<?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 Nixon Augustin on Medium]]></title>
        <description><![CDATA[Stories by Nixon Augustin on Medium]]></description>
        <link>https://medium.com/@nixon_augustin?source=rss-921ad29ea9cd------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*lswIhOgiLvv-QBY9RO_TQw.jpeg</url>
            <title>Stories by Nixon Augustin on Medium</title>
            <link>https://medium.com/@nixon_augustin?source=rss-921ad29ea9cd------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 14:31:05 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nixon_augustin/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[I tried to renew car insurance online. It crashed. So I automated the whole thing in 4 hours.]]></title>
            <link>https://medium.com/@nixon_augustin/i-tried-to-renew-car-insurance-online-it-crashed-so-i-automated-the-whole-thing-in-4-hours-393ce031985b?source=rss-921ad29ea9cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/393ce031985b</guid>
            <category><![CDATA[selenium-test-automation]]></category>
            <category><![CDATA[quality-assurance]]></category>
            <category><![CDATA[test-automation]]></category>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[Nixon Augustin]]></dc:creator>
            <pubDate>Wed, 21 Oct 2026 19:48:13 GMT</pubDate>
            <atom:updated>2026-10-21T19:48:13.507Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g_gvNUSAqlSlk7quuZKUYA.png" /></figure><p>Earlier this year I sat down to do something boring: renew my car insurance. I went to HDFC Ergo, filled in my details, clicked through the buy flow — and the website crashed. Not a validation message. A full dead end:</p><blockquote><strong><em>Error!</em></strong><em> An error has occurred. (Error Code: f51438ad-…)</em></blockquote><p>A RETRY button that didn’t help. The URL said SessionExpired. I was a customer with money in hand, and the website wouldn&#39;t take it. I figured it was a one-off glitch, closed the tab, and got on with my day.</p><p>A few days later I went back to finish the purchase — and hit the <strong>exact same wall, at the exact same step.</strong> Once is a glitch. Twice, identically, is a pattern. That’s when the test engineer in me took over: this isn’t a fluke — <em>can I reproduce it on demand?</em></p><p>So, a few hours after that second failure, I got to work. In about <strong>four hours, with GitHub Copilot riding shotgun</strong>, I built a full automation of HDFC Ergo’s car-insurance quote-to-buy flow.</p><p>Four hours for a full quote-to-buy automation — OTP and all — sounds fast, and it is. It’s fast for two reasons, and I’ll be upfront about both. First, I didn’t build it from scratch: I built it on <strong>Flex</strong>, a cross-surface test framework I’ve been developing. Flex handles the boilerplate — drivers, waits, the web/mobile/API/hardware plumbing — so I wasn’t wiring up a harness, I was describing the flow. Second, <strong>GitHub Copilot was pair-writing with me the whole way</strong> — scaffolding the page objects, suggesting selectors, filling in the repetitive code while I steered the hard parts. The framework removes the boilerplate; Copilot speeds up what’s left; and I still had to make the calls — how to structure it, and the fiddly bits like the OTP bridge that needed a bit of figuring out.</p><h3>What it does — end to end, no human touching the keyboard</h3><ul><li>Opens the site, navigates to Car Insurance</li><li>Fills the quick-quote form — name, mobile, city, vehicle number</li><li>Clicks <strong>Send OTP</strong>…</li></ul><p>…and this is the part I’m quietly proud of:</p><ul><li>It connects to my Android phone over WiFi (ADB), waits for the HDFC SMS, <strong>reads the one-time password straight off the device, and types it into the form.</strong> Fully unattended.</li></ul><p>OTP is where most automation quietly dies. The code lands on a phone, a human reads it, the robot stops. Bridging that gap — phone → script → browser — is the difference between a demo and something that actually works.</p><pre>SmsOtp.wait_for_otp: sender=&#39;hdfcgi&#39;  timeout=120s<br>AdbWifi.connect: &lt;phone-ip&gt;:&lt;port&gt;<br>SmsOtp: OTP received: ******<br>QuoteForm.fill: entering OTP<br>QuoteForm.fill: waiting for OTP verification tick  ✓</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rvi_J30yrIT1JzBAZEITEw.png" /></figure><p>From there it walks the rest of the journey: verifies the vehicle details the site pulls back, picks the policy type, answers the previous-policy and no-claim-bonus questions, pulls the quote, and reaches <strong>BUY NOW</strong>. Every step screenshotted, every network call captured, a clean log from start to finish.</p><p>Here’s the whole run at a glance (times are from the start of the run):</p><pre>00:00  ▸ open hdfcergo.com                        ✓ home loaded<br>00:03  ▸ BUY ▸ Car Insurance                      ✓ quick-quote form<br>00:07  ▸ fill: name, mobile, city, vehicle no<br>00:07  ▸ Send OTP  →  connect phone via ADB/WiFi<br>00:16  ✓ OTP read from phone &amp; entered [******]   ← 9s, fully unattended<br>01:06  ✓ vehicle auto-detected: Maruti Vitara Brezza ZXI+ (2021, MH-12)<br>01:39  ▸ policy type: Comprehensive<br>02:16  ▸ previous policy: Comprehensive / Not Expired<br>02:21  ▸ claim: No · NCB: 20%<br>02:22  ▸ VIEW QUOTE<br>02:31  ✓ BUY NOW reached<br>──────────────────────────────────────────────────────────────<br>154 seconds · 0 human keystrokes · 543 network calls + 7 screenshots captured</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HnsO-uYOwzbj3YRpGrF4uA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XeHF76jwQJRELlo6Hw6qQw.png" /></figure><h3>Did the automation reproduce it? No — and that’s the interesting part.</h3><p>Here’s the twist: on both automated runs, the flow completed cleanly. No crash. Yet by hand, just hours earlier, I’d hit the wall every time — at the <strong>NCB step, the moment after I answered “No” to the claim question.</strong> Same path, same inputs, opposite outcome. I genuinely expected the bot to crash where I did; it sailed straight through.</p><p>At the time, I had no idea why. My first instinct was the usual toolkit — add a few delays, add a retry, poke at it until it broke. It was only later, looking again at that SessionExpired sitting in the URL, that a theory clicked into place: maybe it was never about the steps at all — it was about <strong>pacing.</strong></p><p>When a human reaches that step, we pause: reading the options, deciding, typing slowly. Those pauses give a session token (or a backend integration step) time to <strong>expire</strong>, and the next click lands on a dead page. The automation doesn’t pause — it lands the right strokes at the right moments and clears that step almost instantly, so it never idles in the window where the session dies. A slow human hits it; a fast bot sails right past. It’s a guess, not a verdict — but the error page naming itself SessionExpired is a strong hint.</p><p>And proving that guess would mean doing exactly the wrong thing: deliberately slowing the script down, step by step, to widen the window and force the crash on demand — against a live production system I don’t own, firing a real OTP to a real phone on every run.</p><p>So I stopped, on purpose. A couple of those early runs were just me getting the script working, and each one already sent real traffic to their servers and a real OTP to my phone. Once it ran clean end to end, I had everything I actually needed: a log that proves the automation works, and enough detail to report the bug. Chasing the crash further — hammering a system that isn’t mine just to force an error — had no good justification. Knowing when <em>not</em> to run the test is part of the job.</p><h3>The funny part</h3><p><strong>While the automation was running — mid-flow — my phone rang.</strong> HDFC’s sales team, calling to help me buy a policy. There I was: a script quietly filling their quote form on one screen, a salesperson offering me the same policy on the phone. I could have just bought it through him right then — but my head was deep in reproduction mode, so I explained what I was actually up to: an engineer trying to pin down a crash in their buy flow. Their broken checkout couldn’t close the sale, but their lead funnel found me in real time anyway.</p><p>I did report it — with the reproducible step, the run log, and an offer to point my suite at their test environment for free so they’d get a definitive root cause. The reply came back: a <em>temporary network glitch.</em> Maybe it was. But a network blip doesn’t usually drop you on a page literally named SessionExpired, at the same step, right after the same click. I&#39;ll let the error message speak for itself.</p><p>That’s okay, though — I didn’t build this for them. I built it because this is the kind of problem I enjoy, and because four hours of work turned a frustrating dead end into something I could actually show.</p><h3>So here’s my honest question</h3><p>If I built something like this for <strong>your</strong> product — your checkout, your OTP flow, the one critical path that absolutely cannot break — would it be worth anything to you?</p><p>Would you just scroll past… or would it be worth a thank-you, a coffee, a conversation?</p><p>I genuinely want to know. I can produce logs and automation like this, and I can do it fast. If that’s useful to you, or to someone you know, I’d love to hear from you.</p><p>— Nixon</p><p><em>Built with Python, Selenium, a bit of ADB, and Flex — a cross-surface test framework I’ve been building that handles the boilerplate. ~4 hours, Copilot-assisted.</em></p><p><em>Want the receipts? The full run log — every step, every wait, every network call, nothing hidden (personal details masked) — is here: </em><a href="https://gist.github.com/naugustin/507c85753dd9c16c94f6ee666100ee51">https://gist.github.com/naugustin/507c85753dd9c16c94f6ee666100ee51</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=393ce031985b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What causes test automation chaos in most companies?]]></title>
            <link>https://medium.com/@nixon_augustin/what-causes-test-automation-chaos-in-most-companies-f9f05a0a4b87?source=rss-921ad29ea9cd------2</link>
            <guid isPermaLink="false">https://medium.com/p/f9f05a0a4b87</guid>
            <category><![CDATA[testing]]></category>
            <category><![CDATA[test-automation]]></category>
            <category><![CDATA[pytest]]></category>
            <category><![CDATA[ci-cd-pipeline]]></category>
            <category><![CDATA[python]]></category>
            <dc:creator><![CDATA[Nixon Augustin]]></dc:creator>
            <pubDate>Thu, 01 Jun 2023 17:34:04 GMT</pubDate>
            <atom:updated>2023-06-01T17:35:50.330Z</atom:updated>
            <content:encoded><![CDATA[<p>In my opinion, test automation is an area in the IT industry that has seen limited innovation. While this statement may sound unusual, it is based on my 20 years of experience in the field. Whether it is a well-established company with a long history or a newly formed startup, both face challenges in finding a reliable test automation framework. Although the dynamics differ between established companies and startups, the need for a robust automation framework is crucial for both.</p><p>It is widely acknowledged that a good automation framework can significantly improve productivity, reduce costs, and enhance product quality. Regrettably, I have witnessed companies that have suffered due to the lack of a solid test automation framework, resulting in inadequate testing and delays in product development.</p><h4><strong>Established company</strong></h4><p>Let’s begin by examining the issue within established companies. Typically, these companies tend to have multiple frameworks, and in some cases, each team within the company may even have its own framework. The development of new frameworks often occurs as the company diversifies its product offerings or through acquisitions. Unfortunately, due to the lack of a viable framework or unifying mechanism, attempting to consolidate these frameworks becomes practically impossible. As a result, the proliferation of frameworks persists, leading to increased operating costs for the companies involved.</p><p>The idea of “consolidating” or “unifying” test automation frameworks within established companies is often discussed. It involves the effort to merge multiple test frameworks into a single framework. Typically, at the request of management, different teams gather to present their frameworks to others. However, this endeavor is likely to face resistance, as teams tend to cling to their existing frameworks and resist any attempt to impose a new framework upon them. Having ownership of a framework is seen as a way to enhance job security, making it understandable why teams are reluctant to let go. In some cases, management might enforce the adoption of a specific framework on all teams, but this can lead to challenges in the long run. The teams on the receiving end may attribute any issues in their testing process to the “new framework,” resulting in frustration and blame being placed on the framework itself.</p><p>I believe there are two primary reasons why consolidation efforts often fail and why test automation can be a messy endeavor in many companies:</p><ol><li><strong>The absence of a comprehensive or near-comprehensive framework within the company:</strong> In most cases, the existing frameworks within the company have their own limitations and shortcomings. It becomes challenging to select a single framework that addresses all the requirements and resolves the existing flaws effectively.</li><li><strong>The lack of qualified individuals with an independent perspective:</strong> Without the involvement of knowledgeable and unbiased individuals, it becomes difficult to make informed decisions about selecting an existing framework or developing a new one. The absence of qualified experts who can provide guidance and contribute to the framework selection process hinders progress and leads to confusion.</li></ol><p>To summarize, the absence of a suitable Uber framework and the lack of qualified individuals to guide the selection process contribute to the challenges and messiness experienced in test automation consolidation efforts within companies.</p><h4>Startup</h4><p>The startup environment offers a relatively simpler and more straightforward approach. As startups build everything from scratch, test automation is also developed from the ground up, free from any existing baggage. It provides the opportunity to adopt the latest technologies available.</p><p>However, challenges arise due to the complexity of the product being tested and the expertise of the hired automation engineer, which significantly impacts the outcome. Finding a qualified engineer can be difficult, leading to the development of sub-optimal frameworks that are reluctantly accepted. In some cases, an engineer may struggle to produce desired results within a specified timeframe, resulting in their departure and the need to restart the process. Outsourcing test automation to third parties can further exacerbate the problem, as lofty promises often lead to less satisfactory outcomes. When faced with difficulties, third-party engineers may simply quit, leaving the company in a precarious situation.</p><p>The issue of engineers quitting their jobs is not exclusive to startups; it can also occur in established companies, although it happens less frequently. In established companies, the situation is slightly different as they have more resources and the ability to tolerate a substandard framework due to limited visibility beyond the team manager. Consequently, the manager becomes the de facto owner of the framework and may choose to rely on a mix of automation and manual testing or hire additional resources to compensate. However, this approach comes at a significant cost to the company.</p><h4>What is the next step forward?</h4><p>The purpose of this article is not to cause panic among those seeking to hire test automation engineers, but rather to sincerely address the pitfalls of test automation and assist companies in making better choices. While there are many dedicated test automation engineers out there who are contributing to the success of companies, not every company is fortunate enough to have a skilled automation engineer. Furthermore, there are numerous committed engineers who are willing to work tirelessly, either through manual testing or adhoc scripts, to compensate for any shortcomings in automation.</p><p>The initial question that may arise is how an external company can assist in testing a wide range of diverse products. The second question may revolve around trust, particularly for those who have had previous experiences with exorbitant charges and subpar services. To address these concerns, I look to the successful business model of Mercury Interactive (acquired by HP), particularly their product WinRunner, a GUI automation tool. Their approach involved providing a base product, accompanied by sample scripts and documentation, which allowed users to delve into the product. Additionally, they had a team of support engineers who were readily available to answer questions and provide solutions for complex scenarios. I believe this model is key to resolving automation challenges, where a customizable base product can be tailored to meet the specific requirements of each customer, providing them with the necessary support to succeed.</p><p>As I work on my startup to offer test automation services, I am more than willing to help if you are facing any test automation challenges. Feel free to reach out to me via email, as I am just one message away.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f9f05a0a4b87" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>