ChatGPT Plugin Follow-up
How my little experiment with ChatGPT Plugins has gone so far.
I wanted to share a follow up to my post on ChatGPT Plugins with some additional learnings, now that I have a plugin that’s available in ChatGPT.
As I mentioned in that original post, plugins to me seem like one of those rare things that happens every so often where a platform is in its early days and you can tap into that as a great distribution channel. At the time, it was mostly speculative since the plugin store was brand new and I had no actual first hand knowledge of what the market would be like.
Since then, however, I took a product I have been building over the past two years called Decision Journal and added a ChatGPT plugin to it to test out my hypothesis. That plugin went live on ChatGPT this past weekend and sure enough, the results have been pretty incredible.
What’s hard to tell from that chart is that even though it looks like there were no users for about every day until two days ago, that’s not actually true. It’s just that the last day of users has been so high that it makes the rest of the chart look like it’s been 0 the rest of the time.
To provide context around Decision Journal before the ChatGPT Plugin release, it has always been a pretty niche little product. While I do believe everyone in the world would benefit from using a decision journal (whether it’s paper-based or digital), I understand that not everyone’s going to take the time and do the work to become better decision makers.
Since Decision Journal’s initial launch, outside of an initial influx of people checking it out from Product Hunt, it’s been a slow trickle of users over the last couple years. I’d say roughly about 1-10 new users trying it out per day on average. Not terrible over a couple years, but nothing to write home about, that’s for sure. There were definitely a number of improvements and features that I could’ve made to help increase that but it became a side project of mine and honestly was just a product that I built for myself to solve my own needs, which I was fine with, so I never implemented any of those things.
By releasing the plugin, I was able to double the total number of users all-time in a day. Pretty wild.
Please note that this is only a few days worth of data so who knows what will happen after that. It’s most likely that this spike of new users will die off over the long run once the plugin has been around for awhile and once the plugin store gets many more offerings. Still, I wanted to share with you some of the early results since they confirm my suspicions that the plugin store is a great opportunity for distribution of applications right now.
There are still a ton of improvements that can be made to the plugin experience too. The obvious improvements on the ChatGPT end user side of things is just how barebones the plugin store is. This makes complete sense of course since I’m sure it’s low down on the list of tasks for their team to take on, but until they add search and categories to the plugin store, it can be a bit hard to find things.
Another improvement that could be made for both end users and plugin developers is having a bit better of an onboarding experience. Right now, it’s a bit confusing for new users trying out a plugin for the first time because ChatGPT doesn’t give you much information after installation (I’ve submitted this feedback to OpenAI).
Finally, it’s still the case that GPT kinda gets things mixed up when working with plugin APIs and I haven’t been able to figure this out 100% yet.
The last thing to note is that this is only talking about the initial acquisition of users. Obviously, there’s a lot more that goes into building a successful product/business that goes beyond acquisition (usage over time, churn rates, conversion rates, etc). But so often, we tend to fall into the trap of “if we build it, they will come” mindsets that I think sharing these sorts of things helps put things in perspective. Regardless, I thought the early results were worth sharing with others also because I haven’t read a ton of information yet from the developer side of plugins.
Give the Decision Journal plugin a try and let me know what you think!
I’ve seen various things. Some I’ve highlighted in my previous post but more recently I’ve seen it hallucinate ids to use with the APIs, which is not great since you could potentially end up in some bad scenarios if you update the wrong decision or attach a review to the wrong decision. There doesn’t seem to be a method to the madness that I can figure out yet. I assume it will get better over time but in the meantime, playing around with error messages to see if I can coach it back.