Week 9: Launch Day, Panic Fixes, and Zero Support Tickets

There is no feeling quite like launch day in a fast-paced startup. But before I could officially merge my code and ship this massive feature to SideShift’s users, I had to execute one last 24-hour engineering sprint. If you recall my conversation with the “super user” at Stanford last week, he pointed out that brands desperately needed exportable PDF receipts for their bookkeeping once a payout was processed. Instead of wasting time building a complex backend rendering engine from scratch, I pulled in a frontend library called “jspdf”. It allowed me to automatically format and generate clean, downloadable receipts directly from the campaign data already loaded in the user’s browser. It was a quick “fast-follow” feature, but it was the final puzzle piece the marketplace needed.

On Tuesday morning, I took a deep breath, merged the feature branch into our main repository, and watched the Vercel CI/CD pipeline push the code live to production. And, because this is the real world of software engineering, something immediately broke.

Almost right away, a power user with a massive roster of active campaigns logged in, and their dashboard simply refused to load. My stomach completely dropped. I dove into the Firebase error trackers and immediately realized my mistake: while testing locally, I had created a specific database rule to sort documents, but I had completely forgotten to deploy the required Firestore Composite Index to the live production environment. I scrambled into the Firebase console, built the index manually, and within five agonizing minutes, the data populated and the dashboard was working perfectly. That was a long five minutes.

With the system finally stable, I spent the rest of the week answering the core question driving my entire research thesis: did any of this actually solve the problem? I built out analytics dashboards to track our new user funnels, and the quantitative data absolutely blew my mind. Back in Week 2, our baseline metrics showed that 81% of creators were wasting over three hours a week updating Notion boards and tracking down rogue payments. Post-launch, the average creator session length on the new Earnings Dashboard is exactly 42 seconds. They log in, instantly see their yellow “Processing” or green “Paid” pill, and log out with total peace of mind.

The impact on the brand side was just as dramatic. Remember that A/B test I ran last week where I moved the “Approve” button directly into the Whop Chat window? That single UI tweak dropped the average time brands take to approve content from several days down to just 14 hours. But the absolute best metric of the week? Support tickets. Before this launch, SideShift leadership was fielding about 14 frantic messages a week from users asking if content was approved or if money was sent. In the four days since we deployed, payout-related support messages dropped to zero. The automated UI statuses and Whop webhook integrations completely cannibalized the need for manual support.

To cap it all off, I deployed a lightweight survey in the app asking users to rate the new system. We scored a 4.8 out of 5.0, and one creator left a comment that made all the late nights entirely worth it: “The new messaging system is incredible! Everything is in one place and the money hits instantly when they click approve. Finally. Props to the team”

The software is live, and the data proves it works. Next week, I just have to compile all these graphs into my final academic case study and prepare for my presentation at the Senior Project Symposium. See you at the finish line!