science today. Instead of using standard tools, people are engineering their own workflows, tools, and frameworks.</em> And it doesn’t help that, this time around, these people come from vastly different backgrounds. Quick and dirty solutions are once again amassing.</p><blockquote>People are sharing “version controlled experiments” in <strong>Excel sheets </strong>over <strong>Slack</strong> with links to <strong>Jupyter</strong> notebooks stored in <strong>Dropbox</strong>. Have we learned nothing from the last 30 years?</blockquote><p>Jupyter notebooks in Dropbox shares are like the Scientology of Deep Learning — they have their fanatic supporters, but most of us aren’t throwing our money at it. And it’s not the tools that are wrong — it’s what you use them for.</p><p>Before we select our tools, we have to agree on the problems we want to solve. I think there are three main goals we want to achieve:</p><ol><li><strong>Quick Experimentation</strong> — we want data science to be quick and agile. People should be able to test out things without spending time on boilerplate code or DevOps work.</li><li><strong>Reproducibility</strong> — we want to ensure reproducibility and an audit trail for every experiment that we conduct. Teams should learn from each other and previous experiments so that we don’t have to reinvent the wheel over and over and over and over again.</li><li><strong>Standardized ways of working </strong>— we want work to be standardized. So when new people join they know how things work. And when somebody leaves, we know what they’ve done previously.</li></ol><p>Let’s have a look at these one at a time!</p><h3>Quick Experimentation🔬</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z7aahOChZZxNfMInuGUN4w.jpeg" /><figcaption>A Data Scientist wants to quickly try out new models and see if they work. And in case they don’t tweak, re-iterate and improve!</figcaption></figure><p>In Deep Learning the core of quick experimentation depends on what stage of your model building you are. In the beginning, you need to be able to explore your data, visualize it and get to know it. A tool such as <a href="https://www.h2o.ai/">H2O</a> is superb for getting to know your data and building your first hypotheses.</p><p>When you get further, you might want to experiment a bit in a Jupyter notebook with Pandas. But the further you get to building production scale models, and especially if your team consists of more than just you, you definitely want to move over to an IDE with proper autocompletion and a way to run your experiments on clusters of more powerful GPUs than the one on your local machine.</p><p>The key to quick experimentation is thus fully automated machine orchestration that is as transparent for the single data scientist as possible. Clicking one button or running one command on the command line would be optimal.</p><h3>Reproducibility👯‍</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sUPIYRsPGlIyI6UkrgmTJQ.jpeg" /><figcaption>Even Darth Vader seems to understand the importance of reproducibility once an optimal solution is found!</figcaption></figure><p>The key to reproducibility in any scientific work is rigorous and complete bookkeeping, i.e. version control, for every experiment. Having to do version control manually is not an option as that would not be your primary focus during model development, leading to random snapshots instead of full reproducibility.</p><p>But unlike with software engineering, reproducibility should not be limited only to your training code but must also include your training &amp; test data, external hyperparameters, software library versions and more.</p><p>In this case, automatic version control for every part of each and every training run is the optimal solution.</p><h3>Standardized Pipeline Management💩</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ApcgSzd_yg5WQ1XK9UVMtQ.jpeg" /><figcaption>Machine Learning pipelines are just like plumbing – you want to ensure the output from your toilet smoothly gets to the cleaning facility – without getting clogged on the way!</figcaption></figure><p>Having the whole team work the same way, with some given degrees of freedom is a necessity. How do you store data? Where do you deploy models and code? Where do you train them? Which frameworks do you use and how do you chain together feature extraction with model training?</p><p>If everyone in the team solved these by themselves, not only would it be a huge waste of time, it would also make collaboration virtually impossible.</p><p>The solution is decoupled pipeline steps with standard ways of chaining and orchestrating them. As a trivial solution, it could just be a script that calls every step of the pipeline in order. But the central part is that it’s standardized within the company and within the team.</p><h3>The Rebirth of Deep Learning: AI Platforms🏆</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EOwaHx1ZPC5fYQqnpIjqxQ.jpeg" /><figcaption>AI Platforms standardize your way of working, abstract away unnecessary complexities and give you the power to move from idea to solution at the speed of a Shinkansen!🚄</figcaption></figure><p>To bring clarity and structure to machine learning, the technology unicorns have been building their own overarching platforms that tie solutions to all of the above challenges together, usually in the form of an AI platform that includes libraries, machine orchestration, version control, pipeline management, and deployment.</p><p><a href="https://code.fb.com/core-data/introducing-fblearner-flow-facebook-s-ai-backbone/">FBLearner Flow</a> is Facebook’s unified platform for orchestrating machines and workflows across the company. <a href="https://soundcloud.com/twiml/bighead-airbnbs-machine">BigHead</a> is AirBnB’s Machine Learning platform for standardizing the way to production, built largely on Apache Spark. <a href="https://eng.uber.com/scaling-michelangelo/">Michelangelo</a> is Uber’s machine orchestration and ML training platform for quickly training ML models.</p><p>The same is true for Google, Netflix and pretty much all larger players that have understood the competitive advantage gained through quick model building.</p><p>But what about the rest of us? What about those who can not invest 10 man-years into building our own orchestration but need results today?</p><p><a href="https://valohai.com/">Valohai</a> is what FBLearner Flow, BigHead and Michelangelo are for the tech unicorns, but built for the rest of us. It is a cloud-based service that can run on AWS, GCP, Azure or your on-premises server farm. Valohai lets you run your models in the cloud, just as if you were running them on your localhost as individual steps or as pipelined workflows. It automatically snapshots every training run so that you can always grab a model running in production (on Valohai’s scalable Kubernetes cluster), click a button and trace back to how it was trained, by whom, with which training data, which version of the code and more.</p><p>Valohai is however not your only alternative — you could build much of it yourself. The important thing is you ensure quick experimentation, reproducibility of your experiments and standardized ways of working. But the real question is, do you want to hit the road running or do you want to start by designing and building your own running shoes?</p><p><em>Originally published at </em><a href="https://blog.valohai.com/from-dl-experimentation-to-production-ready-models"><em>blog.valohai.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d9fe294d6be6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/the-journey-from-deep-learning-experimentation-to-production-ready-model-building-d9fe294d6be6">The Journey from Deep Learning Experimentation to Production-Ready Model Building 🤓</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[Data Scientists are Rocket Surgeons stuck with Stone Age Tools ]]></title>
            <link>https://medium.com/swlh/data-scientists-are-rocket-surgeons-stuck-with-stone-age-tools-e1b92fe171b7?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/e1b92fe171b7</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[developer-tools]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Fri, 07 Sep 2018 08:29:28 GMT</pubDate>
            <atom:updated>2018-09-07T10:49:04.015Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>A software engineer’s view at data science.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hbCg76sT14eqieubp4qMSA.jpeg" /></figure><p>If developers used to be the rock stars of the dotcom era, Data Scientists are quickly overtaking them as the new <a href="https://duckduckgo.com/?q=whitesnake&amp;iax=images&amp;ia=images">Whitesnake cover bands</a> of the 2020s. Although both might be sporting the same hobo beards, Data Scientists are getting their work done with just sticks and stones as their tools while us Software Engineers have every tool in the universe.</p><h3>Data Science Tools</h3><p>Whether you accept our new overlords or not, there’s one thing you, as fellow software engineers, will agree with me on. Data Scientists may be building the next Tower of Babel for all we know, but they’re stuck with steam engines and pitchforks as their tools.</p><p>Where us software engineers couldn’t live without version control for our code, reality for data scientists too often is manual bookkeeping of experiment data, model algorithms, testing environments and training parameters in an Excel file tossed around over Slack to other team members. It’s like Clearcase and Lotus Notes had an illegitimate baby.</p><p>When us Software Engineers want to deploy an app to production, we don’t need to build our own servers anymore — we just deploy to the cloud. Data Scientists, on the other hand, too often have to do the equivalent of mining their own ore and dancing a rain dance before training their models on a server. They need to SSH into a server, install the latest Nvidia drivers, Python dependencies and a clusters of 100 GPUs hosting docker containers over Kubernetes. I know you’d know how to do it, but the difference is, that’s our job. Wanna guess how relevant this is for getting their job done and how much more they could get done if they could just double the 10% they spend on actual machine learning algorithms?</p><p>And it doesn’t stop at the tools, they same is true for team collaboration tools and methodologies. When us software engineers use JIRA and go agile, data scientists need to invent their own ways. In every fricking company and project. Imagine if you had to re-learn everything every time you switched projects! (oh, wait… I guess you <a href="https://github.com/samiheikki/javascript-guessing-game">bleeding-edge JS haxxors</a> do know what I mean…).</p><h3>The ML infrastructure at Facebook, Amazon, Netflix, Google et.al.</h3><p>Most big players in machine learning have seen this problem and started solving it for themselves. For example, Über has built their own <a href="https://eng.uber.com/michelangelo/">Michelangelo</a> toolset for doing version control and server management. Likewise Airbnb has <a href="https://www.slideshare.net/FeiChen29/ml-platform-q1-meetup-airbnbs-endtoend-machine-learning-infrastructure">Bighead</a>, Netflix has something of their own, as do Google and Facebook. These are all in-house proprietary toolsets to make sure their data scientists’ time is not wasted on pipeline orchestration and management tools.</p><p>Not every company however has the muscles to invest into ML orchestration in the same way and many large corporations still fail to see big impact it will have on business. “Fat pigs become lazy”, as they say… As a result these early movers have a HYUUUUGE advantage over the rest — not to mention how impossible it will be for startups to get started when most of your time goes into building tools for themselves.</p><p>Data science is not a one off thing — the world changes, data changes and you constantly need to retrain your model. Unlike software engineering it’s not a simple if-clause somebody wrote in a stored procedure in 1979 running on your DB2, but something that you constantly need to re-train. And the further away you start, the more impossible it becomes to catch up with the competition.</p><h3>The knight on the white horse</h3><p>There are however solutions that don’t involve inventing it yourself. For example at <a href="https://valohai.com/">Valohai</a> we’re in the business of building the same tools for the masses that the big players have built in-house for themselves.</p><p>Whether you’re a software engineer moving to data science or a data scientist tired of “not-invented-here” discussions, you’ll get automatic version control from training data and training environment to training script and algorithms.</p><p>You could of course build it yourself as well — and at the end of the day you already know what to build: Version control, server orchestration and support for ML frameworks and distributed learning. The important thing is you’ll have something to help you concentrate on your actual work: Building predictive models.</p><p>Valohai, on the other hand, is free for all research purposes and super easy to get started with.</p><p><a href="https://app.valohai.com/">Try it out,</a> <a href="https://valohai.com/machine-learning-platform-demo/">schedule a demo</a> and <a href="https://valohai.com/research-license/">apply for the research license.</a></p><p><em>Originally published at </em><a href="https://blog.valohai.com/data-scientists-are-rocket-surgeons-stuck-with-stone-age-tools-"><em>blog.valohai.com</em></a><em>.</em></p><figure><a href="https://medium.com/swlh"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YqDjlKFwScoQYQ62DWEdig.png" /></a></figure><h4>This story is published in <a href="https://medium.com/swlh">The Startup</a>, Medium’s largest entrepreneurship publication followed by +366,567 people.</h4><h4>Subscribe to receive <a href="http://growthsupply.com/the-startup-newsletter/">our top stories here</a>.</h4><figure><a href="https://medium.com/swlh"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ouK9XR4xuNWtCes-TIUNAw.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e1b92fe171b7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/data-scientists-are-rocket-surgeons-stuck-with-stone-age-tools-e1b92fe171b7">Data Scientists are Rocket Surgeons stuck with Stone Age Tools 📠</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When is the prime time of your life? ]]></title>
            <link>https://medium.com/@fredrikrnnlund/what-was-the-best-time-of-your-life-71877e1296b3?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/71877e1296b3</guid>
            <category><![CDATA[nostalgia]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[success]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[life-lessons]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Fri, 20 Oct 2017 22:20:12 GMT</pubDate>
            <atom:updated>2017-10-21T23:59:39.017Z</atom:updated>
            <content:encoded><![CDATA[<p>Define: Nostalgia – <em>a sentimental longing or wistful affection for the past, typically for a period or place with happy personal associations.</em></p><p>The first time I remember being nostalgic was when I was 11 years old. Me and my brother had just finished building a snow castle in our back yard. We were crouching inside of it and letting hot chocolate warm us while the sub zero temperature was making our sweat, from all the digging, freeze to the backs of our necks. This is when it struck me: We had done something I wanted, but it wasn’t for me – it was for the younger me, years before. I remember then and there thinking that IF I only was four years old again I would have enjoyed this much more. If only I was four, I would have been small enough to enjoy the castle, because of it looking so much bigger for a smaller person compared to the pre-teen I had become.</p><p>But when I was four I didn’t have the strength nor the knowledge of how to build a snow castle so I made a mental note to strive for it later. I had now accomplished what I wanted, but it no longer felt like I wanted it as much.</p><p>(I am, however, yet to grow out of the love for snow castles…)</p><p>Us humans are nostalgic and we love to drift off to past, better times. At least I am. I love old turntables, Madison Avenue Whiskey glasses, VW beetles and Old Fashions. Things from the times I, born in the 1980s, haven’t even experienced in their rightful time. From the times of my grandparents and Don Draper. I also like to look backwards at experiences from the past or dream about the future. I find myself often drifting, like in self suggestion, dreaming about the future.</p><p>And I’m not alone — I know there are people that long to the little houses on the prairie or the time before the gold rush. The last US election showed how many people long to the time of the good old days. To the time of 50s diners, old Chevy’s and coal mines. To the time when we could “make sense” of the world. We combine our idealistic picture of the past to our wrongful impression of the future and think we can get it back somehow.</p><p>It’s hard to always live in the present. All the struggles we go through, we forget about them and we only remember what feels good when we look backwards. We’re wired that way in order to survive — we remember only parts of history.</p><p>But here comes the bitter truth:</p><blockquote>None of our futures is planned. And thinking you can plan the steps to reach your future is as helpful as trying to solve an algebra equation by chewing bubble gum.</blockquote><p>No matter if you believe in karma or that your choices make you, none of it will turn out as you think today. Surprises, failures and hidden dimensions always show up. Because while you’re making your plans, life everywhere else is also happening regardless of your plans.</p><p>We look at our own lives and think that there is a plan for how we should live it. We hear it in children’s stories and we read it in books about successful entrepreneurs. We learn it from the people we take as our mentors. We’re all fooled to think that other people are living their lives according to their plans. And we think that we used to be in the same roadblock as somebody else, but we somehow took a wrong turn and thus ended up where we are today. We think there was a choice and can compare to others — but correlation does not imply causation.</p><p>Celebrities endorse this survivorship bias, by talking about how they had their eyes on the ball since they were two years old. We never think about the bottom 99,999%. And we never stop to think, that even the ones we measure by success in one dimension, have had their failures in hundreds of other dimensions. And those failures ultimately formed them as a person and that one success.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/317/1*GPYRxPcEv3drkAsLOH7G1g.png" /><figcaption><a href="https://xkcd.com/1827/">https://xkcd.com/1827/</a></figcaption></figure><p>Looking at our own lives, nothing is as good a time as right now. Not only don’t you have a choice, but to live in the present, but also: life isn’t about what we want. In the words of the great Opus: <a href="https://youtu.be/2bljji5zdKU"><em>Life is life. Nanaananana.</em></a></p><p>I cherish the memory of barely fitting my legs inside the snow castle, in the freezing cold, while drinking hot chocolate with my brother. I’m nostalgic about being nostalgic back then. It sucks that I never got to experience it as a four year old, but I wouldn’t have thought about this without the experience. It wouldn’t have made me no matter how trivial a thing it is.</p><p>Be grateful for the things around you: rain, nature, your family and your memories. When you accept life as it is, you can take every challenge as it is: a challenge. And you will fail. Even when you succeed it will not look like anything that anyone else’s success has looked like. You are your own success and will be more interesting, do better and enjoy more when you aren’t trying to mimic somebody else.</p><p>It’s liberating knowing that there is no greater plan or meaning. It’s liberating knowing that when everything hits the fan and life throws you melons — you can still make your own genuine melonade.</p><p>And remember: you are only able to be nostalgic about the past when living in the now. In the past you had other future plans and perhaps other things to be nostalgic about.</p><p>Be nostalgic in the now. Use your dreams to plan the future, but don’t expect things to go according to plan – the planning is more important than the plan and remember: The best time of your life is now. As they say: life is what happens when you’re busy making plans.</p><p>— — —</p><p>This pieces was written as a self reflection and much like our minuscule odds of being on this planet and exactly like a broken pencil, this text is pointless.</p><p><strong>About the Author</strong></p><p>Fredrik Rönnlund is the founder of DigitalBooker and the Marketing Lizard at Vaadin. Follow him on <a href="https://twitter.com/freducom">Twitter</a>, <a href="http://linkedin.com/in/fredu">LinkedIn</a> or <a href="https://medium.com/@fredrikrnnlund">Medium</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=71877e1296b3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Branding 2.0: it’s all inside your primate brain]]></title>
            <link>https://medium.com/swlh/branding-2-0-its-all-inside-your-primate-brain-9d664dec541e?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d664dec541e</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[company-culture]]></category>
            <category><![CDATA[branding]]></category>
            <category><![CDATA[founders]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Fri, 05 May 2017 10:13:23 GMT</pubDate>
            <atom:updated>2017-05-05T10:13:23.099Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pu3laKRmT31mu2BK5oQ2pQ.jpeg" /></figure><p>Neuroscience teaches us that all of our fundamental decisions are made in the oldest and most primate part of our brain; the lizard🦎 brain. We also know that our brain makes decisions by answering two simple questions: 1) have I seen this before – yes or no? and 2) was it a pleasant experience – yes or no? If it looks scary or new, we’re immediately on our toes. Making these decisions quickly was elementary for our survival on the savannah and it made us fast at making decisions that saved our lives.</p><p><em>“Does that look like a lion? Yes. Was my previous encounter pleasant? No. RUN RUN RUN!”.</em></p><p>Even though we’d like to think our newer and more conscious “Homo Sapiens” brain is responsible for our decisions by standing to facts and reasons, in the most fundamental decisions its role is only to validate and fill in with facts to support the decision afterwards.</p><p>This is the core of what the late <a href="http://brandflight.com/thomas-gad/"><strong>Thomas Gad</strong></a> taught me about branding. Thomas was a pioneer in his field, having worked with <em>Richard Branson</em>, <em>Bill Gates</em> and companies from BMW to Spotify and Deutsche Bank to Microsoft. I had the pleasure of working with Thomas for a brief few months before he passed away, when we were building the new brand for the startup Vaadin, where I was responsible for marketing. Working with Thomas changed and expanded my view on branding forever.</p><p>But enough about brains and fluffy decision formulas. What does decision making have to do with branding?</p><p>Many people associate branding with the company logotype or the language used in copywriting, but it is much more than that. Thomas argued, that a brand is the DNA of the company and thus the most important part of any company. A brand defines what products the company builds, what channels it uses to communicate with its users and whom it hires. A brand is the company’s culture both internally and externally. A brand is the sales person’s tone of voice and the smell of the company’s products. A brand is essentially the answer to the question “Who are we and why do we exist?”.</p><p>A brand is your differing factor and it makes us as potential customers answer “yes” to the questions “does this look familiar and does it look like a pleasant experience”. That is essentially the purpose of a brand.</p><p>Every company has a brand. But too many companies don’t take their brand seriously and don’t try to define it – especially companies with a strong engineering cultures. But we still have a perception about those companies as well. Think about your local baker or the company next door or the company your dad used to work at – you have a perception about all of them. It’s impossible to not have a perception about your company and that is what branding is all about; <em>it is the management of people’s perceptions. </em>The perception of you – your brand.<em> </em>If you don’t control it, people will fill the void with their own perception about you. And most likely it won’t be what you think it is and most probably it will make them think of you like a line and run.</p><blockquote>A brand is the management of people’s perceptions. If you don’t manage it, people will make up their own perception.</blockquote><h3>Brands matter, you insensitive clog!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VfExOqaK8pxOfoS_6ljwgw.jpeg" /></figure><p>Imagine if Nike built cool shoes and didn’t care for its brand… Crazy right? But this is how most companies start out; they’re going to build the best fricking shoes on the planet! Shoes that increase your performance by 25% as shown by our own studies. Shoes that are the new Ubers for urban Googlers. Shoes that all cool people will buy and you should too!👊 Sound familiar? That’s because most startups start this way – their idea is not to create a brand but a product.</p><p>Without a brand and with a pure focus on product, Nike would mean completely different things for everyone. Its ads would feature one thing, its slogan would be another and its products would communicate a third thing. Shoes are however just the iceberg of what we think of when we see Nike – it is mainly about winning by getting started, or just doing it. And this brand makes Nike larger than just shoes. It is the image we associate with Michael Jordan wearing Nikes or Tiger Woods wearing a Nike cap.</p><p>Or put another way; the only thing separating Nike from Adidas and Reeboks is its brands. All these shoes are manufactured by the same under-age kids in Indonesia. They’re all built out of the same material and designed by equally talented designers. And if somebody invents something new, the others immediately copy it. Their brands are their only differing factors and the brands define who they are and why they exist.</p><p>But conveying this picture is not easy – one thing is defining a brand, another is successfully conveying it in a multidimensional world. Adidas for me for instance, stands for slavic youngsters driving mopeds – because that was their core audience in my home town and I didn’t interact with the brand in my youth otherwise. We all interpret brands differently but the better it’s managed and the more communicated in all of its mediums, the higher the probability it is perceived correctly. The question is, do you want your product to be about #winning or #slavicMopeds? 🇷🇺🏍</p><p>And it’s not just about shoes. Think about any other industry. From IT and open source libraries to manufacturing facilities in China and to ISO, CMMI and 6Sigma processes. Anyone with money can copy the same thing you’re doing. But what lies inside the minds of your customers – that’s impossible to copy. Thus, brands matter.</p><h3>The evolution of brands</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Le2kiHdZ0_UGbrNnaGphBg.jpeg" /></figure><p>All brands were local in the beginning. And most brands were the personal brands of their founders. They were stories told about the person that created the perception about them. We still have this today about our leaders and authority figures.</p><p>But then industrialization happened, and all of a sudden we could multiply our stories to the masses. We were able to deliver the same bread and the same cars from a conveyor belt with complete accuracy and the brand lived on – as a repeatable, predictable product. And we’ve all loved a predictable story, ever since our days around campfires, cooking mammoth. And the products echoed this story; the brand’s story was the product and it was the same product, no matter where we bought it. Thomas Gad called these brands <em>Transaction brands, </em>because they concentrated on completing a transaction.</p><p>Entire companies were working for a repeatable, predictable product experience. BigMacs became the embodiment of McDonalds’ brand, in the same way Beetles became the definition of Volkswagen’s. The core of these brands was the product itself and its success relied on repeatability. We could forgive a BigMac being different 4 times out of a hundred, but not 40. The product was the brand and the way to sell them followed this same schema; by shouting out the product and repeating the message on billboards, TV commercials and magazine ads. This was the birth of <a href="https://en.wikipedia.org/wiki/Don_Draper">Don Draper</a>. Everything from manufacturing to selling cars took the form of an industrialized conveyor belt. And our brains loved it. It was something we had seen before and it was a pleasant experience last time, so we bought ourselves another BigMac.</p><blockquote>But then came the birth of the Internet, and branding changed forever.</blockquote><p>The internet democratized information and all of the sudden, we didn’t need to rely on mass media telling us what to buy. We could listen to our friends (social media), we could read reviews (blogs) and we could challenge the companies directly ourselves. The channels changed and so did our expectations. We wanted to know more than just seeing the products – although most companies are still stuck in repeatable advertising.</p><blockquote>Top-down mass-sales-driven branding stopped working and it forced a more bi-directional brand.</blockquote><p>Thomas Gad called these new brands <em>Relation brands. </em>Relation brands didn’t rely on the product or its repeatability, but on the brand’s relationship with its customers. We wanted to have a predictable experience with the brand – not only a product.</p><p>When transaction brands talked about WHAT they delivered, relation brands talked about WHY they existed.</p><p>Look at for instance Tesla and how it isn’t a car manufacturer touting the “joy of driving” but wants to “accelerate the world’s transition to sustainable energy.” Elon Musk kicked off the Model 3 launch by talking about WHY they exist. And because of this WHY, they’re able to expand their portfolio without disappointing people that expect products.</p><p><a href="https://www.youtube.com/watch?v=sioZd3AxmnE">Simon Sinek</a> summarizes these layers of what inspiring stories are made of, by picturing it like an onion. While most companies start from the What layer and maybe talk about the How layer – the successful and inspiring companies talk about Why and move outwards.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/839/1*KArFHPMwMPihf7biz8FPqg.jpeg" /><figcaption>Transaction brands answer the question “What?”. Relation brands answer the question “Why?”</figcaption></figure><p>Most companies are still stuck in the industrialization stage of branding, and as Thomas noted – will probably stay there forever, or until a newer relation brand kills the older transaction brand.</p><p>Relation brands also moved the conversation from top-down to a dialogue. Think of Apple and how they leak out their launches weeks before an actual launch. They start a discussion with their users, because they value their users and want to surprise them positively. Think of tech companies like Amazon and Google going from Developer Marketing to two way Developer Relations. Or think of airlines trying to surprise their customers by offering personal gifts during Christmas. All these brands are having a conversation but they’re doing it on their terms – it is not left to the customers to decide what the brand is about. Relation brands touch the WHY and do it together with their users.</p><h3>How do I define my brand?</h3><p>Now that we know what a brand is and how a relation brand works, it’s time to define our brand. The minimum requirement is to have a written definition of your brand that everyone can understand. You would ideally come up with a brand motto that everyone in the company can recite and feel as their own. This will then help you hire, decide what your next product will be and whom you’re trying to target.</p><p>Thomas had a methodology for defining the brand through four dimensions. He pictured the brand affecting four dimensions and working through all dimensions to define its core. The dimensions were the functional dimension (the product), the social dimension (customers or users as a group), the mental dimension (people as individuals) and the spiritual dimension (the higher meaning). Thinking of your aspirational brand and how it touches each dimension is a clear framework for digging up the essentials. I warmly recommend Thomas’ book on <a href="https://smile.amazon.com/4-D-Branding-Cracking-Corporate-Network/dp/0273653687/ref=asap_bc?ie=UTF8">4D Branding</a> that talks more about how to walk through the process.</p><p>After having the brand, the more important part is execution and ensuring your story is the same in all customer touch points. A brand’s story has to be coherent, as you saw from the Nike example – but any good relationship also requires managed surprises. This is the age old easter-egg strategy to help your customers feel smart and excited, used heavily e.g. in the mobile gaming industry to create recurrence.</p><p>Another psychological strategy is re-framing. Re-frame all your communication so that you’re associated with something people have seen before and feel positive about, or at least that they don’t associate you with something they felt bad about. When Steve Job’s launched of the iPad; he didn’t launch it as a large phone or a crippled computer – he launched it as a view into the Internet. He reframed it as a completely new category of devices, which he then associated with what we knew from before. People don’t want new things, they want something familiar but better.</p><p>This in essence is what branding is about. Our primate brains love a predictable story, but they love it even more when it’s spiced with positive surprises. The story needs to convey that it’s good and familiar either directly or through a loop. And the story needs to visible in all you do and have a higher meaning than the product you sell. But at the end of the day; branding is setting the expectations and fulfilling them.</p><blockquote>“Branding is setting the expectations and fulfilling them” — Thomas Gad</blockquote><p>In my <a href="https://medium.com/swlh/how-to-not-hire-a-marketing-lizard-and-still-wing-it-506db9181fc8">previous blogpost</a> I introduced two frameworks with which to define your products and know your customers. In this blogpost I gave you a purpose to live. You figure out which ones is more important. 😉 This blogpost was written in memoriam of Thomas Gad — I wish I had known you better.</p><p><strong>About the Author</strong></p><p>Fredrik Rönnlund is the founder of DigitalBooker and the Marketing Lizard at Vaadin. Follow him on <a href="https://twitter.com/freducom">Twitter</a>, <a href="http://linkedin.com/in/fredu">LinkedIn</a> or <a href="https://medium.com/@fredrikrnnlund">Medium</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d664dec541e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/branding-2-0-its-all-inside-your-primate-brain-9d664dec541e">Branding 2.0: it’s all inside your primate brain🐒</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to not hire a marketing lizard and still wing it!]]></title>
            <link>https://medium.com/swlh/how-to-not-hire-a-marketing-lizard-and-still-wing-it-506db9181fc8?source=rss-3132bad6b2c1------2</link>
            <guid isPermaLink="false">https://medium.com/p/506db9181fc8</guid>
            <category><![CDATA[founders]]></category>
            <category><![CDATA[founder-stories]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Fredrik Rönnlund]]></dc:creator>
            <pubDate>Thu, 20 Apr 2017 23:17:20 GMT</pubDate>
            <atom:updated>2017-04-22T03:55:33.912Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to not hire a marketing lizard🦎 and still wing it!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-rXe2DbmFlbE1PSklv5oPQ.jpeg" /></figure><p>I founded my first tech startup 130 years ago. That’s in <em>entrepreneur-years</em>, of course, which we all know are measured by amount of pivots and toilet sessions spent reading Quora discussions. Not to mention lost “free time” doing whatever it is others are doing in their <a href="http://www.tv.com/shows/modern-family/"><em>Modern Family-esq</em></a> lives 👨‍👩‍👧‍👧.</p><p>What I’ve learned however will save you at least a decade of time. I’ll walk you through two fundamentals of creating a business: 1) how to differentiate your product and 2) understanding your customers. Or as the big boys put it; marketing. Under the hood, however, it’s only about three things; knowing your products, knowing your customers and knowing how to connect the two in an inspiring way 🙌. The rest comes naturally.</p><p>I’ll lay out the land for you through my own mistakes, with concrete steps and (best of all) Google Docs templates so you can get started while still sitting on the toilet. I’ll give you the bare🐻essentials! No magic, no jargon, no fluff – just stuff.</p><p>Through it all you’ll learn to walk by taking the following steps:</p><ol><li>Define your key values</li><li>Know your customers and #win🏆</li><li>Really know your customers</li><li>Be bigger than yourself</li><li>Pick your strategy and stick to it</li><li>How to run your marketing programs🏃</li></ol><p>I’ll spare you from all the MBA bulls**t and jump right into concrete steps. In this post we’ll concentrate on the two first steps – the fundamentals of doing anything — the stuff you should have done before starting the company but didn’t. In later posts we’ll dive into the next ones.</p><blockquote>Consider this the secret source code to your customers, market domination and everything.</blockquote><h3>Step 1: Define your key values</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*W8zsacBnKKvq92RpJcTgfQ.jpeg" /></figure><p>I started my first company at the age of 23, making websites for small companies. My business, however, was exactly the same as that of a thousand guys at the same time; we made websites for SMEs. There were probably more website companies in my city than there were any other kinds of companies during the .com. It was a bad idea. Period 💩. I hadn’t selected a niche or defined myself – I had no key values to offer.</p><blockquote>A company without a key value proposition is like a broken pencil… pointless!</blockquote><p>You’d be surprised how many companies get started with no <em>clearly defined </em>key values. As founders you always have a vision, but has your product been clearly defined?</p><p>This is where <a href="https://docs.google.com/document/d/1swnapn9FhpSeak0Pm5noSRqIPPdDKuJU2IIb9oMNhsA/edit#heading=h.o7qjf6nylbcx"><em>Key Value Propositions</em></a> come in. Key Value Propositions are a template I’ve put together based on business canvases and value definition papers. I’ve found the Russian-doll structure, where each consecutive step builds on the previous, works perfectly for a Dilbert-brain such as mine!</p><p>The Key Value Proposition starts from one step earlier than what the product centric engineers inside of us would like to start; the customers point of view. And finally it demands for customer testimonials or proof on how your product solves the customer’s problems. But most importantly, it teaches you about your product or service. It nails down your Key Value Proposition.</p><p>👉 <a href="https://docs.google.com/document/d/1swnapn9FhpSeak0Pm5noSRqIPPdDKuJU2IIb9oMNhsA/edit#">Key Value Proposition template in Google Docs</a>.</p><p>After Step 1, you’ll have a much clearer picture of your product. Next, let’s tackle the second part of the formula for success: your customers.</p><h3><strong>Step 2: Know your customer and #win🏆</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6y7tu85poAxcSjm5myzGYw.jpeg" /></figure><p>In my second startup, we started by making online booking systems for tennis courts. During the time, this was a new idea – there were tons of sports halls around the world and nobody dominated the booking market. But we hit a wall already before going global; we couldn’t get in front of the decision maker – or to be honest, we didn’t know who they were. Many of them were run by committees and it confused the hell out of us. They were bad hombres. Complete failure. Sad🌮! We used various strategies to try to meet with different people on the mini-corporate ladder, but ended up wasting a year of R&amp;D and sales time.</p><p>So we pivoted🙌. We turned into yoga and pilates bookings. Yoga studios usually had one decision maker and the decision maker was a yoga instructor themselves. Instant success! We were able to define the typical buyer; Harriet. Harriets were usually women in their thirties, their passion was primarily yoga and secondarily money. Their calendars were always booked and they didn’t have time to take care of billing or marketing or anything else. So they were struggling in a chicken-egg world of needing more business but having no time to do anything for it.</p><p>This became our secret sauce. We knew their pain points and were able to target that in website SEO. We knew their minds and were able to build the system so that they could get started in seconds and wouldn’t need to invest hours into seeing our benefits. And we hired a part-time yoga instructor to take them through the system introduction. And our new customers loved it. We grew from a $20k turnover company into a $200k turnover business in less than 12 months and continued climbing from there. Hockey sticks for everyone!💰 Only because we went from a fuzzy customer to one we understood.</p><p>The Harriets affected our R&amp;D, marketing, sales and support — much more than general customer feedback ever could. It made us prioritize the right things. Later on we had more and more stereotypical personas as we grew as a business. And we knew the personas by first name which was very powerful💪.</p><blockquote>Ultimately, no matter what marketing budget or sales channels you have, the one that knows the customer the best, wins.</blockquote><p>A <a href="https://docs.google.com/document/d/1aUezdcge79nDHKsAymYTomf61KEJgD767Q0F2D7Zn94/edit#heading=h.tyrngc78o669"><em>Buyer Persona</em></a><em> </em>description will let you define your stereotypical customer, already before you get started. It starts from the person’s demographics and goes on to their fears, values and everything that makes them tick. You’ll refine these along the way as you get under your customers skin.</p><p>👉 <a href="https://docs.google.com/document/d/1aUezdcge79nDHKsAymYTomf61KEJgD767Q0F2D7Zn94/edit#">Buyer Persona template in Google Docs</a>.</p><p>You’ll now have a clear view of your customers and your product – the next steps are to connect these with each other. Value Cards and Personas are like the foundations of your house – if they aren’t in shape, you won’t be able to build on top of them.</p><h3>Step N: n+1</h3><p>Next, you’ll need to connect your products to your customers in an inspiring way 🙌.</p><p>To do this you’ll need to understand your customer’s evaluation process, ie. the customer journey. After that we’ll look at how you can ride on the wave of something larger than your product, i.e. mega trends. And finally we’ll pick your marketing strategy and see how to do it operationally as marketing programs.</p><p>The general 6-step framework is the same for both small garage startups and large airfield-as-a-campus corporations.</p><p>But remember; this is the foundation – you can’t build a Marketing funnel, an Account Based Marketing strategy, a Social Media plan, an ad-words strategy, a content calendar or anything else unless you know the fundamentals. So take your time, a couple of hours should do, and do it well.</p><p>I’ll touch the following teaching steps of the framework in more depth in the next blogposts. Heart this post to be notified when it’s out, or follow me on <a href="https://twitter.com/freducom">Twitter</a>.</p><p><strong>About the Author</strong></p><p>Fredrik Rönnlund is the founder of DigitalBooker and the Marketing Lizard at Vaadin. Follow him on <a href="https://twitter.com/freducom">Twitter</a>, <a href="http://linkedin.com/in/fredu">LinkedIn</a> or <a href="https://medium.com/@fredrikrnnlund">Medium</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=506db9181fc8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/how-to-not-hire-a-marketing-lizard-and-still-wing-it-506db9181fc8">How to not hire a marketing lizard and still wing it!</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>