<?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 Gal Abramovitz on Medium]]></title>
        <description><![CDATA[Stories by Gal Abramovitz on Medium]]></description>
        <link>https://medium.com/@galabra?source=rss-a7d4afb90fa6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*iXx4qYbUyYcUTsHcbYH3oQ.jpeg</url>
            <title>Stories by Gal Abramovitz on Medium</title>
            <link>https://medium.com/@galabra?source=rss-a7d4afb90fa6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 09:09:55 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@galabra/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[Turning raw data into meaningful insights]]></title>
            <link>https://medium.com/akooda/turning-raw-data-into-meaningful-insights-c479c98fbd9a?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/c479c98fbd9a</guid>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[operations-management]]></category>
            <category><![CDATA[research-and-development]]></category>
            <category><![CDATA[performance-management]]></category>
            <category><![CDATA[digital-footprint]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Thu, 07 Apr 2022 11:52:53 GMT</pubDate>
            <atom:updated>2022-04-09T10:20:47.010Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7j0TFO3gPZd9FXgaF9wUhQ@2x.png" /></figure><h4><strong>How we successfully compared apples and oranges</strong></h4><p>Everyone loves chocolate. Everyone loves pizza.<br>But can you tell if Lombardi’s is better than Hershey’s?</p><p><a href="https://akooda.co/">Akooda</a> helps organizations understand what’s going on both internally, within the organization, and externally, with customers or prospects; manage resource allocation according to plan and gain insights to their customers’ health, sentiment, and interactions.</p><p>The challenge of making sense of raw data is a highly complex challenge that the most innovative tech companies face. Now imagine handling different types of data, from multiple sources: conversations, meetings, documents, support tickets, etc.</p><p>In this article I’d like to focus on how Akooda’s flexibility allows us to solve this challenge. More specifically, rather than going into the mathematical methods we use to process our data, I’d like to give a taste of the challenges we face as part of <strong>processing</strong> different data types and their solutions.</p><figure><img alt="Processing the raw data is crucial before we can analyze it and generate insights" src="https://cdn-images-1.medium.com/max/765/0*3VKvMGZmOKU7jFUB" /><figcaption>Processing the raw data is crucial before we can analyze it and generate insights</figcaption></figure><p>Processing the data is crucial for a couple of reasons:</p><ol><li>If we look at each data source separately, we won’t be able to identify patterns that emerge only when looking at the full picture. For example, projects in any organization are discussed in meetings and conversations; and are often described more thoroughly in files or tickets.<br>Connecting these digital breadcrumbs allows Akooda to look at clients, projects and processes as a whole.</li><li>Analyzing each data type separately requires different logic per data source. This architecture is potentially redundant (Confluence and Google docs documents have different metadata, but their contents are inherently similar); and requires significantly more resources to maintain and extend, as each upgrade affects only a single connector.<br>When the data follows a scheme regardless of its source, we can decouple third-party integrations from our business logic. This allows us to constantly improve our logic engine, while independently integrating more data sources quickly, efficiently and in a scalable fashion.</li></ol><h4>Semantically Defining our Model</h4><p>In the spirit of <a href="https://en.wikipedia.org/wiki/Domain-driven_design">Domain-Driven Design</a>, we had to agree on a name that describes all the data we fetch. Given that our business is understanding organizational communications, we decided to name our model “interaction”.</p><p>Now we had to define <em>what an interaction is</em> for each data type.<br>For some - like meetings - it was straightforward: a meeting equals an interaction.<br>However, other types - like conversations - were trickier: while defining a single message as an interaction makes perfect sense, we noticed a single message can be very short and have no context (e.g. “Good morning”) - and is then “outweighed” by a meeting (or a file, or a ticket).</p><p>For this reason, we decided to define a <em>conversation</em> as an interaction. This, in turn, raised the complicated challenge of splitting a stream of messages into conversations. But that’s a topic for another article 🙂</p><h4>Let’s get Unifying</h4><p>Naturally each interaction should have the basic fields, like its source (e.g. Slack), participants and time. Wait, did you say “time”? Let’s consider three examples for defining interactions’ “life spans” by their types:</p><ul><li>The times when a meeting starts and ends are static and well-defined. Nice!</li><li>Conversations start the moment a message is sent. However, we might need to update the end times for conversations when new messages are sent (and figure out when each conversation ends altogether).</li><li>A ticket can be created (i.e. start time) and immediately get moved to the backlog. After a few months someone can change its priority, then a few days later another person starts working on it — until a week later the ticket is resolved. Hence, it makes sense to define a ticket’s end time as its last update.</li></ul><p>Perhaps an interaction should also include some kind of “weight” (e.g. amount of messages, lengths of meetings, etc.). This gets us to the realm of defining all sorts of unintuitive comparisons, such as <em>how many messages equal a ticket status update</em>; But assuming we are given those “subjective” heuristics, we still need to address a fundamental issue: an interaction’s weight isn’t spread evenly throughout its life span.</p><h4><strong>Improving Our Model</strong></h4><p>That was when we figured an interaction’s weight and times are intertwined. We wanted to maintain the semantic meaning of interactions, which meant they might include a collection of singular events (e.g. messages, status updates).</p><p>Eventually we decided to represent this duality using a dedicated data structure we named “activity map”. Similarly to a log, its keys are timestamps and its values are the weights of each event.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8Ldz8thVKBLUIgIO" /><figcaption>An example for a ticket activity map</figcaption></figure><p>Using this activity map, we were able to represent the times in which the interaction was (and wasn’t) active. This allowed us - for example - to analyze and to visualize with better granularity the attention a project gets over time.</p><h4><strong>Conclusion</strong></h4><p>This was a simplified example of a product challenge we had to solve as part of our R&amp;D routine in Akooda. Obviously this hasn’t even scratched the surface of how complex it is to define a unified model. This specific model also takes into account topic and sentiment analysis, privacy restrictions, BI metrics and other specifications, both technical and product related.</p><p>If you find these challenges interesting and are passionate about solving them as we are, please feel free to <a href="https://akooda.co/company/the-team#JoinUs">contact us</a>! Being an ambitious startup, we’ll always hire people with positive vibes and great skills!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c479c98fbd9a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/akooda/turning-raw-data-into-meaningful-insights-c479c98fbd9a">Turning raw data into meaningful insights</a> was originally published in <a href="https://medium.com/akooda">Akooda</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Buddhist Who Went Serverless]]></title>
            <link>https://galabra.medium.com/the-buddhist-who-went-serverless-fab014fdcb60?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/fab014fdcb60</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[buddhism]]></category>
            <category><![CDATA[serverless]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Thu, 28 Jan 2021 17:12:42 GMT</pubDate>
            <atom:updated>2021-01-28T17:12:42.815Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ck3gxKjuYLNQUCNZGuP0ug.jpeg" /><figcaption>Photo by Iván Tejero on Unsplash</figcaption></figure><p>One of Buddhism’s key principles is <a href="https://zenhabits.net/attachments/">letting go</a> of attachments. There’s a long chain of logical reasoning behind it, but essentially it means life would be simpler and happier if we become independent of the <em>stuff</em> we’re used to think of as necessary. That includes physical property — such as a car or a house — and extends to abstract concepts like fortune, nationality, etc.</p><p>However, in software development, dependencies are part of our everyday life. Literally every piece of code relies on another compiler, framework or library; and while using these is allegedly a professional matter, the experience is often coupled with feelings (<a href="https://wiki.c2.com/?WhyWeHateLisp">hate</a>, <a href="https://www.youtube.com/watch?v=84dZddrmZu8&amp;ab_channel=SourceDecoded">love</a> and <a href="https://www.redbubble.com/i/poster/I-code-in-Java-and-I-m-proud-of-it-by-sans-design/51297587.LVTDI">pride</a> are some common examples).</p><h3>The Stuff We Can Touch</h3><p>Up until 20 years ago, servers were purely physical machines (with lights, fans, dust and the whole shebang). Although outdated, this approach ensured maximal security and control.</p><p>As technology progressed, cloud computing services were introduced. Nowadays maintenance of physical servers (and their rooms) sounds like ancient history. The flexibility that cloud services offer made them the mainstream technology in modern research and development.</p><p>The transition from owning a physical object into subscription-based service isn’t unique to the development industry: even the idea of owning a DVD and a bunch of disks makes <a href="https://www.forbes.com/sites/blakemorgan/2019/01/02/nownership-no-problem-an-updated-look-at-why-millennials-value-experiences-over-owning-things/?sh=320eaaaf522f">millennials</a> shift in their seats, as YouTube, Netflix and other streaming services have become our main sources of video content.</p><p>Generation X and their parents <a href="http://www.aabri.com/manuscripts/08050.pdf"><em>needed</em> to own property</a>. In the post-WW2 reality they had to be certain some things will remain theirs, no matter what catastrophes the future holds. Unfortunately, the remaining of this approach still dictates the way we invest money and manage our code.</p><h3>Letting Go</h3><p>Alongside “classic” cloud server services (i.e. Platform as a Service, <a href="https://azure.microsoft.com/en-us/overview/what-is-paas/">PaaS</a>), which allow running containers of code, there’s another model that supports scalable applications out of the box: serverless computing (i.e. Functions as a Service, <a href="https://en.wikipedia.org/wiki/Function_as_a_service">FaaS</a>).</p><p>In a nutshell, according to this pattern, the server’s logic is split to independent, <a href="https://stackoverflow.com/a/844548/3103891">stateless</a> functions that are hosted by a third-party service. The provider guarantees the availability of the functions’ endpoints and allocates resources dynamically for each run. Hence, scale-related issues and virtual machines’ configurations become transparent to developers.</p><p>In other words, after we stopped tracking <em>where</em> our code runs, we now also give up knowing<em> how </em>it runs. Just like a Buddhist monk who gave up his property, we’re now free of this burden and can focus on the essence: our business logic.</p><p>In both worlds letting go requires a certain degree of confidence, and indeed there are some <a href="https://hackernoon.com/what-is-serverless-architecture-what-are-its-pros-and-cons-cc4b804022e9">drawbacks</a> to using serverless architecture in your application. Nevertheless, it sounds like a choice we need to carefully study and consider. Even if we don’t <a href="https://www.serverless.com/blog/6-things-to-know-before-migrating-an-existing-service-to-serverless">migrate our whole backend services</a>, it’s still a strong tool to have in our system design arsenal.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fab014fdcb60" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Empathy: The Unsung Hero of Development]]></title>
            <link>https://galabra.medium.com/empathy-the-unsung-hero-of-development-a29114fea20?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a29114fea20</guid>
            <category><![CDATA[teamwork]]></category>
            <category><![CDATA[code]]></category>
            <category><![CDATA[clean-code]]></category>
            <category><![CDATA[coding]]></category>
            <category><![CDATA[development]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Thu, 21 Jan 2021 08:58:50 GMT</pubDate>
            <atom:updated>2021-01-21T08:58:50.919Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0UbWjhrwNh3uMMpW9zVSbA.jpeg" /><figcaption>Photo by Kristen P on Unsplash</figcaption></figure><p>Who do we write code for? Do we satisfy the product managers? Do we write features for the company’s clients, who pay our salaries? Perhaps for the CEO?</p><p>At least when considering code, we - developers - are our own most loyal clients. Hence, the code we write is primarily for future us.</p><p>A code is <a href="https://nypost.com/2019/07/29/this-scary-pot-of-soup-has-been-cooking-for-45-years/">like a tradition</a>, passed among generations of developers. It’s continuously evolving, but the basis always consists of older (if not ancient) <a href="https://en.wikipedia.org/wiki/Stratum">strata</a>. Unlike every other person - from PMs, through senior management and up to end users - only developers read code; and it actually takes a significant chunk of our time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/475/0*VZDJdUToSvmpdMl8" /><figcaption>We spend much more time reading code than writing it</figcaption></figure><p>The common (software) engineer is portrayed as a socially awkward nerd, where the <em>good</em> engineers are also smart. The thing is that intelligent people often come up with either complex verbose solutions, or they use very specific terms that require deep knowledge of their domain.</p><p>Every developer had come across code that requires familiarity with esoteric patterns, or just notorious single-line <a href="http://gist.github.com/galabra/57bea6bc88dbb48696ce52f9e5c2e1bc">code snippets</a>. How did it make you feel about yourself, or about the person who wrote it? I hated them (and was pretty vocal about it - you can ask my colleagues).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/662/0*EuEN003PXGt5yk_T" /></figure><h3>Being Empathetic Is Key</h3><p>Extraordinary developers see the code they write through the eyes of others. The tactical tasks that we implement are just a part of the bigger picture - one day someone will extend our code, or will use it as a module in another complicated feature.</p><p>Legacy code that is hard to read significantly slows down companies, or worse - it halts people from changing stuff, just because they&#39;re afraid of the unknown. Hence, regardless of algorithmic skills or agility, if you&#39;re not conscious about others you often do more harm than good.</p><p><a href="http://cleancoder.com/">Uncle Bob</a> claims that &quot;Clean code always looks like it was written by someone who cares&quot; and I&#39;m certain that <em>cares</em> doesn&#39;t only imply one&#39;s high standards, but also their caring for both current and future colleagues.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a29114fea20" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Establish your narrative to outperform job interviews]]></title>
            <link>https://medium.com/akooda/establish-your-narrative-to-outperform-job-interviews-a8b4ce65941a?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a8b4ce65941a</guid>
            <category><![CDATA[job-interview-tips]]></category>
            <category><![CDATA[job-search]]></category>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[storytelling]]></category>
            <category><![CDATA[storytelling-for-business]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Thu, 31 Dec 2020 13:57:15 GMT</pubDate>
            <atom:updated>2021-01-01T06:49:59.343Z</atom:updated>
            <content:encoded><![CDATA[<h3>Establish Your Narrative to Outperform Job Interviews</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LJfTfFaAh-sL__iG" /><figcaption>Photo by <a href="https://unsplash.com/@fiidesign?utm_source=medium&amp;utm_medium=referral">Federica Campanaro</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Imagine going on a blind date. You’re about to meet the person that might be your life partner - perhaps the parent of your children, the one you will go both to bed and to family dinners with for the rest of your life - but still, the first interaction you’re going to have will probably be comprised of the most predictable and boring set of questions: where did you grow up, what did you study, etc.</p><p>You better find a way to spice up this conversation - otherwise you’ll spend at least a few more evenings answering these predictable questions over and over.</p><p>The exact same thing applies to job interviews. After being on both sides of the interview table, I can definitely say that my successful interactions were the ones that were - simply put - fun.</p><h3>The path to success</h3><p>It’s widely accepted that in order to pass an interview successfully, one must have an impressive résumé, spend at least a few days preparing and then present knowledgeable answers that would dazzle the interviewer.</p><p>I humbly claim that these guidelines are distracting, even misleading, as they cloud a simple truth: the content of the conversation is just the setting for the tone, the energy and the excitement that extraordinary candidates generate.</p><p>Obviously you have to put effort into studying and improving your professional skills, but that goes without saying. If you come to an interview professionally unprepared, the conversation will probably get awkward regardless of what you say, or how you say it.</p><p>However knowledge and experience simply don’t guarantee a successful interview. One must wrap them with an intriguing story and complement it with the right energy.</p><h3>What’s your story?</h3><p>If interview questions had a common ancestor, it would be “tell me about yourself”. The fact that it’s as open-ended as it gets makes this question tricky and even complicated to approach. However if you prepare for it in advance, this can become an amazing opportunity to get the other person engaged and set the tone for the interview.</p><blockquote>A good story cannot be devised, it has to be distilled (Raymond Chandler)</blockquote><p>Crafting your story is the single most important thing when applying for a job. Prior to this point in time you had plenty of experiences, so don’t worry about not being “experienced” or “interesting” enough. Rather, the important thing is to tailor the best fitting narrative to the position (and organisation) you apply to. Once you’ve constructed your narrative’s wireframe, you can come up with ways to further exemplify it.</p><p>Hence, the process’ baseline is defining what kind of a person you would like to present yourself as. Please note that I’m not referring to what you did or achieved, but to the personal qualities you bring to the table.</p><p>For example, if you apply for an entry-level position, you should probably consider portraying yourself as a curious person. This can be further demonstrated by different hobbies you have and take seriously: cooking, traveling, wood crafting, gardening - you name it! It’s not as important what you like to do, but how excited you are about a topic, how eager you are to expand your knowledge of it and what you do in order to achieve better proficiency.</p><p>One of my best friends decided to change his career path and become a software developer. He had absolutely no relevant prior career experience, which made him speak unconfidently about himself. This feeling is commonly mistaken with dullness; and indeed he didn’t even get through the screening interviews. After reflecting on his infelicitous experience, he decided to try and talk about his academic thesis, on the physical properties of human blood vessels. As exotic as this subject sounds, once he demonstrated his passion and curiosity, his tone became vibrant - and things just clicked! Regardless of the topic itself, it characterised him as a great candidate and landed him a great role in one of the world’s most innovative tech companies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dMiXuuD6SQCEQ0Uy91ILGQ.jpeg" /></figure><h3>Putting it all together</h3><p>Once we decided what’s the main story we want to present ourselves with, we need to structure how are we actually going to tell it.</p><p>In my opinion, <em>this</em> is the tricky part. There are infinite ways to tell a story; and respectively there are many decisions to consider:</p><ul><li>Should the narrative be linear (i.e. chronologically ordered) or non-linear?</li><li>If we take the chronological approach, <em>when</em> do we start?</li><li>What stories best demonstrate your values? Maybe some of them are still a little less relevant?</li><li>Where do you prefer to skip a story, and just state the bottom line explicitly? (e.g. “I love working as a part of a team”)</li></ul><p>The process of optimising your narrative’s structure requires multiple trial and error iterations; And the key to improvement is listening to other people’s feedbacks. This allows to continuously tweak all the parameters above, so you end up leaving an impression that’s as close as possible to the description you’ve initially defined.</p><p>For example, the most basic topic to tackle is probably the first thing you tell about yourself (i.e. your story’s prologue). The interviewer wants to know <a href="https://www.youtube.com/watch?v=PJKYqLP6MRE">who you are</a> rather than just what you did. You’re far more than a features list (e.g. “Martin, a 45 years old financial advisor from Chicago”) and might as well use your special and unique experiences to intrigue your interviewer right from the beginning of the conversation (e.g “after getting tired from the city life in Chicago, I moved my family to Missouri, where my wife and I made a serious of investments in local businesses, including a casino and a funeral home”).</p><p>One interviewee that I’ll never forget literally kicked off the interview by telling me how he almost lost his ability to walk, after going through a parachuting accident. He was hospitalized, spent a couple of months in a coma; and only after a year-long physiotherapy treatment he was able to stand on his feet. Nevertheless he insisted on going back to military service and is now a ranked officer in one of the IDF’s elite units. It has been almost 10 years since I interviewed this guy, who was then a young ambitious soldier, but I’ll never forget his passion nor his story.</p><h3>Epilogue</h3><p>Before applying to your next position, spend some time reflecting about how you want to introduce yourself. Your interviewer would like to hear about <em>you </em>-<em> </em>not only about the things you did - and telling your story correctly can definitely set the tone in your favor.</p><p>Consider your interviewer as your potentially best friend, with whom you might spend the next few years. They most certainly have peculiar stories and share some weird interests too, so try to think of this preliminary interaction as a lively conversation, rather than a test.</p><p>Breath, smile, laugh - let them know who you really are. If all goes well, you both should have a great time. Good luck!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a8b4ce65941a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/akooda/establish-your-narrative-to-outperform-job-interviews-a8b4ce65941a">Establish your narrative to outperform job interviews</a> was originally published in <a href="https://medium.com/akooda">Akooda</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Most Generic Algorithm for Removing Outliers]]></title>
            <link>https://galabra.medium.com/the-most-generic-algorithm-for-removing-outliers-362e064f8a4c?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/362e064f8a4c</guid>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[data-visualization]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Wed, 27 Nov 2019 09:59:22 GMT</pubDate>
            <atom:updated>2019-11-27T09:59:22.347Z</atom:updated>
            <content:encoded><![CDATA[<h4>A generic algorithm for cleaning outliers from a one-dimensional set of numbers</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*agwk1Xm6FPWIl2Ce" /><figcaption>Photo by <a href="https://unsplash.com/@will_myers?utm_source=medium&amp;utm_medium=referral">Will Myers</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Working on a large dataset is always tricky. It’s often impossible to manually examine the data and determine whether it even makes sense.<br>Personally, I’m a big fan of data visualisation, that allows intuitive and swift view of otherwise incomprehensible data. The challenges and dilemmas of the visualisation process is well-documented and dealt with; Hence I encourage you to explore and experiment this topic, so you can add it to your data science toolbox and use it as a base for any data examination in the future.</p><p>Getting back to our point, no matter if you use scatter plots, histograms or heat maps, <strong>outliers</strong> will probably make your axes too wide and will make your chart out of proportions. To make matters worse, they might sabotage your understanding of the dataset (by distorting the mean, median, etc.).<br>But what even are these little demons?</p><blockquote>From <a href="https://en.wikipedia.org/wiki/Outlier">Wikipedia</a>:<br>“An <strong>outlier</strong> is a data point that differs significantly from other observations”</blockquote><p>For example, consider the following ascending set:<br>(1, 2, 2, 5, <strong>115</strong>)<br>The outlier in this first set is clearly 115, as it’s much bigger than the other numbers in the set. While it doesn’t change the median (2), it does change the mean from 2.5 to 25, which is conveniently 10 times bigger.</p><h3>Dealing With Outliers</h3><p>It’s worth mentioning that outliers aren’t always mistakes. They can be an excellent indicator for measurement errors, or they can just point an extreme result, which is much better/worse than the others.</p><p>However, the inspiration for this article was a work I did on a very diverse dataset (the columns/dimensions significantly differed from each other in scale) that involved creating normal distributions charts for each dimension’s histogram. In this case, the outliers often distorted the variance<em> </em>values and made the line charts too wide.</p><p>Hence, I was interested in detecting outliers and removing them from the dataset.<br>The <em>good</em> news is that in most cases outliers are easy to spot visually.<br>The <em>bad</em> news is that this isn’t scalable.</p><h3>Common Methods For Detecting Outliers</h3><p>As mentioned above, outliers are easy to point out but pretty elusive to define.<br>Most of the algorithms that handle with outliers require a threshold that defines the minimal “distance” of outliers from the rest of the set (<a href="https://en.wikipedia.org/wiki/Standard_score"><em>Z-score</em></a>).<br>This makes sense, but doesn’t allow a generic solution to a wide range of data in different scales. For example, in order to find the outliers in the two following sets, one will need to define two different thresholds:</p><ol><li>(1, 2, 3, 4, <strong>60</strong>)</li><li>(100, 150, 160, <strong>25000</strong>)</li></ol><p>Another approach is to determine the threshold according to the distribution of the data (<a href="https://en.wikipedia.org/wiki/Interquartile_range"><em>IQR score</em></a>). Using this method is very efficient; although a bit rough, as it doesn’t reflect the distance of data-points outside the range it handles (between the 75th and the 25th percentiles) and could identify the points that are closer to the median as outliers, while there are other points much farther away.<br>For example, in the set (1, 2, 3, 4, 5, 6, 100) the interquartile range is (2, 3, 4, 5, 6), hence both 1 and 100 are outside it, even though 1 isn’t an outlier and 100 indeed is.</p><h3>The Generic Algorithm For Detecting Outliers</h3><p>Trying to avoid using the thresholds mentioned above, I came up with a greedy algorithm that - on each iteration - identifies the farthest point(s) from its neighbour. In other words, on every step it tries to eliminate as much blank space as possible.</p><p>But… I cheated. This algorithm actually does require one threshold: the percentage of original data-points that must remain in the output set.<br>The lower this value is, the more outliers will be removed, while increasing the risk of removing more non-outliers points as well.<br>I just figured that this threshold is easier to use, as it’s mostly a definition which is externally provided, rather than requiring intimate acquaintance with the data’s specific properties.</p><p>The algorithm is as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U6zr1SWcReTtp2aTxndo9g.png" /></figure><p>As explained, it generically removes the outliers, even when they’re grouped (“far” from the main data group) and portray a batch of relevant data by themselves.</p><p>I’d love to hear comments, critics and use cases that can benefit from the use of this algorithm.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=362e064f8a4c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[2019’s KPI Report is Out!]]></title>
            <link>https://galabra.medium.com/2019s-kpi-report-is-out-4db4cdb84157?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/4db4cdb84157</guid>
            <category><![CDATA[benchmarketing]]></category>
            <category><![CDATA[retail]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[reporting]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Wed, 07 Aug 2019 18:43:33 GMT</pubDate>
            <atom:updated>2019-08-07T18:43:33.627Z</atom:updated>
            <content:encoded><![CDATA[<p>Dublin-based Digital Marketing agency, Wolfgang Digital, has published their KPI report, which contains insights about the hot trends and their strategic meanings in the digital world of retail, e-commerce and travel industries.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Nz7Ypr__yfduXEmF" /><figcaption>Photo by <a href="https://unsplash.com/@srd844?utm_source=medium&amp;utm_medium=referral">Stephen Dawson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Their report highlights the importance of measuring and comparing your brand’s performance to your industry and competitors.<br>But measuring is not enough. It should be done correctly and methodically, leading the company’s decision making process.</p><p>Among the KPIs in the report you can find benchmarks of traffic sources, revenue sources, conversion rates, engagement rates, etc.</p><p><a href="https://www.benchmarketing.io/?utm=medium">Benchmarketing</a> makes these KPIs accessible in real time, so you can stay on top of your industry.</p><p>To view Wolfgang Digital’s KPI report, <a href="https://www.wolfgangdigital.com/kpi-2019/?utm=benchmarketing">click here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4db4cdb84157" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Make Your App Load Quicker]]></title>
            <link>https://galabra.medium.com/make-your-app-load-quicker-7d903c17be9f?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/7d903c17be9f</guid>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[optimization]]></category>
            <category><![CDATA[css]]></category>
            <category><![CDATA[react]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Thu, 25 Oct 2018 18:00:13 GMT</pubDate>
            <atom:updated>2019-08-10T09:57:23.138Z</atom:updated>
            <content:encoded><![CDATA[<h3>Quicker loading with ReactJS</h3><p>After a couple of days of continuous research and studying about page-loading time optimizations, I wanted to sum it up - so the next Front-End developer won’t have to go through all the fuss of reading like I did (or at least they’ll be quicker).</p><p>I’m working on <a href="https://www.benchmarketing.io/">Benchmarketing.io</a>, which is an (amazing) tool for benchmarking, hand-tailored for e-commerce marketers and managers. The app itself is a single-page application (SPA), which means there’s plenty of room for loading-time optimizations, as explained <strong>beautifully</strong> in <a href="https://serverless-stack.com/chapters/code-splitting-in-create-react-app.html">this article</a>.</p><h3>Figuring what’s going on</h3><p>First of all, Google’s <a href="https://developers.google.com/speed/pagespeed/insights/">PageSpeed Insights</a> provided me with a basic ground for my research. It has all the terminology and some basic explanations and action items, which are perfect to get you started.</p><p>In my case, the issues that held me back the most were the <em>blocking resources</em> of both JavaScript and CSS.</p><p>I’m using Create React App (CRA) to compile my app, so I started with 2 relatively big files that were created by <em>react-scripts build</em>: one that merges my whole CSS files and the other one merges all the JS files.<br>Obviously these are pretty nasty files to work with, as they include not only my app’s design and scripts, but also the whole Bootstrap’s CSS, along with the node modules I’ve used.</p><p>Because the browser <a href="https://developers.google.com/web/fundamentals/performance/critical-rendering-path/render-blocking-css">blocks its rendering process</a> until it finishes downloading all the required files (including the huge CSS file), the merged file posed a real problem.</p><h3>Identifying the challenge(s)</h3><p>After some more research, my insights boiled down to 2 main goals:</p><ol><li>The first challenge was to split the gigantic files into smaller, manageable chunks (which is actually how they’re called 🙃).</li><li>The second challenge was to actually make React load these chunks of JS and CSS files only when they’re relevant; Hence sparing their loading time when the app initially loads.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rtQBQiFPGPmTU-5J" /><figcaption>Breaking into pieces (by <a href="https://unsplash.com/@markusspiske?utm_source=medium&amp;utm_medium=referral">Markus Spiske</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a>)</figcaption></figure><h3>Splitting</h3><p>The mechanism that actually generates the main files is CRA’s <em>build</em> function.<br>It magically goes through the app’s source code and identifies what JS files are imported into each component - then it merges the files by the components they’re imported into.</p><p>In my case, because I’m working on a SPA, the code was written in a way that just loads <em>everything </em>on initialization. It still took only ~3 seconds to load; However, according to many articles (such as <a href="https://www.machmetrics.com/speed-blog/average-page-load-times-websites-2018/">this one</a>), users will leave when an app takes over 3 seconds to load. And here we are.</p><p>So the first step was to create asynchronous components - just as explained <a href="https://serverless-stack.com/chapters/code-splitting-in-create-react-app.html">here</a>. This alone made a big impact, as the app no longer had to load at once.</p><p>The bigger issue was handling the CSS files: my app looked great, but still took long time to load, because its design was scattered across a few CSS files - which, in turn, made them relevant for all the components and therefore resulted in the huge merged CSS file.<br>I needed to clean the CSS design mess and re-organize it in separate files, by its relevance: each “page” in a different CSS file.</p><p>Finally I wanted to make the initial user experience as best as possible, so I separated the design of the elements that were above-the-fold from the rest of the landing page’s elements.</p><p>This was actually most of the work that needed to be done. The CSS and JS chunks were imported only where they’re relevant, resulting in a much more modular app. The only issue that was left was loading the rest of the landing page’s CSS (<strong>after</strong> the above-the-fold part is loaded).</p><h3>Loading code only when it’s relevant</h3><p>I read A LOT about the possibilities of doing this, but eventually I came with a very simple work-around that work really nice.</p><p>Let’s say we have a landing page component (that takes a while to load), with the following structure:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1e3f9b6584dbf3e20c5a1ccae8fb3b97/href">https://medium.com/media/1e3f9b6584dbf3e20c5a1ccae8fb3b97/href</a></iframe><p>After splitting the main CSS file, we’ll have two files: one for the initial load (above-the-fold content) and the other for the rest of the page’s content.</p><p>Then all we need to do is lazy-load the bottom part of the landing page. In my opinion the simplest method to do so is by using React’s built-in <a href="https://reactjs.org/blog/2018/10/23/react-v-16-6.html#reactlazy-code-splitting-with-suspense">React’s code splitting</a> (added in v16.6.0).<br>Then the whole code, including the CSS, won’t block the whole page from rendering, but will load after the above-the-fold section is presented to the satisfied user.</p><h3>Bonus: How to configure React-scripts to build the CSS file inside the public directory</h3><p>As we probably <em>will</em> edit and update our CSS files, if we use Sass we’ll need to build at compilation time those that are in the <em>public</em> directory as well.</p><p>Luckily, we only need to add to our <em>package.json</em> file’s “build-css” and “watch-css” properties the relevant parts. For example, using <em>node-sass-chokidar</em>:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1598cac33c342cf34937db2c2b05639d/href">https://medium.com/media/1598cac33c342cf34937db2c2b05639d/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7d903c17be9f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Web Scraping 101 (Using Selenium for Java)]]></title>
            <link>https://galabra.medium.com/web-scraping-101-using-selenium-for-java-9d37c52ce7a2?source=rss-a7d4afb90fa6------2</link>
            <guid isPermaLink="false">https://medium.com/p/9d37c52ce7a2</guid>
            <category><![CDATA[scraping]]></category>
            <category><![CDATA[selenium]]></category>
            <category><![CDATA[web-scraping]]></category>
            <category><![CDATA[selenium-webdriver]]></category>
            <category><![CDATA[java]]></category>
            <dc:creator><![CDATA[Gal Abramovitz]]></dc:creator>
            <pubDate>Mon, 13 Aug 2018 01:24:34 GMT</pubDate>
            <atom:updated>2018-08-13T10:05:57.310Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*92J7XHuDmcDeCjvl" /><figcaption>Photo by <a href="https://unsplash.com/@rawpixel?utm_source=medium&amp;utm_medium=referral">rawpixel</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Web Scraping is one of the most useful skills in today’s digital world.<br>Basically it takes web-browsing to the next level, by automatizing everyday actions, such as opening URLs, reading text and data, clicking links, etc.</p><p>Even though web scraping is relatively easy to learn and execute, it’s a powerful tool that you can use to <strong>collect existing data from websites</strong>, then easily manipulate, analyze and store it to your liking; or to <strong>automate workflows</strong> in web-based platforms.</p><p>In this tutorial I’ll show you step-by-step:</p><ol><li>How to set up Selenium in IntelliJ platform.</li><li>How to build a basic web scraper, that can read data from a webpage.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/200/1*46jMgMa0_rnpNahKwfGGEw.png" /></figure><h3>First things first, we’ll set up Selenium in IntelliJ environment:</h3><p><em>* The stages are followed by matching screenshots</em></p><ol><li>I found it easiest to use <a href="https://www.seleniumhq.org/download/">Selenium Standalone Server</a>. Download the JAR file in the <a href="https://www.seleniumhq.org/download/">given link</a>.</li><li>Open IntelliJ IDEA and create a new project.</li><li>Right-click one of your project’s directories and click <strong>Open Module Settings</strong>.</li><li>In the <strong>Modules </strong>section click the <strong>Dependencies</strong> tab, then click the “<strong>+”</strong> button and choose <strong>JARs or directories…</strong></li><li>Choose the JAR file you’ve downloaded in stage 1, then click <strong>OK</strong>.</li><li>The <strong>External Libraries</strong> should now contain the JAR file.</li><li><a href="https://github.com/mozilla/geckodriver/releases">Download <strong>Geckodriver</strong></a> (choose the file according to your operating system) and put the executable file in the project’s folder.</li></ol><p><strong><em>That’s it - you’re all set!<br></em></strong><em>You can take a look at the self-explanatory screenshots and move on to the second part of the tutorial.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NvNA0d4H0spUqHDhE9lIcw.png" /><figcaption>Stage 3: click <strong>Open Module Settings</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p6Qw7FZYxrrGqXqLQ_JnqA.png" /><figcaption>Stage 4: in the <strong>Modules </strong>section click the <strong>Dependencies</strong> tab, then click the “<strong>+”</strong> button and choose <strong>JARs or directories…</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/772/1*IrKq_StXHvuSAnETXgvNVA.png" /><figcaption>Stage 6: The <strong>External Libraries</strong> should now contain the JAR file</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/726/1*PyjDH6JZCgYoLc7mzlXjIw.png" /><figcaption>Stage 7: your project’s directory should look like this</figcaption></figure><h3>How to build a basic web scraper</h3><p>After we’ve set up our working environment, let’s use Selenium to build a basic web scraper.<br>We’ll use the HTML structure of a webpage and read specific data (we’ll actually take advantage of the CSS code, but the principles are the same).</p><h4>Preparing for web scraping</h4><p>Before we actually start to scrape, we need to understand what we’re looking for. This part will be much easier for you if you can read HTML fluently.</p><p>In this tutorial we’ll scrape my website and extract the list of languages I use. It might seem a tough task at first, but a glance into the paragraph’s HTML code will reveal a surprising structure:</p><pre>&lt;div class=&quot;line&quot;&gt;<br>    In my work and personal projects I use <br>    &lt;p&gt;Java&lt;/p&gt;, <br>    &lt;p&gt;JavaScript&lt;/p&gt;, <br>    &lt;p&gt;jQuery&lt;/p&gt;, <br>    &lt;p&gt;HTML&lt;/p&gt; and <br>    &lt;p&gt;CSS&lt;/p&gt;. My university projects also use <br>    &lt;p&gt;C&lt;/p&gt;, <br>    &lt;p&gt;C++&lt;/p&gt;, <br>    &lt;p&gt;Assembly&lt;/p&gt;, <br>    &lt;p&gt;SQL&lt;/p&gt; and <br>    &lt;p&gt;Python&lt;/p&gt;.&lt;br&gt;<br>    As I&#39;m highly aware to the elegance of my products, I keep learning and pushing myself towards cleaner code and more beautiful UI.<br>&lt;/div&gt;</pre><p>You can tell immediately that each language name is wrapped in a separate&lt;p&gt; tag. We’ll use this fact later in our scraper’s code.</p><p>I obviously already know my own website, but unravelling each page’s code is the first challenge that you’ll need to solve. In most cases there is an obvious logic to the HTML structure, that we can take advantage of.<br>Understand the HTML logic - and your life would get much easier.</p><p>The next step would be to plan the scraping workflow. I suggest you go through your workflow manually a couple of times before writing your web scraper.</p><p>In this basic example, this is the workflow we’re going to implement:</p><ol><li>Open a new browser window.</li><li>Navigate to <a href="http://www.galabra.co.il/">http://www.galabra.co.il</a>.</li><li>Click the menu option “ABOUT”.</li><li>Copy the languages names, according to the logic we’ve described.</li></ol><h4>Selenium Building Blocks</h4><p>I like to think about a (basic) web scraper as a synchronous program that mainly uses two objects instances:</p><ul><li>A <a href="https://www.seleniumhq.org/projects/webdriver/"><strong>WebDriver</strong></a> instance, which is our code’s connection to the browser’s window: it allows to read its code and simulate user’s actions.<br>Each WebDriver instance corresponds to an individual browser window.<br><em>(in this tutorial I’ll use a </em><strong><em>FirefoxDriver</em></strong><em>, just because I usually use Chrome and I like to have a separate browser for my web scraping applications)</em></li><li>A <a href="https://seleniumhq.github.io/selenium/docs/api/java/org/openqa/selenium/support/ui/WebDriverWait.html"><strong>WebDriverWait</strong></a> instance, which allows us to synchronize our actions (e.g. click a button only <em>after</em> it loads and is clickable).</li></ul><p>It might seem a bit abstract, but don’t worry - in a moment we’ll use both of them in context.</p><h4>Finally - let’s write some code!</h4><p><em>(The full code is provided </em><a href="https://github.com/galabra/WebScraping101"><em>in this repository</em></a><em>)</em></p><ol><li>We’ll start by initializing a <strong>FirefoxDriver </strong>and a<strong> WebDriverWait</strong>:<br>FirefoxDriver driver = new FirefoxDriver();<br>WebDriverWait wait = new WebDriverWait(driver, 30);<br>Note that <strong>WebDriverWait </strong>remains the same, regardless of the browser you choose, and that it’s connected to the FirefoxDriver instance. The second argument in its <a href="https://seleniumhq.github.io/selenium/docs/api/java/org/openqa/selenium/support/ui/WebDriverWait.html#WebDriverWait-org.openqa.selenium.WebDriver-long-">constructor</a> represents the maximum amount of time that our program should wait for an action (e.g. a page load).</li><li>After initialization, we have one new Firefox window that’s controlled by our program. Now let’s navigate to a specific webpage:<br>driver.navigate().to(&quot;http://www.galabra.co.il&quot;);</li><li>Now we want to click an element. We’ll need to specify a one-to-one address that’ll point this exact element. There are a few addresses we can use, but I find <a href="https://en.wikipedia.org/wiki/XPath"><strong>XPath</strong></a> to be the most convenient one.<br>In order to find the button’s XPath I’m going to use Chrome’s Developer Tools: right-click the element, then click <strong>Inspect</strong>.<br>The page’s HTML code will be opened and the inspected element will be highlighted. Right-click it, click <strong>Copy</strong> and then <strong>Copy XPath</strong>.<br>Now you have the element’s XPath that can be used as its address! Let’s save it as a String:<br>String aboutButtonXpath = &quot;//*[@id=\&quot;about\&quot;]div/a&quot;;</li><li>The code is executed regardless of the browser’s state, so we need to make sure the button element is loaded before we try to click it. As mentioned, we can use our wait instance just for that (one of the reasons I love Selenium is because the syntax is self-explanatory):<br>wait.until(ExpectedConditions.elementToBeClickable(By.xpath(aboutButtonXpath)));</li><li>Our WebDriverWait instance blocks our thread until the button element is clickable. Hence, right after it we can click it:<br>driver.findElement(By.xpath(aboutButtonXpath)).click();<br>Note how the click is called via driver.</li><li>Using the same method, we make sure the relevant paragraph is loaded before trying to copy its data:<br>String languagesParagraphXpath = “//*[@id=\&quot;page1\&quot;]/div[2]/div[5]”;<br>wait.until(ExpectedConditions.visibilityOfElementLocated(By.xpath(languagesParagraphXpath)));</li><li>Now we can use the logic that we’ve defined in the workflow. Using driver.findElements() method, we’ll retrieve a list of <a href="https://seleniumhq.github.io/selenium/docs/api/java/org/openqa/selenium/WebElement.html">WebElement</a>s, each representing a &lt;p&gt; that contains a language name:<br>List&lt;WebElement&gt; languageNamesList = driver.findElements(By.xpath(&quot;//*[@id=\&quot;page1\&quot;]/div[2]/div[5]/p&quot;));<br>Note that the last level in the XPath’s hierarchy doesn’t have an index. That’s because this XPath points to the array of all elements that satisfy this HTML tags hierarchy.<br>Another detail worth mentioning is that the method used here is driver.findElement<strong>s </strong>(in plural form), which suggests that the expected return value is a list.</li><li>We’re finished scraping, so we can close the browser window:<br>driver.close();</li><li>Lastly, I chose to just print the language names. However, you can do anything you want with them, as they’re now accessible from within your code!</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pfeFC8Lu8aZFYczjdnjR0w.png" /><figcaption>Stage 3: finding an element’s XPath</figcaption></figure><p><strong><em>That’s it.</em></strong> I hope I was able to give you a basic understanding of Selenium for Java, its applications and its implementation.<br>Please don’t hesitate to contact me for any questions.</p><p>Thank you for your time!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9d37c52ce7a2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>