<?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 Martin Campbell on Medium]]></title>
        <description><![CDATA[Stories by Martin Campbell on Medium]]></description>
        <link>https://medium.com/@macrotentional?source=rss-6ef92153be2f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*awFbXyXwK3gK6UQ_p--SaA.gif</url>
            <title>Stories by Martin Campbell on Medium</title>
            <link>https://medium.com/@macrotentional?source=rss-6ef92153be2f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 23:24:00 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@macrotentional/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[Investigating OkCupid profiles: can we predict someone’s city?]]></title>
            <link>https://medium.com/@macrotentional/investigating-okcupid-profiles-can-we-predict-someones-city-31a4734e96dd?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/31a4734e96dd</guid>
            <category><![CDATA[okcupid]]></category>
            <category><![CDATA[dating-app]]></category>
            <category><![CDATA[logistic-regression]]></category>
            <category><![CDATA[k-nearest-neighbours]]></category>
            <category><![CDATA[decision-tree-classifier]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Thu, 21 Jul 2022 21:13:21 GMT</pubDate>
            <atom:updated>2022-07-21T21:13:21.780Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/760/1*OP-J7xA1nN_f0OJp6DMKbQ.jpeg" /><figcaption>Los Angeles, CA (image credit: travelinusa.us)</figcaption></figure><h3>Introduction</h3><p>This project analyzes data from online dating app OkCupid. In recent years, there has been a massive rise in the usage of dating apps to find love. Many of these apps use sophisticated data science techniques to recommend possible matches to users and to optimize the user experience. <strong>These apps give us access to a wealth of information that we’ve never had before</strong> about how different people experience romance.</p><p>The goal of this project is to scope, prep, analyze, and create a machine learning model to solve a research question.</p><h4>Project goals</h4><p>In this project, the goal is to utilize the skills learned through Codecademy and apply machine learning techniques to a data set. The primary research question that will be answered:</p><blockquote><strong>Can an OkCupid’s user’s city be predicted using other variables from their profiles?</strong></blockquote><h4>Data</h4><p>The project has one data set provided by Codecademy called profiles.csv. In the data, each row represents an OkCupid (OKC) user and the columns are the responses to their user profiles which include multi-choice and short answer questions.</p><h4>Analysis</h4><p><strong>This solution will use descriptive statistics and data visualization</strong> to identify key figures in understanding the distribution, count, and relationship between variables. Since the goal of the project is to make predictions on the user’s city, classification algorithms from the supervised learning family of machine learning models will be implemented.</p><h4>Evaluation</h4><p>The project will conclude with the evaluation of the machine learning model selected with a validation data set. The output of the predictions can be checked through a confusion matrix, and metrics such as accuracy, precision, recall, F1 and Kappa scores.</p><h3>Exploratory data analysis</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_j-eu9VjKss4QdSnAnM66Q.png" /><figcaption>First 5 rows of profiles.csv</figcaption></figure><p>There are <strong>31 features</strong> and <strong>59,946 rows</strong> in this dataset, which should be ample data to draw statistically significant conclusions. Other than age, height, and income, all of them are categorical and there are also nine short response questions. Onward!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/258/1*cIqLJ-_cpHnWW031NxOl6A.png" /><figcaption>data types of each column of the dataset</figcaption></figure><p><strong>Features</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/387/1*4fJDLCMMgpS541px7paZ3Q.png" /><figcaption>Age distribution of OkCupid users</figcaption></figure><p>From this information we can see that a vast majority of OKC users are in their 20s or 30s, and there is a steep drop-off after age 40. Like most dating apps, OKC caters to young adults.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/379/1*rG9wlkJYPdsrQkdSSF6dTA.png" /><figcaption>Count plot of sex of OkCupid users</figcaption></figure><p>There is a pronounced skew towards male users, which means straight males may have more difficulty finding partners, and straight females can be more selective.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/441/1*HSbGt8apB5xAGLA4c3HZgw.png" /><figcaption>Count plot of body type of OkCupid users</figcaption></figure><p>Naturally the most common body type is “average.” Athletic and fit are also popular descriptors, while users who are overweight are more likely to describe themselves as “curvy” than any other adjective.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/461/1*rA6pkPZSG89yDoTFaf6Z9w.png" /><figcaption>Count plot of diet of OkCupid users</figcaption></figure><p>When it comes to diet, OKC users are not particular choosy — a large proportion of them characterizing their diets as eating “anything,” “strictly anything,” or “mostly anything.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/429/1*v9v8JKV__x6hz8MTyH5s9w.png" /><figcaption>Count plot of alcohol consumption of OkCupid users</figcaption></figure><p>Most OKC profiles indicate the user drinks socially. There are more heavy drinkers than teetotalers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/424/1*UDaHYI7nC_rloqjROTyFNw.png" /><figcaption>Count plot of drug consumption of OkCupid users</figcaption></figure><p>A strong majority of users never do drugs, and only a very small percentage use drugs often.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/554/1*N6tO9SQJZRW9f_vpWwSIYA.png" /><figcaption>Count plot of education of OkCupid users</figcaption></figure><p>OKC users are a fairly educated bunch, with the common responses being “graduated from college/university” or “graduated from master’s program.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/423/1*3MF2snzja7DptJdD9jPlkw.png" /><figcaption>Count plot of astrological sign of OkCupid users</figcaption></figure><p>As one would expect, the 12 astrological signs are roughly evenly distributed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/441/1*P4h14bK3_VO9Kbp7bP7f8Q.png" /><figcaption>Count plot of smoking prevalence of OkCupid users</figcaption></figure><p>Here we find that the majority of people on OKC don’t smoke, but interestingly only a minority of smokers are trying to quit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/451/1*Bc8XUiW0uk4qL6dmpczS5w.png" /><figcaption>Count plot of relationship status of OkCupid users</figcaption></figure><p>Unsurprisingly, something like 95% of users are single or available. OKC is a dating site after all!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/411/1*iPZXtplORiGz_lCAFLyahA.png" /><figcaption>Count plot of ethnicity of OkCupid users</figcaption></figure><p>OKC skews white, and there are more asian and fewer black and hispanic users than one would expect given the population demographics of a US-based dating platform.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/645/1*hyhybjgEX1KqW76o2hc3oQ.png" /><figcaption>Count plot of offspring of OkCupid users</figcaption></figure><p>While most users don’t have children,the interesting insight here is that given that ~95% of the userbase is single/available, a significant proportion of them <em>do</em> have kids!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/411/1*y5MiiYjk7u2YgoumSxX-Tg.png" /><figcaption>Count plot of sexual orientation of OkCupid users</figcaption></figure><p>Heterosexuals are about 10x as common as homosexual users, which goes along with the oft-cited statistic that <a href="https://www.theguardian.com/society/2015/apr/05/10-per-cent-population-gay-alfred-kinsey-statistics">10% of people are gay.</a> Curiously, bisexual profiles are roughly half as common as gay ones.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/411/1*2Uk7GJ9Pnv-2jrWButLYkQ.png" /><figcaption>Count plot of sex + sexual orientation of OkCupid users</figcaption></figure><p>Digging a little deeper, we learn that men are more likely to identify as gay, but women are more likely to identify as bisexual.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/521/1*81xzMtJtRMEIEKI30EBpsQ.png" /><figcaption>Count plot of pets of OkCupid users</figcaption></figure><p>Pets are popular across the board, but more users like dogs than cats.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/430/1*rPWzo7km4WiJklIdlUjvLg.png" /><figcaption>Count plot of religion of OkCupid users</figcaption></figure><p>Here we discover that when it comes to religion, OKC profiles are <a href="https://news.gallup.com/poll/358364/religious-americans.aspx">dramatically different from the overall population</a>, with a plurality of users ascribing to agnosticism, and christianity being less popular than atheism (!).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/528/1*HyB9sfou7KDa2wTyx5CSAw.png" /><figcaption>Count plot of job of OkCupid users</figcaption></figure><p>Beyond “other” and “student,” the most common jobs are science/tech/engineering, computer/hardware/software, and artistic/musical/writer.</p><h3>Who are we really looking at?</h3><p>Eagle-eyed readers may have noticed that the first 5 rows of the dataset were all profiles from California. Indeed, the dataset is <em>extremely</em> unrepresentative of the US population, with &gt;99.9% of profiles being from the Golden State:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/334/1*nMt0bI01s42JMRRh9aXZnw.png" /><figcaption>Breakdown of profiles by state</figcaption></figure><p>Suddenly the ethnic, religious, and vocational makeup of the users we examined makes much more sense!</p><p>This means <strong>the data should NOT be taken as representative of the overall OKC userbase, but *only* OKC users in California.</strong></p><p>(With that in mind, we will adjust our scope to only include profiles from CA.)</p><p>Drilling down to the city level, we find that not only is the dataset exceptionally biased towards California, <strong>the makeup of the sample is drawn predominantly from San Francisco:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/451/1*DucIoJPrFJyZeaV16yrZlw.png" /><figcaption>Breakdown of profiles by city</figcaption></figure><p>This is the point when I started to think seriously about what kind of research question to pose.</p><p>Other Codecademy users had already scrutinized features such as age, alcohol use, religion, and astrological sign, so I wanted to do something different.</p><p>Given the most striking aspect of the sample is most of the profiles are users who live in San Francisco<strong>, I decided that it would be interesting to determine whether it’s possible to predict a user’s city from the other variables they share.</strong></p><p>To further narrow the focus, I chose to only analyze the cities home to more than 1000 profiles, which were the following:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/374/1*xRagsPhlcsQez19OZr3VUQ.png" /><figcaption>Cities selected to analyze</figcaption></figure><p>I wanted to use machine learning techniques (not natural language processing), which means I had to remove the essay questions.</p><p>This is what the profile rows looked like after adjusting to only include features that might be conducive to machine learning:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-w1A2AqtyXHO4B08boL3QA.png" /></figure><p>The next step was to figure out how many empty entries we were dealing with. Turns out for some columns there were quite a few:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/334/1*yGANoLAcLSF9uFZGFCUgdg.png" /><figcaption>NaNs by column</figcaption></figure><p>In the interest of simplicity and in order to ensure the machine learning techniques ran smoothly, I chose to drop all the rows with missing values. This was okay because it still left me with more than 16,000 rows to play with:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/718/1*I61VweiKML7emDip1JfrRw.png" /><figcaption>shape of final data sample: 12 columns and 16,170 rows</figcaption></figure><p>Once the data was was fully cleaned, it was separated into train and test sets with validation size of 25% of the sample. This was a relatively straightforward process conducted with the train_test_split utility:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6WDXpXLro4Mx5J9UiULC-g.png" /><figcaption>Train-test-split on cleaned data</figcaption></figure><h3>Machine learning models</h3><p>For this project four common algorithms were used to make predictions:</p><ol><li><strong>Logistic regression</strong></li><li><strong>K-nearest neighbors</strong></li><li><strong>Decision trees</strong></li><li><strong>Random forests</strong></li></ol><h4>1. Logistic regression</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9U1KYr8fG2WqGuZy7y3buQ.png" /><figcaption>Classification report for logistic regression</figcaption></figure><p>Analysis of logistic regression initially indicated the easiest city to predict was San Francisco. As San Francisco constitutes &gt;67.5% of the sample, this was not as impressive as it appears. The model struggled to identify Palo Alto (the smallest city included) at all, mistakingly identifying it as SF.</p><p>That said, considering it’s the second smallest city with only 502 users, the <strong>predictions for San Mateo (67%) were very good. </strong>Ones for Oakland were also not bad (17.0% of the sample, yet reaching 57% precision).</p><h4>2. K-nearest neighbors</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gKvpGzcKiJUjud75WXX9zA.png" /><figcaption>Classification report for K nearest neighbors</figcaption></figure><p>Analysis of k-nearest neighbors revealed a similar yet slightly worse result. Although predictions for SF were at the same level as the logistic regression (that is, no better than chance), <strong>precision predictions for Oakland and San Mateo took a nosedive</strong>, this time tracking half as good.</p><p>Overall: KNN was not quite as effective as logistic regression.</p><h4>3. Decision trees</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CAnU40m8kapj2CtpmKYtIA.png" /><figcaption>Classification report for decision trees</figcaption></figure><p>Decision trees produced a worse result than either of the previous algorithms. Could it be that that depth of the decision tree was to blame?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/308/1*GZMT5tA4IEyPHNIbpyumng.png" /><figcaption>Default decision tree depth</figcaption></figure><p>To determine the optimal depth, I opted to run a cross validation test. Plotting the cross accuracy per decision tree depth suggested the best tree depth was 3:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4DvZJhdsx7d9RgDtkZpLPw.png" /><figcaption>Decision tree cross validation test</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1010/1*9mun-SRlkhaDZ_hFvpKdtg.png" /><figcaption>Decision tree classification report with an applied tree depth of 3</figcaption></figure><p>At a big picture level, using an ideal tree depth ended up helping our model. Yet <strong>in the process it improved unevenly </strong>— the model grew better at predicting Oakland at the cost of misidentifying the 3 smaller cities 100% of the time.</p><h4>4. Random forests</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/829/1*5Eusm3TSPU8Aji2KCGgDIQ.png" /><figcaption>Classification report for random forest classifier</figcaption></figure><p>As a final algorithm, I selected random forests to see how it compared with decision trees.</p><p>Here we see that <strong>random forests opted to prioritize the top 3 cities</strong>, meaning it was more accurate than our ideal tree depth decision tree for Berkeley, but less so for Oakland. Palo Alto and San Mateo were onced again neglected.</p><h3>Insights</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ExljjVN2n-qDMRFCugQ68Q.jpeg" /><figcaption>San Francisco, CA (image credit: getyourguide.com)</figcaption></figure><p>Ultimately the question remains:</p><blockquote>Is it possible to accurately predict someone’s city from the other data they include in their dating profile?</blockquote><h4>At this point, the answer is <strong>not so clear</strong>.</h4><blockquote>Working with an unrepresentative dataset introduced limitations to the project, which affected the models and their predictions.</blockquote><p>That said, each of the models approached the data in a different way, and some of them were pretty good at correctly guessing whether you’re from San Francisco or Oakland.</p><p>If your city is oversampled (like SF was here), it seems like using some of the most common algorithms can provide relatively accurate forecasts. If your city is undersampled, it runs considerable risk of being mistaken for a an oversampled one.</p><p>Overall, <strong>logistic regression</strong> and <strong>optimized decision trees</strong> produced the best accuracy. However, the former generated predictions for 4/5 cities while the latter honed in on the two largest.</p><p>In other words, there are tradeoffs to both models:</p><blockquote>If you’re interested in maximizing accuracy for your largest samples, choose optimized decision trees.</blockquote><blockquote>On the other hand, go with logistic regression if you want to maximize number of predictions.</blockquote><p>When evaluating the quality of the models, here’s another point worth considering:</p><blockquote>It’s certainly possible that the selection of users (all from the Bay area) was too intrinsically similar, which made forming distinctions between them difficult.</blockquote><p>If the data were evenly split between, say, SF, New York, Chicago, and DC, it’s conceivable that the models would perform better than when the cities were all from California.</p><h3>Implications</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/991/1*l8F9DpWya9HvfWqRkriPuA.jpeg" /><figcaption>Berkeley, CA (image credit: workforcehub.com)</figcaption></figure><p>The findings of this report may partially allay <strong>concerns over identity tracking.</strong></p><p>You can share a multitude of info about yourself yet keep your location status undefined, and based simply on the other info you share, it’s not immediately clear if someone would be able to identity your current city.</p><p>Of course, there are other forms of online tracking, but OKCupid users can be confident that while it’s not impossible, it’s also not easy for someone to infer your location from the other info you provide.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/834/1*MhVfphlk-fLCA4Bknl6emg.png" /><figcaption>SMOTE analysis to upsample smaller cities (didn’t improve improve on results)</figcaption></figure><h3>Thoughts &amp; next steps</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J5fiYaAiwWmVCPNwoLWmOg.jpeg" /><figcaption>Oakland, CA (image credit: mwg.aaa.com)</figcaption></figure><p>The longer I worked on this project, the more apparent it became that I was working with <strong>unrepresentative data</strong>. With that in mind, it’s difficult to draw any long-ranging conclusions when you remember that this was essentially just a sample of the Bay area.</p><p>Nonetheless, this was an exciting project that gave me the chance to examine self-reported data on a massive scale. There remains much potential for uncovering other insights beyond those related to one’s location.</p><p><strong>It would definitely be worth revisiting this project with another dataset featuring a more diverse range of cities.</strong> Doing so could reveal further observations about society and help to resolve the question of whether your location can be predicted from the other details you reveal about yourself.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*ZtLDaGx1Hfe973d7diAkZg.png" /><figcaption>Wanted: data from more cities (image credit: <a href="https://www.pixeltrue.com/premium/modern-illustrations-2">https://www.pixeltrue.com</a>)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=31a4734e96dd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Does working in tech take a toll on your mental health?]]></title>
            <link>https://medium.com/@macrotentional/does-working-in-tech-take-a-toll-on-your-mental-health-af22a301fa35?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/af22a301fa35</guid>
            <category><![CDATA[tech-industry]]></category>
            <category><![CDATA[mental-health-awareness]]></category>
            <category><![CDATA[surveys]]></category>
            <category><![CDATA[lambda-school]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Fri, 29 May 2020 15:36:18 GMT</pubDate>
            <atom:updated>2020-05-30T20:33:43.617Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BJE3zle8s-t2Im26NJpYyg.jpeg" /><figcaption>What are the mental costs of working in tech? (Image credit: <a href="https://www.pexels.com/@andrew">Andrew Neel</a>)</figcaption></figure><p><strong>A career in tech has its perks, but for many they come at a heavy personal cost. </strong>The entry barriers are high, and there tends to be intense pressure to stay on top of the latest trends to keep your skills current. Age discrimination is a well-known concern, as are gender and ethnic inequities.</p><blockquote>Companies don’t entice applicants with benefits like meditation stipends and <a href="https://www.wethos.co/career-product-designer/?SQF_SOURCE=DenverUX">mental health days</a> for nothing.</blockquote><p>Among those who clear the hurdles and land a tech position, many suffer from impostor syndrome and the persistent fear that sooner or later they’ll be laid off in favor of the latest crop of comp sci grads.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9DPuMf4z1neCtpIFhL2zmQ.png" /><figcaption>Feel like you don’t belong in tech? <a href="https://www.teamblind.com/blog/index.php/2018/09/05/58-percent-of-tech-workers-feel-like-impostors/">You’re not alone</a></figcaption></figure><p><a href="https://medium.com/s/story/techs-long-hours-are-discriminatory-counter-productive-17dc61071ed5">Across the industry, long days are the norm.</a> This is particularly true at startups, in which a traditional 9–5 schedule is arguably incompatible with the gritty spirit of entrepreneurial enterprise.</p><blockquote>If your payoff with an aspiring Y Combinator unicorn is measured by how many hours you invested in its mission, you’re incentivized to stretch yourself to your limit.</blockquote><p><strong>Good luck making it to the top if you only do the minimum required.</strong> And it’s really not very different at established companies, which with amenities like catered meals and on-site gyms, aim to keep you within their bubble for as much of the day as possible.</p><p>As a budding tech employee, I found myself drawn to this topic because of its relevance to my career aspirations.</p><ul><li>Could all the horrors stories be true?</li><li>What could do I do to avoid experiencing burnout (or worse)?</li></ul><p><strong>With these issues in mind, I decided to take a closer look at data on the state of mental health in the tech industry.</strong></p><h3><strong>Background research</strong></h3><p>For the last few years an organization called the <a href="https://osmihelp.org/"><strong>Open Source Mental Illness team (OSMI)</strong></a> has conducted an annual survey to assess how mental health is viewed within the tech/IT workplace, and the prevalence of certain mental health disorders within the industry.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tVZyzLFF4tBW4p89OrKGxQ.png" /><figcaption><a href="https://leanpub.com/osmi-mental-wellness-in-the-workplace">Guidelines for Mental Wellness in the Workplace</a></figcaption></figure><p><strong>Since my goal was to employ machine learning methods, I chose to examine </strong><a href="https://www.kaggle.com/osmi/data"><strong>OSMI’s 2016 survey</strong></a><strong> because it featured the greatest number of responses—1,433—of any of the years recorded.</strong></p><h3><strong>Exploring and cleaning the data</strong></h3><p>An immediate challenge was this dataset had no continuous values — all of the features were categorical. On closer inspection most of the questions were asked on a <a href="https://en.wikipedia.org/wiki/Likert_scale">Likert</a> scale: <strong>variations of “sometimes”, “maybe”, and “I don’t know” were everywhere.</strong></p><p>In order to compare and contrast questions and look for correlations, I had to reencode nearly each of the responses to a numerical value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3Q9UwbUUMphGs-6g5J8gmw.png" /><figcaption>The dataset before I changed the observations to a number scale</figcaption></figure><p>When dealing with unfamiliar data it’s always important to check for missing values. There were a dozen or so questions in which the vast majority of respondents either didn’t reply or the question was not applicable to their situation.</p><p><strong>The survey was composed of a decision tree of sorts.</strong> For instance, if you asserted you were self-employed, there were a number of questions related to your employer that you would not have answered.</p><p>Nonetheless, the questions with large numbers of missing values didn’t affect my results as I simply didn’t use those columns in my analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Dd4x9IniDm7UE8CvLxgiBg.png" /><figcaption>A heatmap displaying the percentage of missing values</figcaption></figure><p><strong>One of the few open-ended questions asked about a participant’s job title.</strong> Since this topic was not appropriate to numerically encode, the number of positions needed to be scaled down to a manageable amount.</p><p>This was achieved by broadly grouping by keyword, and combining the rest into a category called ‘Other’. The resulting list:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WgK-uecH2jsySSUZK8w6BA.png" /></figure><h3><strong>Research questions</strong></h3><p>My research topics evolved with my understanding of the dataset.</p><p>Although the survey focused on the tech industry, not all respondents answered that they were tech workers—in fact, 20% answered that their employer wasn’t a tech company.</p><h4><strong>1. Are respondents who identify as tech workers more likely to be currently diagnosed with a mental health condition?</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HQImDOnCcfArgBuugDK-9Q.png" /></figure><p>I was surprised to discover that there was not a statistically significant difference in diagnosis rates between tech and non-tech employees.</p><blockquote><strong>The group with the highest proportion of mental health diagnosis was self-employed workers.</strong></blockquote><p>Although the survey didn’t distinguish whether self-employed respondents worked in tech or not, they were a significant proportion of the sample. Since the survey was targeted at tech workers, it’s reasonable to assume many worked in tech-adjacent roles.</p><p>It’s easy to imagine the multitude of stresses related to a freelance career—like irregular schedules and actually getting paid—can wear you down and serve as additional hurdles to mental wellbeing.</p><h4><strong>2. Are certain positions in the tech industry more prone to mental illness diagnoses in general?</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZdR_3dOtMJx8phX4loclLw.png" /></figure><p>From this analysis I learned that <strong>designers are the tech role most likely to suffer from diagnosed mental health conditions. </strong>People who identified as ‘dev evangelists/advocates’ are also at high risk.</p><p>These were the only two categories of job roles in which a majority of respondents were currently diagnosed.</p><p>On the other hand, <strong>front-end developers were the group <em>least</em> likely to have a current mental health diagnosis.</strong></p><blockquote>More than any other tech role, designers are known for their creativity. I can only speculate, but perhaps there’s some truth in the stereotype of <a href="https://en.wikipedia.org/wiki/Creativity_and_mental_health">creativity being paired with mental illness</a>.</blockquote><h3><strong>Predicting modeling</strong></h3><p>Although it was clear from the start that since that were no continuous variables, I would have to apply a categorical model, the best choice was not apparent. With 60+ survey questions, there were quite a few variables to scrutinize.</p><p><strong>I decided to keep things simple and selected a binary classification of the current overall diagnosis rate.</strong></p><p>I wasn’t able to just use “Do you currently have a mental health disorder?” because the survey included a ‘maybe’ option, which would have muddied the results.</p><p>The order of questions:</p><p>Do you currently have a mental health disorder? -&gt; If yes, what condition(s) have you been diagnosed with? -&gt; Currently diagnosed? (yes or no)</p><p>The <strong>baseline</strong> for my model was the majority class of the currently diagnosed column, which was <strong>60.4%:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/666/1*2KBFfigq-5LJaNy2W4k-tA.png" /></figure><p>After running a random forest classifier, I was able to assess the permutation importance of each question.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*abe-hG_b_D_a2nTjvvyNaQ.png" /></figure><p><strong>This visual shows that the branch-off question if you responded ‘maybe’ to “Do you currently have a mental health disorder?” was likely skewing my results, acting as a leaky variable.</strong></p><p>To avoid contaminating my model, it needed to be dropped.</p><p>After removing it, the features for the model were updated. As one might expect, having a mental disorder in the past is a strong predictor of having one today. That explains the top three questions below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MQNfhVI8_LIVNstVEv8pXA.png" /><figcaption>Updated permutation importances (i.e. questions that best predicted answer to “Currently diagnosed?”)</figcaption></figure><p>After accounting for that somewhat obvious correlation, the strongest predictors of a current mental health diagnosis were as follows:</p><blockquote>1. Have you ever sought treatment for a mental health issue from a mental health professional?</blockquote><blockquote>2. What is your age?</blockquote><blockquote>3. If you have a mental health issue, do you feel that it interferes with your work when being treated effectively?</blockquote><blockquote>4. Do you have a family history of mental illness?</blockquote><p>I reran the model with randomized search cross-validation, and ended up with an <strong>accuracy score of 81.8%</strong>—a respectable upgrade on the baseline.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JoCrHvw-bhJjWjZP8F2o-w.png" /><figcaption>Not a bad score considering I didn’t tune any hyperparameters</figcaption></figure><p>Finally, I looked at the interaction of a partial dependence plot (PDP) between different combinations of questions.</p><p>Here you can see a comparison of the number of employees at a company in relation to willingness to discuss mental health issues with supervisors:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sv9S0Qhg4_ppFHaWAIfQjQ.png" /></figure><h4>In general, the larger a company, the more comfortable an employee is to share their mental health issues.</h4><h3><strong>Next steps</strong></h3><p>An inquiry that I didn’t get around to investigating was whether certain tech positions are more prone to specific mental health diagnoses.</p><p>For example:</p><blockquote>Are executives more likely to suffer from a psychotic disorder?</blockquote><blockquote>Do the self-employed tend to have higher rates of depression?</blockquote><blockquote>Are back-end developers at greater risk of a mood disorder?</blockquote><p>I cannot emphasize enough how <strong>I barely scratched the surface of the myriad of mineable insights within this dataset.</strong> There were 62 questions in the survey; I devoted my attention to just a few of them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*W0puejbE_S7dUvLCuIRH-Q.jpeg" /><figcaption>(Image credit: <a href="https://www.pexels.com/@gustavo-fring">Gustavo Fring</a>)</figcaption></figure><p>With the rise of COVID and the industry-wide shift to a work from home lifestyle, the space between home and work life for those in tech has been reduced to zero.</p><p>Now that separating the two is more difficult than ever, reports on <a href="https://www.cnn.com/2009/LIVING/worklife/05/11/weisure/">the phenomenon of ‘weisure,’</a> or the blurring of work and play, have become unsettlingly accurate.</p><p><strong>I’m sure OSMI’s 2020 survey will have much to say on this matter.</strong></p><p>Where do we go from here? I’ll be the first to admit I don’t have the answer.</p><h3>The good news: tech workers don’t appear to face higher rates of diagnosed mental health issues than than those who work for non-tech companies.</h3><p>It’s my hope that by further exploring data generously provided by organizations like OSMI, we can join together to illuminate a path across the workforce spectrum and break down barriers to develop scalable solutions that improve the mental wellbeing for all.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KvIXxH7QPuBShoINDNRPIA.jpeg" /><figcaption>Tech and non-tech workers are in this together (image credit: <a href="https://www.pexels.com/@jibarofoto">Luis Quintero</a>)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af22a301fa35" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What I learned from analyzing my Fitbit data]]></title>
            <link>https://medium.com/@macrotentional/what-i-learned-from-analyzing-my-fitbit-data-f70fb622911d?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/f70fb622911d</guid>
            <category><![CDATA[fitness-trackers]]></category>
            <category><![CDATA[personal-data]]></category>
            <category><![CDATA[fitbit]]></category>
            <category><![CDATA[linear-regression]]></category>
            <category><![CDATA[lambda-school]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Fri, 01 May 2020 16:46:24 GMT</pubDate>
            <atom:updated>2020-05-29T00:34:23.655Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*011g7bn2GcAp0gQcM_X6UQ.jpeg" /><figcaption>Image credit: <a href="https://www.ama-assn.org/topics/patient-privacy?page=1">American Medical Association</a></figcaption></figure><h3>A world of metrics</h3><p>Everything we do online is quantified. Like it or not, there’s more data about us than ever before. Understandably, many of us are concerned about our private information being stolen or exploited.</p><p>Despite the fears of having our lives quantified, we’re simultaneously also drawn to the idea.</p><p><strong>The rise of wearables and DIY DNA tests shows that when access is limited to the individual, personal data can also be used to empower us.</strong></p><p>Every dataset has the potential to tell a story. When we peer under the hood, what hidden insights can we discover about ourselves?</p><h3><strong>Background</strong></h3><p>When I decided to investigate a dataset, I had no idea where to start. There’s no shortage of data on the web, and an incredible variety of sources and subjects to consider. From entertainment, to government, to neuroscience, and everything in between, <strong>if you look hard enough, you can find public data on nearly any topic.</strong></p><p>Frustrated by the seemingly limitless options, I sighed and looked down from my screen. As my gaze shifted to my hands, I suddenly had an idea.</p><blockquote>What if my research topic was right in front of me the whole time?</blockquote><h3><strong>The quantified self</strong></h3><p>For the past six months, I’ve been religiously wearing a Fitbit fitness tracker. My mother received it as a birthday gift, but decided she didn’t want to bother with another device to charge and lent it to me to try it out.</p><p>I’ve found it most useful for incentivizing myself to take a few extra laps around the block and for tracking my sleep.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ieNghYNx12yheza8ZLZ3hw.png" /><figcaption>The fitbit.com user dashboard is fully customizable</figcaption></figure><p>In a certain sense, exploring my own fitness data info seemed like a cop-out. Nonetheless, it was freely available, I was already familiar with the variables, and my personal wellness is a subject about which I’m passionate.</p><p><strong>I know I’m not the only one interested in my own health data.</strong> As someone curious about physiology, brain waves, and improving my sleep, it’s a matter very much near to my heart. What could go wrong?</p><h3><strong>Exercise vs sleep</strong></h3><p>The data provided by fitness trackers allowed anyone to investigate this topic on a personal, granular level.</p><p>From the start, the most natural inquiry to examine was whether exercise affects sleep. Yet I knew I needed to narrow the focus to concentrate on two specific features measured by the Fitbit.</p><p>As such, <strong>I chose to compare my sleep quality score and ‘activity calories’</strong>, a measure of calories burned when the device detects my heartrate is elevated and I’m engaged in prolonged physical activity.</p><p>Of all the variables related to exercise, activity calories was the closest approximation to exercise. Others included daily calories burned, steps taken, minutes lightly active, and minutes very active, but I wasn’t satifisfied that they fully captured the notion of exercise sessions.</p><blockquote>Research question: what is the relationship between sleep quality and calories burned from exercise?</blockquote><h3><strong>Process</strong></h3><p>After settling on my research topic, I logged into the Fitbit website. I assumed that it would be possible to download a record of your data. Sure enough, there was an option to export your data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8V_vTgf1SPA6SVD-3wmZPw.png" /></figure><p>You can either download from a certain time period or export an archive from the lifetime of your account. I decided to do both and compare the info.</p><p>Conveniently most of the data is already in csv format, allowing me to easily upload into a Python notebook with the pd.read_csv function.</p><p>I discovered the data in each was slightly different. My nemesis, the nightly sleep score, was only included in the time period reports.</p><p><strong>In order to cross-examine sleep score vs activity calories, I needed to merge the two dataframes</strong>. There was no matching column between the two, but it was easy enough to add a date feature to the sleep score csv.</p><p>In my new dataframe, each row matching to a different day. When combining the two, the only issue is I lost a few rows because some days lacked a sleep score. (This could have happened if I either had to charge the Fitbit and forgot to put it back on before bed, or I didn’t get a proper reading because it couldn’t pick up a consistent heart rate measurement throughout the night.)</p><p>However, this wasn’t a big deal. It simplified things because I was left with no missing values. Score!</p><h3><strong>Exploring the data</strong></h3><p>In total, my new dataset had 117 days of measurements and 25 columns—comprising all of the features related to sleep and exercise measured by the Fitbit Charge 2.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u8SQRiU4cxnWfwlmLs18pA.png" /><figcaption>Comparing all the numeric variables</figcaption></figure><p>With data cleaning out of the way, now came the fun part. What trends and patterns could I unearth?</p><p>I made a quick scatter plot of sleep quality vs activity calories. At first glance, there didn’t appear to be much of a correlation:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*AO2b-eSy1gzliGT2EHE0ng.png" /></figure><p>Sleep score is measured from 1 to 99. You can see that most of the data points were clustered between 70 and 85.</p><p>At the same time, most of the observations of calories burned from exercise are situated around 400–1000, which makes sense to me given my most common form of exercise is a 4-mile jog over the same familiar loop.</p><p>I decided to run a linear regression to determine the extent of the relationship between the two variables.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/894/1*Nq1ocPOej96q1jADc_cZQA.png" /></figure><blockquote>I learned there was a small negative correlation between Fitbit’s assessment of my sleep quality and the number of calories I burned during exercise.</blockquote><p><strong>I actually slept <em>worse </em>on days when I exercised a lot! </strong>Of course everyone’s body works differently. But this was an eye-opening discovery. What could account for this counterintuitive insight?</p><p>Other variables that could be at play:</p><ul><li>do I drink more caffeine on days when I exercise?</li><li>do I tend to eat more chocolate those nights?</li><li>do I stay up later because I feel like I’ve earned it?</li><li>do I jog as a way to procrastinate and then have more work to do at night? (this seems like the most likely explanation)</li></ul><p>All else being equal, based on my finding, if I run longer than usual, I could reasonably expect to sleep worse than if I go my regular distance.</p><h3><strong>Moving forward</strong></h3><p>I’ll admit it: my primary objective with this project was somewhat self-serving. As someone who rarely wakes up well-rested, I was interested in gaining a better understanding of my sleep patterns.</p><p><strong>In the end, I didn’t arrive at a clear answer.</strong></p><p>That being said, I learned so much more than expected, and feel like I barely scratched the surface of my fitness data.</p><blockquote>data is knowledge -&gt; knowledge is power -&gt; data is power</blockquote><p>Although we can’t predict the future, it’s a safe bet that the amount of personal data collected about us is sure to continue to grow. But that doesn’t mean we have to be passive about it.</p><p><strong>The better you know yourself, the more effective decisions you can make. </strong>Given the popularity of wearables and fitness apps, odds are good you’re already tracking your daily progress on some measure or another. If so, are you using it to grow and improve?</p><p>Ultimately, we can’t depend on external forces to use our personal information for our own benefit.</p><p><strong>It’s up to us to make the most of the data we have at our disposal.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/925/1*RD3s1JgTaSqS_iyDs4mKcg.jpeg" /><figcaption>What will you do with your personal data? (Image credit: <a href="https://burst.shopify.com/@matthew_henry?utm_campaign=photo_credit&amp;utm_content=Picture+of+Woman+Wearing+Winter+Jacket+-+Free+Stock+Photo&amp;utm_medium=referral&amp;utm_source=credit">Matthew Henry</a>)</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f70fb622911d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Remote Collaboration, Real-World Results]]></title>
            <link>https://medium.com/@macrotentional/remote-collaboration-real-world-results-33f60afb6f79?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/33f60afb6f79</guid>
            <category><![CDATA[design-process]]></category>
            <category><![CDATA[lambda-school]]></category>
            <category><![CDATA[language-learning]]></category>
            <category><![CDATA[remote-work]]></category>
            <category><![CDATA[collaboration-tools]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Mon, 07 Oct 2019 13:56:01 GMT</pubDate>
            <atom:updated>2019-10-07T13:56:01.210Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uovU6uydekNPa_Q8OkADlA.png" /></figure><h3>Working remotely on the site redesign for a language-learning company in Bahrain taught me a lot about how the world works, and also how to keep a team together.</h3><p>Over the past two weeks, we moved steadily through each stage of the design cycle: discovery, definition, ideation, prototyping, and evaluation. It was a terrific adventure on so many levels. Some of the skills we developed included working with a real client, sticking to a schedule, communicating with a remote team, resolving differences of opinion, preparing a minimally viable product, articulating a vision, and displaying grace under pressure.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oibDVme3N1QYkA2whDro6A.png" /><figcaption>Discovery: empathizing with the user’s worldview</figcaption></figure><p>The day before the deadline, I found it frustrating that other team members were running out of energy right before the finish line<strong>.</strong> <strong>Just because everyone finished their ‘job’ doesn’t mean a project is ready. </strong>This was partly because several members were unavailable (one was gone completely, another was abroad with limited access to the web, and one was tired after helping out a friend who was dealing with a home fire), so our team wasn’t running on full cylinders. No one else volunteered, so I decided to be one of our presenters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eloHZyfg00YFKbez" /><figcaption>Inspiration: investigating the competition</figcaption></figure><p>I try to step up when I know people need help. Turns out being a serial procrastinator is not necessarily a bad trait when working with a team. Where there’s a will, there’s a way. Other people in my group were tired and just want the project to be over, but <strong>the closer to the deadline, the more energized I became.</strong> I’m learning thanks to a lifetime of procrastination, I’m an unusually strong finisher. If I played baseball, I’d be a pinch hitter. Apparently I perform surprisingly well under pressure. I have a talent for summarizing info, staying organized, paying attention to detail, copyediting, understanding a client’s problems, and making conclusive decisions. <strong>When time is short and the stakes are high, that’s exactly when I can be counted on to do my best.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XFbjfIxREstvMgRi" /><figcaption>Definition: charting the user experience</figcaption></figure><p>Another lesson from this project is a lack of centralized leadership meant tasks that weren’t delegated weren’t prioritized. Without a leader, eight people might be too many for an optimally-run remote team. Although we each had certain responsibilities, <strong>what</strong> <strong>we needed were more clearly defined roles. </strong>For example payment specialist, copy editor, landing page designer, prototyper, and product lead. At the same time, it was critical for others to check our work. Nonetheless, we all finished our assigned user flows in a timely fashion. As a design tool, Figma was invaluable. I can now understand why it’s rapidly becoming the industry norm. Sketch and Adobe XD are powerful tools if you’re working on your own, but Figma dramatically simplifies group design collaboration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9vZPkz0L1ENWDsHk" /><figcaption>Clarification: understanding the payment dilemma</figcaption></figure><p>Other than the day before everything was due, <strong>the second week of this project went much more smoothly than the first.</strong> I was amazed at how effective our team was at communicating the morning of the presentation. Everything went like clockwork: messages and deliverables were shared seamlessly. There was a strong collective energy. After some initial friction, we learned from our mistakes and it was awesome to see our team come together.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4NBOtsVuQohLEeXc" /><figcaption>Prototyping: designing the vision</figcaption></figure><p>At the end of the day, the most lasting takeaway from this project is <strong>remote collaboration is the real deal.</strong> Even with every member of a team working from separate locations, you can still produce impressive results and make a striking impact. Although there were growing pains, it was immensely gratifying to hear our client compliment the product we’d been designing over the last two weeks. All of our hard work paid off—we really came into our own. Overall, our team hit a home run.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CMxvSg1t2a3W2l6pBOZiXg.png" /><figcaption>Evaluation: refining the solution</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=33f60afb6f79" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Information Architecture: Building a Sturdy and Dynamic Knowledge Framework]]></title>
            <link>https://medium.com/@macrotentional/information-architecture-building-a-knowledge-framework-thats-sturdy-and-dynamic-7570f3f4bc6c?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/7570f3f4bc6c</guid>
            <category><![CDATA[lambda-school]]></category>
            <category><![CDATA[navigation]]></category>
            <category><![CDATA[information-architecture]]></category>
            <category><![CDATA[card-sorting]]></category>
            <category><![CDATA[labels]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Fri, 16 Aug 2019 00:00:40 GMT</pubDate>
            <atom:updated>2019-08-16T16:20:56.187Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1012/1*vPKJwGhXUD67b4mqGC9CKQ.png" /><figcaption>Seek and you shall find: the hallmark of a well-structured site</figcaption></figure><p>Information architecture refers to the way information is organized to help users achieve the goals they’re seeking to accomplish with your product. Everything from books, to maps, to websites, to mobile apps, and even urban planning involves information architecture in some way or another. <strong>Action labeling, card sorting, and blueprint mapping</strong> are several useful exercises to gain a stronger understanding of how to design your product effectively.</p><p>While planning how to structure the pages of Droom, the job-matching app I’ve been designing, I gained valuable hands-on experience in pitfalls to avoid and best practices to follow:</p><p><strong>Design for All Users</strong></p><p>The first version of my card sort test outlined the path for job seekers using Droom very well. The only problem? It failed to take into account employers. As a matching app, there will not be prospective employees using the service, but also employers. The goal of the app is for it to be able to work both ways, so it’s imperative that mapping the employer experience is also included. Although it’s true the vast majority of users will probably be those hunting for jobs, I also want to emphasize the way Droom can simplify filling a position and ease the work done by recruiters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4vJMsFJ--Wx0FZfs" /><figcaption>Card sort of page categories for Droom (basically Tinder for job hunters)</figcaption></figure><p><strong>Your Market Might Be Larger Than You Think</strong></p><p>Reflecting on that oversight, it’s clear to me that the reason I neglected the employer is I’m personally more familiar with the experience of the job seeker. I have a lot of experience hunting for jobs, not so much with hiring. I don’t know what it’s like to have to sort through applications and decide who will move on. While I’ve already considered the issues and goals of the Droom CEO, CTO, Product Manager, and developed a persona for the primary user, this exercise opened my eyes to the fact that I didn’t account for an important secondary persona — the busy HR professional.</p><p><strong>Information Architecture Is Dynamic</strong></p><p>The first time I conducted the card sort — grouping specific user actions to broader categories — it became obvious that I didn’t need a separate settings page. The second time, it was apparent afterward that I forgot I needed a card to represent the opportunity for users to compose a message. It was only with the third iteration that I finally got it right. That could change after I sleep on it though… And that’s perfectly fine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_P1AqUlZCxOTMlUc" /><figcaption>Blueprint for the Droom app</figcaption></figure><p><strong>The Central Lesson: Improvement Involves Trial and Error</strong></p><p>When designing the information architecture for a product, it’s okay to make mistakes. In fact, during the design cycle, it’s not only encouraged, but absolutely critical. As the understanding of your product develops further, your information architecture becomes more accurate. You have to experiment with what you know to discover what can be improved.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7570f3f4bc6c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bringing Your User into the Spotlight: the Power of Personas]]></title>
            <link>https://medium.com/@macrotentional/bringing-your-user-into-the-spotlight-with-personas-fec65c569210?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/fec65c569210</guid>
            <category><![CDATA[user-personas]]></category>
            <category><![CDATA[personas]]></category>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[user-research]]></category>
            <category><![CDATA[lambda-school]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Sun, 11 Aug 2019 17:46:01 GMT</pubDate>
            <atom:updated>2019-08-11T18:43:01.069Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/655/1*YylFV9aE5QypgEJj3wQkbw.jpeg" /><figcaption>Personas are valuable because they help you empathize with your users</figcaption></figure><p>A persona is a profile of the kind of person that your team envisions would be an ideal user of your product. Crafting a realistic portrait of your end-user can be a painstaking process. Even if you don’t have the time and resources to formulate a fully-researched persona developed through group interviews, behavioral synthesis, and motivation mapping, you can do the next best thing: prepare a proto-persona.</p><p>Before doign any research, I designed a proto-persona for <strong>Droom, a hypothetical mobile app aimed at jobseekers that combines the vocational focus of LinkedIn with the matching functionality of Tinder.</strong> Although there are people of all ages looking for a new job, I decided to narrow my preliminary analysis to folks in their 20s and 30s. Someone who graduated from high school, perhaps completed undergrad, worked an entry-level job for a few years, and now is interested in moving their career forward. I imagined younger jobseekers would be more willing to seek out fresh ways to find the right job — such as trying out a new app — and also be familiar with Tinder and its novel swiping interface.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FqF3mjD3mRNSbl8ym7JWXw.jpeg" /><figcaption>Meet Matt, the persona developed for Droom</figcaption></figure><p>Everyone has a story to tell, and that most assuredly also includes personas. After conducting a user survey, completing an interview, and investigating stakeholder goals, <strong>I created a persona for Droom named Matt Jackson. Matt is a 27-year-old from Milwaukee who studied zoology and works as a research assistant.</strong> After earning his degree he was happy to land a job near home, but now it’s getting a little stale and he’s disappointed there don’t appear to be opportunities to move up in his field unless he goes back to school. He feels unfulfilled in his current position, a research assistant role at his alma mater where he’s been employed in some capacity or another since graduation. Matt isn’t quite ready to take out student loans, and even if he did return for a graduate degree or PhD, he doesn’t know where he’d study.</p><p>Although we think of STEM majors as uniformly high-paid professionals, the truth is more complicated. Depending on your area of expertise, the results can be very different. In fact, <a href="https://www.forbes.com/sites/prestoncooper2/2018/06/08/underemployment-persists-throughout-college-graduates-careers/#52a818267490">a recent study reports biology has the second-highest number of underemployed graduates of any major, second only to psychology. </a>My understanding of the workforce is that the underemployed are naturally the most motivated to search for other possibilities in the job market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*MCPTH5ppMh_KhvXHaPurxw.png" /><figcaption>Contrary to popular belief, not all STEM majors are raking in big bucks</figcaption></figure><p>This premise inspired me to construct a persona who decided to follow his passion and major in zoology. He also minored in psychology, which didn’t particularly help either. It made sense to me for Matt to be working in Milwaukee because it’s a large metropolis with a sprawling university system that nonetheless is best-known for its manufacturing past, and as such offers limited career options for someone like him.</p><p>My persona is open to moving elsewhere to expand his job selection, but is reluctant to leave his girlfriend, who he met on Tinder two years ago, and has been seeing ever since. Indeed, she is also someone could benefit from an app like Droom. As a hard-working go-getter, Matt is someone in need of a service that can help him simplify the job search process, ideally including a feature to rank the positions where he’d be a good fit according to his own personal values.</p><p><strong>All in all, Droom is an app tailor-made for someone like Matt, and that is entirely by design.</strong> <a href="https://www.cbinsights.com/research/startup-failure-reasons-top/">The number one reason startups fail is not competition, but lack of demand for their offering</a>. User personas are invaluable to ensure a business considers the motivations, behaviors, and emotions of the user so that it can deliver a compelling solution that meets actual product-market fit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/844/1*YuzPOrMmol0yCYhxqLLisQ.jpeg" /><figcaption>Note: these aren’t the personas you’re looking for</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fec65c569210" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From User to Interviewer: a First Time for Everything]]></title>
            <link>https://medium.com/@macrotentional/from-user-to-interviewer-a-first-time-for-everything-c2716453604a?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2716453604a</guid>
            <category><![CDATA[user-research]]></category>
            <category><![CDATA[user-interviews]]></category>
            <category><![CDATA[transcription]]></category>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[lambda-school]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Fri, 09 Aug 2019 13:11:01 GMT</pubDate>
            <atom:updated>2019-08-09T13:22:25.499Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HTYtBn8QFCDrYX-LjXa8VA.png" /></figure><p>When it comes to interviews, my primary experience is being the person in the hot seat. If I’m being honest, nearly all of my encounters have been similar—speaking with a potential employer at a prearranged time, telling my story, explaining my skills, and doing my best to demonstrate that I’m qualified for the position.</p><p>Recently I had the chance to turn the tables and conduct my first user interview. It was neat to switch roles and discover what it’s like to interview others for once. Here are a few lessons I learned during the process:</p><p><strong>You Can’t Rush Good Work</strong></p><p>Although there wasn’t a job on the line, I still felt the pressure to live up to personal expectations. Instead of supplying compelling replies, my job was to prepare great questions. Since I’m interested in pursuing a potential career in UX research, I felt personally invested even before doing anything. No one ever told me that asking questions can be tougher than answering them. The fact of the matter is coming up with quality questions that you can count on to provide meaningful responses isn’t easy. It takes practice, and despite your best intentions, things can go awry. However, that’s not always a bad thing.</p><p><strong>Interviews Never Go According to Plan</strong></p><p>You can do all the preparation in the world, but you never quite know what to expect from a job interview. There are so many variables that you can‘t foresee. Will there be chemistry with the interviewer? Will your personalities mesh or clash? Will they be tired and moody? Peppy and satisfactorily caffeinated? Will <em>you</em> be tired? Will you have downed too much coffee beforehand? The same is true during a user interview. There are unpredictable social dynamics at play, and neither of you quite knows where the conversation will lead. In the interviewer’s case, questions you thought would be useful can end up upstaged by follow-ups to intriguing comments by the participant. You have to adapt to the occasion and trust yourself to keep the focus on the issues there are most salient to what you’re trying to learn about your users.</p><p><strong>Watch Out for Bias</strong></p><p>Pay close attention to inadvertently introducing leading questions. This is particularly a challenge when you know the participant. It’s so easy to relax, lapse into agreeable banter, and talk like good friends instead of maintaining a professional demeanor. To avoid accidentally interjecting subjectivity into the process, it’s best to only interview people with whom you’re unfamiliar. In my situation, I needed to find someone who could serve as a model user, and even though we knew each other, they were the best option given the constraints at the time. In my case, the benefits of gaining critical qualitative data outweighed the issue of occasional opinionated framing of questions.</p><p><strong>After the Interview</strong></p><p>I have to admit it feels great to finish an interview. However, when you’re a user researcher, much of the real work takes place <em>afterward</em>. The conversation itself only lasted 20 minutes. Accurate transcription of a recording takes much longer. I spent several hours carefully replicating our exchange. <a href="https://www.descript.com/">Descript</a> is a surprisingly useful tool for helping you manage this process, but you still need to listen for small, quiet bits that software doesn’t pick up. Then there’s also the time you have to invest searching for patterns, synthesizing your findings, and identifying themes. Despite my errors and unimpressive delivery, it’s amazing how much useful information you can glean from a relatively short dialogue focused on understanding the values and pain points of a potential user.</p><p>The biggest takeaway for me: over the course an interview, there’s SERIOUS untapped potential for using data analysis to detect patterns that are not otherwise apparent<strong>.</strong> Are professional user researchers already coding their data this way? Given my practice with SQL and Python, simplifying my workflow with data analysis techniques is something I definitely need to try out next time!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Jm9jvgDu2tDbhAJGwUK1hQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2716453604a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Responsive Design: Serving Up User-centered Fluency]]></title>
            <link>https://medium.com/@macrotentional/responsive-design-a-lesson-in-serving-up-user-centered-fluency-2305baaa155a?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/2305baaa155a</guid>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[lambda-school]]></category>
            <category><![CDATA[prototyping]]></category>
            <category><![CDATA[responsive-web-design]]></category>
            <category><![CDATA[responsive-design]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Fri, 02 Aug 2019 20:44:28 GMT</pubDate>
            <atom:updated>2019-08-02T22:39:36.498Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ovEhSJpaE7VKmkgXs1uuTg.jpeg" /><figcaption>Well-done responsive design and fresh waffles: the same warm feeling</figcaption></figure><p>There’s a theory that the simpler and more intuitive a design appears, the more thought and resources were invested to produce it. Does this maxim also hold true for online work? In most cases, the answer is yes. When responsive site navigation accommodates the goals of users and those of site owners across platforms, that’s effective user experience design in action.</p><p><strong>The Project: a user-friendly recipe sharing experience</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lJ8-7DftPVdGjYwlF4jAVg.png" /><figcaption>The logo, using the Nunito typeface</figcaption></figure><p>Over the past few days, I’ve been flexing my responsive design skills, creating wireframes and prototypes for a hypothetical client interested in launching a recipe sharing web app called Foodreau. The target audience enjoys cooking and collecting recipes, but has a hard time keeping track of them and sharing their favorites with distant friends and family. Considering I was only given a week and many of the steps in this exercise were learned on the fly, I’m relatively satisfied with the finished product.</p><p><strong>The Process: from drawing board to functional prototype</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FX8GXhjeUKpa6lyf5G-h-g.png" /><figcaption>My Foodreau mobile wireframes</figcaption></figure><p>I imagined myself as a UX designer for Foodreau, and was fortunate to have many of the background details for this project prepared by my client in advance. This was more of a test of my design skills and time management than copywriting or conceptualization ability. With a focus on creating a simple and inviting user interface, I began by working on wireframes for the mobile version of the web app. Once those were completed I moved on to the desktop wireframes. The windows, graphics, and design elements were refined in the prototype stage. After the prototypes were finished, <a href="https://invis.io/DETA0Q5A6KQ">I simulated the user navigation on InVision</a>.</p><p><strong>1. The Landing Page: sparking a strong first impression</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*0xgt-gC3WTtIpPdgF2HJzA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4qKBOlWMduGINrfKFaXgVQ.png" /><figcaption>Foodreau: helping you cook up memories with friends and family™</figcaption></figure><p>The emphasis on the landing page is to provoke the curiosity of visitors and prompt them to join the site. With this intent in mind, the spotlight is on the buttons encouraging potential users to create an account. Accordingly, “Let’s get cooking!,” “Join Free Plan,” and “Join Pro Plan” each link to the Sign Up page.</p><p>The unique value proposition is featured above the fold to ensure people understand what Foodreau is all about. Below are headers outlining the top three selling points of the site. To help visitors relate to the service and envision themselves as members, several testimonials are included. As users scroll down the page, there is a natural progression of interest that culminates in a call to action to join Foodreau. Two options are presented: an ad-supported free plan, and a premium option that allows users a greater degree of influence by offering them unlimited sharing of recipes.</p><p><strong>2. Sign In: inspiring loyalty</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JPWzoBOUKgTak_ckAUeMvw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*FEugPUAUrnWIiX-mBTlkOg.png" /></figure><p>The sign-in screen is streamlined to promote user logins. The additional horizontal space on desktop lends itself to a few lines reminding users of why they signed up in the first place, relying on the instinctive drive to explore to build anticipation for the potential of Foodreau to fulfill their hunger for new experiences.</p><p><strong>3. Sign Up: promoting onboarding</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*FDdOQtIITS5Q8ECXGKDMsw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dGUFNePIm-UUqbsREbX62A.png" /></figure><p>The sign-up page is configured in much the same way as the sign in. In both mobile and desktop, a testimonial was featured to boost relatability by appealing to the desire to connect with relatives. The extra real estate for the desktop version came in handy again, this time allowing for another line about what makes Foodreau appealing. Including the photo in the mobile prototype was considered, but with the limited screen size was regarded as distracting from the primary objective of filling out the sign-up form.</p><p><strong>4. Recipes List: tantalizing the taste buds with endless possibilities</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bfLDpAm1gMeXn9mOzxVZ2Q.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*XPKk5Uambhveu1xqA4kGuA.png" /></figure><p>This screen was *definitely* the most technically challenging to produce. With so much visual information to convey, layering and styling the attributes took a great deal of persistence. I learned a lot about masking, spacing, alignment, and the importance of proper ordering to ensure the right objects are visible to the viewer. The mobile prototype necessitated a narrower organization of two columns of recipes. With the greater input precision inherent to a desktop design, I felt comfortable adding the ability to delete, edit, or share straight from the recipes list.</p><p><strong>5. Sample Recipe: let’s get cooking!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/375/1*nHNXpmxhgnYA6dsIMeN4bA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ODAgA-HfYQE38xbEDny2dA.png" /></figure><p>The end in sight, at this point we’re finally at the kind of page of this web app where people are likely to spend the most time. Here we have the screens for an individual recipe. At this time of year, a peach cobbler seemed ideal (full disclosure: the bag of peaches sitting in my fridge waiting for the right opportunity to escape may have influenced my decision-making). It was tough to fit all the text on mobile, and unfortunately, this is a problem that we find on virtually any website with recipes. On desktop there is plenty of room to display the ingredients and directions simultaneously. Delete, edit, share, and print buttons are present on both, but on mobile they’re positioned further apart to account for the reduced accuracy of touch targets.</p><p>If you’re interested, feel free to <a href="https://invis.io/DETA0Q5A6KQ">take a closer look at my user flow.</a> All feedback is greatly appreciated!</p><p><strong>Afterword: user flows, communication, and design fluency</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FjM5KyfX12zb62aDoxyU8A.jpeg" /></figure><p>As toddlers can attest, If you’re immersed in a language, it only takes a few years to figure out the basics. On the other hand, to convey complex ideas, reach cross-cultural understanding, and finely develop your interpersonal skills can take a lifetime. After all, even adults speaking in their native tongue sometimes struggle.</p><p>When it comes down to it, adapting a design from mobile to desktop is a lot like the process of becoming fluent in another language. Before you can communicate effectively, you need to have a solid grasp of the grammar mechanics and vocabulary of the language you’re aiming to learn. Once you have those down, you can reproduce your feelings and get the general meaning across to others.</p><p>Similar content, just a different voice and style. Mobile and desktop interfaces aren’t worlds apart, but when crafting a message, it pays off to take into account the constraints and opportunities each offers for design professionals to achieve user-centered fluency.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2305baaa155a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[UI Design and Gestalt Theory: Reddit Revisited]]></title>
            <link>https://medium.com/@macrotentional/investigating-gestalt-principles-reddit-revisited-d606aa304d0c?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/d606aa304d0c</guid>
            <category><![CDATA[user-interface-design]]></category>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[reddit]]></category>
            <category><![CDATA[gestalt]]></category>
            <category><![CDATA[lambda-school]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Thu, 18 Jul 2019 00:45:09 GMT</pubDate>
            <atom:updated>2019-07-20T19:51:54.114Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IcYMmBlnzx8SRYCEhgOy4A.jpeg" /><figcaption>No need to connect the dots. Gestalt Theory reveals why you can still see the Reddit alien</figcaption></figure><p>As humans, we respond to visual information better than any other type of stimuli. By some estimates, <a href="http://www.t-sciences.com/news/humans-process-visual-data-better">90% of the data we process is visual.</a> To make sense of it, we’re heavily inclined to perceive patterns in the optical input we absorb.</p><p>Many of the cognitive shortcuts we use are summarized by the laws of Gestalt Theory, which are also known as the psychological principles of grouping. Although our understanding of cognitive perception has matured dramatically since the Gestalt principles were first proposed in the late 1800s, they continue to be useful as guidelines for visual design.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*MBPNEUTwBqKZ0Bo507SVHg.jpeg" /><figcaption>The Gestalt Principles: <a href="https://en.wikipedia.org/wiki/Principles_of_grouping">https://en.wikipedia.org/wiki/Principles_of_grouping</a></figcaption></figure><p>Yesterday I examined Reddit through the lens of visual hierarchy. Since I couldn’t think of another site, I figured I might as well choose it once again, this time deconstructing its design with the Gestalt principles.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wDAzem6McpwKyA6cGIB8ug.png" /><figcaption>My Reddit homepage today. Looks a lot like yesterday’s</figcaption></figure><p><strong>Proximity</strong></p><blockquote>“Objects or shapes that are close to one another appear to form groups.”</blockquote><p>Proximity concerns how groups of elements are arranged together. When objects are close, we perceive them as united; when distant, they are interpreted as disconnected. On the Reddit homepage, there are multiple examples of proximity. Each update in the feed is adjacent, forming an orderly column of endless distraction from whatever you’re supposed to be doing. In addition, we can see this principle in the list of suggested trended communities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*muuAtM7VvcmvBqWXt9m5Ww.png" /></figure><p><strong>Similarity</strong></p><blockquote>“Objects that look similar appear grouped, all other things being equal.”</blockquote><p>When the characteristics of elements strongly resemble each other, they are mentally grouped. In the case of Reddit, we again are drawn to the update feed, as each story takes up a similar amount of space. In user interface design, this is not unexpected, since the principles of similarity and proximity often work together to form a visual hierarchy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*UNMk4OvDWGHdqYCJpo2x-g.png" /></figure><p><strong>Continuity</strong></p><blockquote>“Our visual perception is biased to perceive continuous forms rather than disconnected segments.”</blockquote><p>When a design, contour, or outline is incomplete, our minds naturally fill in the gaps to assemble a complete impression. Continuity is evident in the upvote and downvote arrows that allow people to raise or lower the profile of a story. Despite not being linked, the user can infer that they are related by the number between them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ISRPjCUUGI0yXGrhHRDy3w.png" /></figure><p><strong>Closure</strong></p><blockquote>“Related to Continuity is the Gestalt principle of Closure, which states that our visual system automatically tries to close open figures so that they are perceived as whole objects rather than separate pieces.”</blockquote><p>On this page, closure and continuity are analogous. We see the arrows, we connect the arrows, we understand they’re a way for the community to assess the value of a story that’s reflected in the upvote total associated with it. In short, people have a tendency to connect the dots.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mRUe2Iy-Y0c9filasF9G7Q.png" /></figure><p><strong>Symmetry</strong></p><blockquote>“The data in our visual field usually has more than one possible interpretation, but our vision automatically organizes and interprets the data so as to simplify it and give it symmetry.”</blockquote><p>Symmetry is often used to represent 3-D objects on a 2-D plane like a screen. The search bar is situated at the top center of the page, and the feed is roughly in the middle, but otherwise there’s not a whole lot going on with symmetry on Reddit. This is likely intentional, as the primary intention of the user is to scroll through the headlines until they see something that warrants closer inspection.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3tlB-7yG0vs4i5Z1snUVSQ.jpeg" /></figure><p><strong>Figure/Ground</strong></p><blockquote>”Our mind separates the visual field into the figure (the foreground) and ground (the background). The foreground consists of those elements of a scene that are the object of our primary attention, and the background is everything else.”</blockquote><p>Figure and ground can be employed to express layers or give a design a sense of depth. No doubt about it, the focus here is on the updates in the front and center. With a text-heavy site like Reddit, there’s little reason to complicate things further by introducing additional elements in the background.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fh-TnVO95CeKmOTSXNGqAQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ezWGIa50bok_WH7Xm4LEPA.png" /><figcaption>The rule of figure/ground in action, as seen on <a href="https://theoutline.com/">theoutline.com</a></figcaption></figure><p><strong>Common Fate</strong></p><blockquote>”The Common Fate principle affects whether we perceive objects as grouped. It states that objects that move together are perceived as grouped or related.”</blockquote><p>A group of objects moving in the same direction at a similar speed is categorized as a group. Birds of a feather flock together. Reddit.com is one the most-visited URLs on the web, but much like Craigslist, it has refused to radically reinvent itself, instead sticking to a familiar tried-and-true user interface that has served it well over the years. Other than the occasional user-generated GIF, you won’t find many dynamic elements on the site.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Nd0sXyJ1yguWIY0-uRKRRw.jpeg" /></figure><p><strong>The Lesson for UX Designers</strong></p><p>As the success of Reddit shows us, you don’t need to incorporate all of the Gestalt principles to produce a highly engaging experience. In fact, you’re probably better off only using a few of them. When users visit your site or download your app, first impressions are everything. Don’t scare them away with cluttered visuals or disjointed elements.</p><p>Less is more.</p><p>At the end of the day, what really attracts an audience is a community where they feel they belong and quality content that reflects their passions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/896/1*-mc1UVsDkKZxvTVi2_UTZA.jpeg" /><figcaption>Reddit is built on the dedication of its users and the stream of fresh content they submit</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d606aa304d0c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Reddit Can Teach Us About Visual Hierarchy]]></title>
            <link>https://medium.com/@macrotentional/what-reddit-can-teach-us-about-visual-hierarchy-1bf9b530e5c5?source=rss-6ef92153be2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/1bf9b530e5c5</guid>
            <category><![CDATA[personalization]]></category>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[reddit]]></category>
            <category><![CDATA[feeds]]></category>
            <category><![CDATA[visual-hierarchy]]></category>
            <dc:creator><![CDATA[Martin Campbell]]></dc:creator>
            <pubDate>Tue, 16 Jul 2019 22:15:08 GMT</pubDate>
            <atom:updated>2019-08-02T22:40:38.248Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1015/1*2zcAiDYGujaAmLjkqwl-bQ.png" /><figcaption>You get an upvote, you get an upvote, everybody gets an upvote</figcaption></figure><p>What makes a site easy to navigate? What compels us to return time after time? Aside from personal interest in the featured content, a clear value proposition and strong visual hierarchy go a long way in developing user loyalty. Today I chose to take a closer look at what is possibly my favorite site, Reddit. Not just a news site and not quite a social network, it fits in somewhere in between. For better or worse, I often find myself visiting the web’s most popular news aggregator community multiple times a day.</p><p>The visual hierarchy of Reddit is somewhat unique because signed-in accounts have the ability to personalize the content shown and how it’s exhibited. The stories and images displayed depend on which specific topic-based communities (subreddits) you follow. In addition, in the top-left corner you’re able to change how the main page displays these updates by selecting the dropdown menus View (to arrange according to Card, Classic, or Compact) or Sort (to filter by Best, Hot, New, Controversial, Top, or Rising).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*br2KWBpPmHr4N9UQ" /><figcaption>My Reddit homepage: featuring Pokémon Go, memes, and more memes</figcaption></figure><p>At first glance, Reddit is an unusually visually dense site. There’s a ton of information to absorb, no clear central focus, and to users unfamiliar with its interface, it might appear unnecessarily cluttered. When examining the homepage while squinting, the standout elements are the contrast-driven buttons on the right side encouraging you to create a post or try Reddit Premium. Its unique value proposition is positioned just above: “Come here to check in with your favorite communities.” However, what <em>really</em> inspires people to visit are the tailored updates, which are emphasized by the prominent use of white space that attracts your eyes to the visuals and text featured on the left.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RZM7rTssK7AKJlanWqE80A.png" /><figcaption>Hooking new users with custom news feeds since 2005</figcaption></figure><p>My intentions when visiting Reddit, coupled with my typical interactions with its contents (such as endorsing content I like through upvoting, or opening links to stories from headlines that capture my attention), ensure the right side of the page is nearly always ignored. Despite this banner blindness effect, I’m not surprised that Reddit is organized in this manner. The site’s bread and butter is a continuous stream of fresh content, so it makes sense that it would urge you to create a post to add your own contribution to keep the content farm running. The promotion of Reddit Premium is also easily understandable because it is the most cost-effective way for the owners to generate revenue.</p><p>As we can see, Reddit operates according to an idiosyncratic visual hierarchy that’s heavily influenced by the endless update feeds with which most internet users are already well-accustomed thanks to the dominance of Facebook, Twitter, and Instagram. Sometimes you have to break the rules to stand out. In the case of Reddit, which <a href="https://wersm.com/sorry-facebook-reddit-is-now-the-third-most-popular-site-in-the-us/">briefly overtook Facebook in US traffic</a> and sincerely bills itself as “the front page of the internet,” this risk has clearly paid off.</p><p>The cute alien mascot also doesn’t hurt.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YWsJvv08-9aKcaY0os1XlQ.jpeg" /><figcaption>Snoo, mascot in residence on planet Reddit</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1bf9b530e5c5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>