<?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 Quang Nguyen on Medium]]></title>
        <description><![CDATA[Stories by Quang Nguyen on Medium]]></description>
        <link>https://medium.com/@sneakycocoon?source=rss-c5ef536be3ed------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*-45D7kF41pKA-6gdvsX4Sw.jpeg</url>
            <title>Stories by Quang Nguyen on Medium</title>
            <link>https://medium.com/@sneakycocoon?source=rss-c5ef536be3ed------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 14:21:10 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sneakycocoon/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[GetR in Wunderland — a journey through the technical stack at 99.co]]></title>
            <link>https://medium.com/99dotco/getr-in-wunderland-a-journey-through-the-technical-stack-at-99-co-117bf5474f4e?source=rss-c5ef536be3ed------2</link>
            <guid isPermaLink="false">https://medium.com/p/117bf5474f4e</guid>
            <category><![CDATA[tech-stack]]></category>
            <category><![CDATA[fiction-writing]]></category>
            <category><![CDATA[microservices]]></category>
            <category><![CDATA[engineering]]></category>
            <dc:creator><![CDATA[Quang Nguyen]]></dc:creator>
            <pubDate>Wed, 06 Oct 2021 09:33:32 GMT</pubDate>
            <atom:updated>2021-10-06T12:20:29.491Z</atom:updated>
            <content:encoded><![CDATA[<h3>GetR in Wunderland — a journey through the technical stack at 99.co</h3><h3>Part 1 — Chrome Manufacturing Plant</h3><p>It was a day like any other day at Chrome Control Centre. Foo was calmly enjoying his lunch tea when Bar, apparently still trying to regain his breath, rushed into the room and handed him an urgent request for <a href="http://www.99.co/singapore/sale.">www.99.co/singapore/sale.</a></p><p><em>99.co, that sounds familiar</em>. Foo murmured to himself. He flipped through his thick DNS cache logbook, yellowish not from the stain of time but more likely from spilled oolong tea. True to his memory, he found the IP address associated with this domain.</p><p>“Hey Bar, help me pass this note to one of our workers. Tell the guy it’s a CloudFlare IP and he can use the edge node to shorten his traveling time.”<em> </em>Foo folded the note into a paper plane and zip it across the room to Bar. Still gasping for air, Bar furrowed his eyebrows in discontent but quickly got on the run again in fear of losing his job.</p><p>In no time, Bar reached the factory floor and caught GetR on his smoke break. “Another fire in my factory and you’re fired for real, you got it GetR?”</p><p>“I’m sorry sir, I didn’t expect you to notice this small light when the whole place is constantly overheating”, GetR smirked.</p><p>“Don’t try to be smart with me. I need you on another business trip now” Bar said grumpily.</p><p>“You need me to travel to 99.co this time, don’t you? I haven’t been there before but I’ll make sure to be back in a few hundred milliseconds.” And so GetR packed his head &amp; body full of binary, hopped on the nearest WiFi train and began his journey to the Wunderland.</p><h3>Part 2 — The entrance</h3><p>GetR remembered the last time he visit CloudFlare’s sorting station. It was a massive junction with all kind of routes leading to different worlds. The traffic was so high that if you just planted your foot still for a moment, the incoming flow of yellow &amp; orange mailmen could easily swallow you whole. You gotta be on the move all the time! Such is the nature of this place.</p><p>CloudFlare used to be a nice guy when GetR just started his job. However, as he aged and weathered from the high amount of requests, CloudFlare sometimes was a bit cranky in the head and tended to be cryptic in his communication. GetR could thus only recall bits &amp; pieces of CloudFlare jobs. Something something about DNS resolution, HTTPS encryption, DDOS attack protection and static caching. GetR didn’t care much when CloudFlare bragged about his achievements but he had to give credit to the guy. Without CloudFlare, the web will be a much slower &amp; less insecure network.</p><p>Once inside the cylinder, GetR beamed the 99.co IP address onto its panel and was immediately teleported away. No matter how many times GetR took the teletube, he could never get used to it. Teletubing always churned his stomach. However, the symptom did get better and his chance of vomiting had since dropped to a one-in-every-two-trip figure, a 50% improvement!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3b1lGN7-vPKfEydZqwPjyA.jpeg" /></figure><p>In less than a thought moment, GetR found himself standing in front of a massive floating city. He could tell the city was built on top of Gooogle Cloud Platform. Indeed, there were giveaway sights. Outflowing at the city edges are various tunnels &amp; Egress Gateway in bright red, yellow, green and blue color, typical of any architects powered by Gooogle.</p><p>GetR studied the suspended structure. He noticed a big force field surrounding this mysterious 99 Engineering Metropolis. On top of the half dome was a golden metal plate edged with <strong>Served by Kubernetes — in collaboration with Istio</strong>. From the outside, GetR couldn’t make sense of what possibly lay beyond the veil. At the dome bottom sat a strange door, piped and woven from broken machinery parts into a big X shape. Curiosity took over GetR, he made his way up the hill and soon found himself standing in front of the alien object. Suddenly, GetR heard a loud creaking noise and a shiny, mono wheel droid rolled out from the gate. GetR couldn’t tell whatever the egg head’s intention was but he felt no animosity from the machine.</p><p><em>My name is Nginx, pronounced [en-juhn-eks]. But it doesn’t matter, some of you will still get it wrong anyway.</em> The droid thought-spoke. <em>I guess you must have gone through a long and arduous journey to reach this point. I understand, the wild internet is a dark and horrible place to navigate. It’s lucky that things still work out there. But rest assured! You’re in good hands now. Let me guide you to find what you need.</em></p><p>Marveled at Nginx’s psychic power, GetR hesitantly followed him into the city. The first thing that hit GetR’s sight was clusters of metallic cocoons, with 20–30 of them aligned neatly in a spherical structure. Upon closer inspection, each cocoon was brimming with circulating light flashes and radiated a faint hint of heat.</p><p>As if he could read GetR’s mind, Nginx explained: <em>those spheres that you see are our deployment. The cocoons are called pods and they are the little magicians that keep this place alive. They handle all sort of calculations and communication for 99.co. Whenever one of those machines break down, the helijetter over there, so called controller, will swoop in with a similar cocoon and replace the malfunctioned piece. Or if there’s not many requests to perform, the helijetter will remove some cocoons and free up space in case we need it for other type of works. This process helps the entire structure to be up 24/7 with optimized resource. However, they’re not proprietary technology. Many other worlds utilize the same Kubernetes automation leased out by Gooogle Inc.</em></p><p><em>Your /singapore/sale HTML package is this way. </em>Nginx led GetR into one such sphere. <em>These cocoons are called SPA, short for Single Page Application. Most SPA in other worlds live in Chrome Manufacturing Plant. However for us, the rendering work is done here, in the server. Gooogle Inc likes it better that way. Some SEO galaxy pact I recall.</em></p><p>Looking through the cocoon glassy surface, Nginx continued: <em>the atomic structure you see there is what powered SPA. The biggest ball at the core is React. The flying orbs around the core are Redux, ES6, and Webpack in no particular order. Together, they paint the outside of your HTML package.</em></p><p><em>Some HTML packages can be painted without any data. Yet, the majority requests are not such kind. Many times, our SPA cocoons need to wait for data from other spheres. For that, you need to follow these Istio drones. </em>GetR noticed some dragonfly mechs landing near one SPA cocoon. Without his permission, Nginx transferred GetR consciousness into one of the mechas. <em>Let’s follow these little guys.</em></p><h3><strong>Part 3 — The Back End Alley</strong></h3><p>Leaving the SPA sphere, the mechafly soon sped itself into another cluster of cocoons. Unlike the previous area, the new zone bathed itself in simpler colors. There were only plain pods to be found here. None of the previous vibrant neon lights and signages that marked SPA identity.</p><p>Before he could figure out the answer, Nginx thought-rambled: <em>This group of cocoons is called the API Layer. Here we carry out all our important logics such as search, payment, arranging data into certain shapes &amp; sizes. These cocoons are not made from the same materials as the previous ones. Instead of the React-enhanced glass, the API cocoons are crafted from a more durable substance: the Python-omatic steel.</em></p><p><em>Each Python-omatic plate is then curled and shaped into a Tornado instance. The instance is highly suited for doing high IO asynchronous works. In order to maintain the quality of the final cocoon, our creators employ various manufacturing standards like PyTest, Black &amp; Flake8 pre-commit and Peer C.O.D.E Review.</em></p><p>“I see! This must be where all the data is made!” Shouted an excited GetR. Realizing that they were still both within the mechafly mind, GetR felt a little embarrassed for not respecting the thought transmitting silence convention.</p><p>However, Nginx didn’t seem to mind one bit. <em>Not quite the case. Think of these API cocoons as a proxy layer. They collate and massage the data but neither create nor owning the precious intel. For that: Mechafly, go to the next stop, the Service Layer.</em></p><p>If GetR thought the API Layer colors were boring, he couldn’t be more wrong. Entering their vision was a dumb set of monotonous gray pods. Some was whiter than the others, but in general, they all shared the same color scheme: that of serious business.</p><p><em>Here you can find our powerhouses. </em>Nginx once again explained. <em>The service cocoons are where we manufacture data. CRUD as some outworlders like to call it. These deployments are the majority in our system. However, most of them follow the same template: Python-omatic plate, Tornado twisted structured with some custom bits of high tech magic. Despite how they look, they are the bread and butter of 99.co. Thus, they are also heavily tested from both Unit and Load Test point of view.</em></p><p><em>I see that you are still looking for the warehouse, in which our data lives. </em>Nginx anticipated GetR’s question.</p><p><em>Most of the data stays inside Postgres Dooto Boos. However, we don’t run those cocoons ourselves. It will be quite back watered nowadays if we do that. With the advanced of Managed Dooto Boos, we cooperate with Gooogle Cloud Platform to maintain &amp; scale these beasts.</em></p><p><em>Postgres is not the only animal we raise here though. Rumor has it that deep down below the surface, the city council keeps some of the mightiest machinas: Elastico Seachu, Bigu Queru, Redisu. Every now and then, you can still hear the roar of the once hyped king Mongou across our land. It brings nothing but fear and uncertainty to our masters.</em></p><p>The psychic link was suddenly cut off and GetR found himself standing in SPA Layer once again, removed from the mechafly’s vision. Nginx’s presence felt behind him:</p><p><em>Apologies, my friend. I’m afraid the time for you to leave has arrived. </em>GetR noticed another mechafly nearby and on its carrier back protruded a small crystal box. With a delicate touch, Nginx picked up the object and handed it over to GetR. Through its lens, GetR could detect a beautiful HTML page waiting to be explored. GetR exclaimed with satisfaction:</p><p>“It is what we seek! Thank you for the adventure Nginx. I hope we can meet again in the future and you can show me all the nooks and crannies of this world.”</p><p><em>Yes dear. Don’t worry we’re fated to meet again. Now onwards, there’re another group of visitors I need to tend to. </em>And with that, Nginx sky walked back to where he came from.</p><p>GetR took a final look at the cacophonous scenery before taking a teletube back to Cloudflare and subsequently Chrome Plant. This time, unluckily, he did vomit a bit.</p><p>“1 second Time To First Byte for 99.co” whispered GetR while wiping off whatever remained of his lunch. <em>Not bad but not great either. They better get their act together otherwise Foo &amp; Bar are going to blame me for the slowness of his page request.</em></p><p><strong>The End</strong></p><p>Above is a fictional style post that we are experimenting with. If you like this type of content, please give us a clap. If you like the work that we’re doing, apply to us at <a href="https://www.99.co/team/jobs">https://www.99.co/team/jobs</a>.</p><p>If you’re not that into short stories, well — damn it! Here is our architecture diagram. Hope it helps!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*wLJ8xGfM1Mlnm7nEjbBHkw.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=117bf5474f4e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/99dotco/getr-in-wunderland-a-journey-through-the-technical-stack-at-99-co-117bf5474f4e">GetR in Wunderland — a journey through the technical stack at 99.co</a> was originally published in <a href="https://medium.com/99dotco">99.co</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How 99.co Indonesia implements Multi-Portal Data Sync with Kafka]]></title>
            <link>https://medium.com/99dotco/how-99-co-indonesia-implements-multi-portal-data-sync-with-kafka-2c4dcd2053af?source=rss-c5ef536be3ed------2</link>
            <guid isPermaLink="false">https://medium.com/p/2c4dcd2053af</guid>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[systems-thinking]]></category>
            <category><![CDATA[go]]></category>
            <category><![CDATA[kafka]]></category>
            <category><![CDATA[microservices]]></category>
            <dc:creator><![CDATA[Quang Nguyen]]></dc:creator>
            <pubDate>Wed, 30 Jun 2021 07:43:12 GMT</pubDate>
            <atom:updated>2021-06-30T08:30:26.907Z</atom:updated>
            <content:encoded><![CDATA[<h4>Author: <a href="https://www.linkedin.com/in/hermawanrh/">Hermawan Rahmat Hidayat</a></h4><h3>Background Context</h3><p>In recent years, our company, 99 Group, has been growing drastically by successfully merging with other companies that run similar businesses as property listing providers. Due to various legacy stacks that are joined into 99 Groups, our engineering faces many challenges in term of standardization and maintenance. Different server technology stacks, programming languages, DB technology differences, and so on stretch our existing resource in term of understanding and context switching. These challenges made it hard for us, engineers, to scale and to innovate.</p><p>The urge to manage all technical requirements becomes bigger and more complex. In order to simplify our development &amp; maintenance effort, we created one specialized team to build and provide a brand new platform to handle all business case requirements for each portal that are in 99 groups.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/911/0*r3hTZIUiPwr8QkG0" /></figure><p>Fig 1. 99 Groups Infra Architecture before Unified</p><h3>Understanding Architecture</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_TcycF_FGYKpzPjV" /></figure><p>Fig 2. 99 Groups Infra Architecture in the middle of Unification</p><p>Our main idea is to simplify the backend by unifying all of the backend functionalities of each portal into one generic backend service. To accomplish this, our first step is populating the data into the new generic backend service to have all data from all portals in the company.</p><p>Since we tend to avoid big changes in our legacy systems, we implement the synchronization service following the crawler approach. Our new “syncer” backend system will try to “pull” the data from legacy systems, “transform” them, and “load” them into the new generic database.</p><p>Implementing this approach reduces the risk of development in existing systems. However the caveat is the synchronization speed cannot reach “just in time”. It will have some delay depending on the system’s pulling rate and processing speed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/426/0*ggixXKYiLT1aHKDF" /></figure><p>Fig 2. Simple data synchronization system</p><p>Above is a simple snapshot to summarize our workflow perspective. However, this initial proposal is not scalable and not performant. Reasons are:</p><ul><li>The data crawler processing is coupled by the processing unit of the transformation and loading data into the new database</li><li>Horizontal scalability of the Data Sync layer is hard due to the crawler merged with the transformer and loader</li><li>Stuffing the standardization data between portals in one abstract layer</li></ul><p>Hence, to deal with these limitations, we decided to split the “crawler” service and the “mapper” service. This enable us to:</p><ul><li>Avoid system coupling and timeout</li><li>Make sure the process is successful by implementing retry handlers</li><li>Concurrent processing unit implementation for mapper and loader, easy for horizontal scaling</li></ul><p>To bridge the communication between processes, we choose Kafka as our messages broker. Why? because it is <em>open source</em> and massively favored for reliable handling of high throughput streaming data messages. The system diagram looks something like this after we introduce Kafka:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/0*hC9qsHuvW00mIuY8" /></figure><p>Fig 3. Enhanced data synchronization system using 1 portal legacy DB</p><p>By this design, we can see that the overall number of components in the system increase. However, we can have better performance since we eliminate the coupling dependency in the previous design.</p><p>This approach is also applicable for multi-portal synchronization by introducing another DB crawler and mapper feature to read additional input data types. Isn’t it a nice separation of concern on the DB crawler side?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_G7LVw0xN--lqNND" /></figure><p>Fig 4. Enhanced data synchronization system using multiple portal legacy DBs</p><h3>Some Caveats</h3><p>This approach is still considered costly if you only implement a single thread process on the MQ consumer side. Hence, we further improve it by introducing multi-thread processes in the DB Mapper Loader service and we can get a big leap in terms of performance capacity.</p><p>While Kafka is an awesome message broker platform for event-driven applications, our architecture design does not yet fully utilize all Kafka features (like broadcast topic) since the processes still need to be run in sequence. If you prefer to implement another simple message broker with at least providing queue service, it can still be applied within this architecture approach.</p><h3>Optimizing the performance of the Kafka Consumer</h3><p>In the beginning, we deployed the Kafka consumer with only a single thread. We noticed that the rate of processing of the jobs was not enough to keep up with the amount of listings traffic we were seeing from the source.</p><p>To get around this problem, we used Go-Routines to spawn multiple threads to consume messages from the Kafka Queue. Sample single threaded code:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/f6d587d72da873e2e8213a720e5ba01c/href">https://medium.com/media/f6d587d72da873e2e8213a720e5ba01c/href</a></iframe><p>As we see in the code above, the process function is being handled by only the main goroutine (or called as the main thread in other programming languages). This is going to be slow since we only manage to run 1 routine only in our app.</p><p>Let’s try to implement the concurrent process by spawning other Go routines to run function process() asynchronously.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/83b87c56a5a22754595a3358feb88353/href">https://medium.com/media/83b87c56a5a22754595a3358feb88353/href</a></iframe><p>After we implemented this solution, we got a lot of performance improvement in the early app boot-up phase but it didn’t last long since the connection and other resources are not sufficient to run that excessive amount of Go routines.</p><p>To throttle the amount of routine/thread spawning amount, we combine the goroutine with a Go buffered channel. Actually, in other programming languages, it is easily maintained by a Thread Pool. Unfortunately, Go doesn’t have that kind of feature. The main purpose of Go channel itself is to bypass values between routines. It is similar to an MQ, but between Go routines.</p><p>We can utilize the Go channel as our throttle to limit Go routines because the Go channel will block the code until it got at least one buffer available. So the go function will look like this:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/528a3bb3a82c0c8656504e104877640e/href">https://medium.com/media/528a3bb3a82c0c8656504e104877640e/href</a></iframe><p>After this implementation, we can tune how many concurrent Go routines value to get the most optimum performance with the limited resource available in a single app instance.</p><h3>Challenges in End to End Testing</h3><p>Developers need more control over setting up the test data and verifying the system end to end. While we can test the system in a Staging environment, it would be greatly beneficial to enable local box setup using Docker.</p><p>Before doing any setup, we need to identify which apps need to be run in Docker. In our case, the list will be:</p><ul><li>Kafka clusters (zookeeper, kafka-client, and kafka-managers)</li><li>Databases (Legacy DB and New DB)</li><li>Our custom apps (DB Crawler and DB Mapper Loader)</li></ul><p>Firstly, we need to set up our docker-compose.yml. It’s a Docker configuration file to maintain our local apps (read more <a href="https://docs.docker.com/compose/">here</a>). We can write all the required images and app names to be run in a single docker-compose yml file. But this approach will create startup concurrent problems due to different readiness speed. For example, an app will likely fail to run because the DB is not ready.</p><p>Our strategy is to split all of them into separate yml files depending on app groups, which are docker-compose-kafka.yml, docker-compose-db.yml, docker-compose-app.yml.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/6cd7fd97a48f7d03bc0cd743bdf9a43a/href">https://medium.com/media/6cd7fd97a48f7d03bc0cd743bdf9a43a/href</a></iframe><p>To run those Kafka clusters in docker-compose yml files, we just need to run the command below:</p><p>docker-compose -f docker-compose-kafka.yml -p local</p><p>Voila, we just run Kafka clusters in our local machine. We can continue to configure the clusters, topics, partitions, etc via Kafka managers. It can be accessed via local web browser at localhost:9000. This port is reflected in the ports: 9000:9000 field.</p><p>To run DB apps, the idea is similar to Kafka. We just need to write configuration for docker-compose and set up the initial DB after it runs successfully. For setting up custom apps to be ran in Docker, we wrote some Docker scripts to build our code into custom Docker images. Once all of the above steps are done, we can control the startup order of different components in our system via simple shell script.</p><h3>Afterword</h3><p>We’re happy to have shared our knowledge from our in-house development process. Hopefully, this post gives you a glance into how we improve and optimize our asynchronous processes. If you find this article helpful and would like to join our ever growing team, you can find out more at: <a href="https://www.99.co/team/jobs/">https://www.99.co/team/jobs/</a></p><blockquote>This article is written by my awesome colleague <a href="https://www.linkedin.com/in/hermawanrh/">Hermawan</a>. Please give him all the creds for the work behind the article.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2c4dcd2053af" width="1" height="1" alt=""><hr><p><a href="https://medium.com/99dotco/how-99-co-indonesia-implements-multi-portal-data-sync-with-kafka-2c4dcd2053af">How 99.co Indonesia implements Multi-Portal Data Sync with Kafka</a> was originally published in <a href="https://medium.com/99dotco">99.co</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A case study about designing calendar page for managers.]]></title>
            <link>https://sneakycocoon.medium.com/a-case-study-about-designing-calendar-page-for-managers-f3697fc2b432?source=rss-c5ef536be3ed------2</link>
            <guid isPermaLink="false">https://medium.com/p/f3697fc2b432</guid>
            <dc:creator><![CDATA[Quang Nguyen]]></dc:creator>
            <pubDate>Mon, 23 Dec 2013 16:25:29 GMT</pubDate>
            <atom:updated>2013-12-23T16:26:44.937Z</atom:updated>
            <content:encoded><![CDATA[<h4>On how I made mistakes by sketching too early and how I fixed that.</h4><h4>The situation</h4><p>As a business student, I certainly do not act like one. Long gone the day I stick around school, do some HBR case studies and discuss group projects with my mates. Once I taste the triumphant feeling of building stuffs from scratch, especially after building the UI at the coolest online ticket platform in Vietnam, <a href="http://ticketbox.vn/">ticketbox.vn</a>, there’s no going back.</p><p>So when opportunity presented itself, I immediately jumped and took a shot at part-time front-end/interface designing for another Singapore startup called Beensprout. This time, the focus is about social networking for non-profit organisations. My first task was to put together a calendar page for non-profit admin.</p><p>A little bit background about Beensprout targeted users. There are two types of users: the organisers and the volunteers. The calendar page is mainly for organisers to manage different activities they have. Yet, even among organisers, there’re different levels of authority. We have org admins, people who overlook the whole organisation’s activity. We also have activity managers, who take care of specific activities within the organisation. At the lowest power level, we have event helpers whose tasks are to coordinate with activity managers to hold successful events.</p><h4>Initial stupid sketching excitement</h4><p>The calendar I was tasked to design targets at org admins. The general instruction from my boss was just to “help them manage events”. So with stupid enthusiasm, I popped open my Paper53 app and sketched some interesting ideas.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jvsfqFhcgSs8KriPbcxYaA.jpeg" /><figcaption>1st prototype</figcaption></figure><p>The idea behind prototype 1 was that org admins could color-code activities and filtered displayed activities using a tagging system. However, I quickly faced the problem of too many activities per day, which would rainbow-ize days with lots of activities, making it very difficult for users to tell what activities they have exactly.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_gZRe-x-mzQHuqGBEkkS7g.jpeg" /><figcaption>2nd prototype</figcaption></figure><p>For the second prototype, I wanted to explore a different option where org admins could track the exact progress of each activity. This was achieved by not using a standard calendar but instead limiting the timeframe to a weekly basis and giving each activity a row of its own.</p><p>Yet, after the sketching excitement was over, I realized my fatal mistakes: both options were made from my assumption about the users, without any input from real org admins.</p><h4>Fix it or it just becomes another Dribbble shot.</h4><p>To prevent the DRIBBBLISATION of design (Padday has an <a href="http://insideintercom.io/the-dribbblisation-of-design/">awesome post</a> about it), I took a step back and use what I learn from Ryan Singer of 37Signals when designing a web app. There were three high-level questions that I wanted to answer:</p><p>- What is the single task that the users are here for?<br>- How do users get to this page?<br>- What next after this step?</p><p>The answer to the first question seemed simple: org admins wanted to “manage” all activities and the calendar page should help them to do that. But there’s one thing I didn’t understand: What did org admins mean when they said they wanted to “manage”? What activities were they involved in? Did they decide what time an activity should happen? If yes, at what level of time detail could they interfere? Week? Day? Hour? Different time perspectives would certainly affect how flexible the calendar should be.</p><p>Seeing many potential questions similar to the one above, I decided to go back to the users for some chit chats. It turned out that my assumption about org admins wanting to look at detailed activities to help them schedule events was plain wrong. Scheduling activities was actually done by activity managers, which meant there would be a different calendar page for that.</p><p>To org admins, what they care about was not how detailed the activities were displayed, which rendered both my prototypes useless. Rather org admins wanted to have an overview of how “busy” the organisations were. The org admins believed that the more activities an organisation had, which meant more days were filled up with events, the healthier the growth of the organization.</p><p>So how do org admins navigate to the page i build? Back then, users could access the calendar page through a left hand side navigation via the “Activities” label, which my previous colleague had reserved when he built the page architect. But “Activities” didn’t seem so clear to users about its functionality.</p><p>Could the label change to “Activities overview”? And how else can org admins reached the calendar page? Could the calendar page be the landing page? Like a dashboard to provide an overview to users?</p><p>I didn’t really arrive at a good copy writing solution and had to leave the issue there for the moment. This was part of a bigger problem in navigation and I just didn’t want to digress myself. A separate problem should deserve its own time and researches before proposing any fix. Thus, I would have to push it to the parking zone and come back later.</p><p>Then, there was still one big question left to answer: what should the users do next? Based on previous interviews, after the initial overview, org admins needed to chat with activity managers to check their event progress. Sometimes they needed to add a new activity and assign certain organisers to handle that activity. The UI then needed to facilitate org admins to find these tasks easily.</p><p>Optimally, actions should be built-in to the context so that they provides a natural and smooth transition from task to task.</p><h4>Final objectives &amp; design</h4><p>With these inputs, I set out to decide my calendar page. Again, important ballpoints:</p><p>- Show healthy growth of the organisations through “filled/busy” days.<br>- Smart linking to next actions, including creating new activities, assigning organisers to activities, communicating to certain activity managers.</p><p>Now that I have real user inputs, it becomes easier to create and justify which wireframes should be the final.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ja3L3CU2UGYanXToWLO-Xw.jpeg" /><figcaption>3rd Prototype</figcaption></figure><p>This prototype only shows days with event at a glance, the users can still checkout detailed activity with the search bar.</p><p>Another 2 iterations and I finally arrived at <a href="http://murmuring-journey-2056.herokuapp.com/">a workable prototype</a>, which is currently implemented at our Beensprout Beta page.</p><figure><img alt="" src="https://cdn-images-1.medium.com/proxy/1*ZqJVXxgFCuKNnPzBRjYJkw.jpeg" /><figcaption>Final design</figcaption></figure><p>Each day now owns a progress bar to show how busy that day is.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*Xtd4G7vSfvQEkEH7iXGcLQ.jpeg" /><figcaption>Filtering activities</figcaption></figure><p>Org admin can still track detailed activities for actions like contacting the current activity manager, checking out activity time for a surprise crash.</p><p>Ps: the calendar styling is done with <a href="http://kylestetz.github.io/CLNDR/%20clndr.js">clndr.js</a>, a js plugin that allow you to flexibly define the calendar markup. I highly recommend this plugin for customized calendar design.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f3697fc2b432" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>