APIcastor’s Unanticipated Launch
Notes from being discovered by ProductHunt
October 25th. Coffee mug is empty. Laptop battery is following suit. The APIcastor app state is rough but functional. Decided to gather early feedback by showing Hacker News. A critical bunch, they’re usually quite good at poking holes and finding flaws in anything from UX issues to business plans. Initial reception was mixed, nothing I would describe as negative. A few pointed out services in the same space and others expressed interest if we considered adding a specific feature or two.
Late at night, while checking on traffic, I noticed a few requests coming from the ProductHunt upcoming page. Apparently, one of the PH mods had seen our HN posting and submitted us. A sense of dread washed over me as I anticipated the flood of traffic that might follow. I quickly made a short list of easy polish items that I could still tackle that night. Cracked out 80% of them, accepted our fate, and called it a night.
Mental note 1: Consider limiting signups to a set # and having a waitlist feature in place prior to launch to ensure traffic is predictable and potential signups aren’t lost.
Woke up early to see the damage. We received a few upvotes but had not moved to the front page. Checking again around lunch — same results. Hardly any traffic, my fears of not being ready or polished enough subsided and changed to disappointment at the thought that we might not get another shot at the PH frontpage for awhile.
Running late. Full day of meetings ahead. No time to check site traffic. At some point in the morning, the flood gates opened. Notifications from across all channels started pouring in, AWS, Twitter. It made the frontpage! The race was on, but now the initial fears returned. “Is the product ready for this kind of traffic? This is still closer to a proof of concept than a full-fledged product.” Server health is holding steady. A few error reports start trickling in — no time to dig into them now.
Mental note 2: Error reporting is next to useless in a minimal state, find time to flesh them out.
Upon finding a few minutes, realized the issue was with our database — our connections were fine, the instances were up, but they had run out of storage. I hadn’t proactively allotted enough space. Wonder how many people we lost to that mixup.
PH traffic is still strong but waning. I was able to spend a little time making the site more robust and reflect on the future.
Our immediate concerns are three-fold. One, improve customer communication channels. Proposed solutions include posting to this blog, incorporating transactional emails into the product, adding a “what’s new” section to the site, and potentially a public Trello board for long term roadmap viewing. Two, redo the single API page design. Analyzing user traffic patterns highlighted a few areas on the page that weren’t being conveyed properly. Three, general site robustness, improve testing site-wide and more gracefully handle errors when they do occur.
Beyond those concerns, focus will be spent on two areas. Re-architecting how our APIs are stored to better handle large files (> 50MB) while not sacrificing any speed. And completing integration with Stripe to allow our users to add a Paywall in front of their APIs.
We’ll detail more of the who and why in posts to come.
— Team APIcastor
PS — Mental note 3: Write Product Hunt and suggest giving makers more of a heads up than the morning of. In fact, they’re pretty good at spotting social stuff maybe I’ll just collect these notes in a blog post… =)
TL;DR: PH launch was unanticipated. Lots of good results from it, some bad. Immediate concerns include improving customer communication channels, redesigning the single API page, and improving site-wide robustness. Near term goals include supporting large files (>50MB) and adding configurable paywalls in front of user APIs.