Sell it before you make it right?
I'm not convinced it's the right approach but it sure is better than building them all only to find out it's not what people want. I'm having doubts about this as a marketplace. Machine learning is fickle, it's changing constantly, and very difficult to keep up with. I think until we hit a critical mass of about 100 (making that number up), it's going to be hard for people to find what they're looking for.
So I put about 15 or so models that I think would be interesting road map items and allow users to get alerts when they're ready for prime time. I'll only build the one's people have already upvoted on and hopefully can find product market fit in this way.
What happens though if few or no people vote? We'll it may be back to the drawing board, and that's alright too.
I just moved our emails off mailchimp to substack. It took about 15 minutes and I sent my first email asking users about our pivot.
Substack is free and easy to use.
Mailchimp is cumbersome and expensive.
Substack (when used for company emails) is a classic minimal product in an overserved market. Wish I thought of it...
2 years ago I started building SugarKubes as a machine learning marketplace. I quickly pivoted to a code marketplace (generic code marketplace) and had a hell of a time getting developers who wanted to sell code onto the platform.
Now, after about a year of realizing that's not going to work (at least not yet) I'm pivoting back to a machine learning marketplace. The value prop and idea will be straightforward... We sell machine learning models packaged as docker containers.
A few insights:
it's a marketplace but we're building all the ml models (nobody else will be selling on the platform for the foreseeable future)
double sided marketplaces are tough, let alone for single founders
I raised prices. There's nothing on the site for less than $100.
I'm playing with a non-recurring business model. There are many ways I could package this product up but I'll start with the simplest one which is people pick and pay for the model they want with a one time fee.
Yesterday I took away all signup buttons on our landing page.
Being "launched" wasn't doing anybody any good. As a marketplace, in the beginning, most interactions are disappointing. Producers don't get enough sales, and Consumers can't find what they're looking for.
So we've unlaunched to gather more data. The home page now encourages 1 action: Searching.
We're just going to log what people search for in the field for the time being so that we know what people would be interested in buying if it existed on the site.
I spent a little time this week updating our landing page to address some of the complaints we've had.
It's always amazing to me how hard it is to get an idea across on a landing page.
Our message is simple. Buy & Sell Code. The way we do it is slightly different than other players in the space but that's just an implementation detail.
Anyways, hope this version converts a little better!
I added a new product to SugarKubes. Our latest microservice is a simple microservice that does passwordless authentication.
It's simple to set up and supports MySQL, PostresQL, MSSQL, MariaDB and SqLite.
Take it for a test drive and let me know what you think!
More importantly, I'd love to hear what you think I should add next.
Growth for the marketplace has been slow but steady. I think a lot of it has to do with how unnatural it is for developers to buy anything other than templates or starter projects.
So that is going to have to be the focus moving forward. Perhaps we can change behavior later by having really specialized and awesome React components, etc, but until then I don't think there will be any other way to grow than to lean into templates and starter projects despite that not being what I had in mind.
That being said, we now have a total of 10 others selling on the platform so I don't have to build everything for sale by myself!
103% sounds like a lot but definately didn't feel like it. I use Mongodb Atlas for our database and they have some simple analytis tools. I didn't realize the growth rate until I changed the chart from showing daily to showing monthly.
Unfortunately, our revenue didn't change at all this month but we were able to add 2 more sellers and add 2 more projects to the platform.
Hopefully, there is some momentum now for people to discover SugarKubes on their own through blog posts, or word of mouth, but I doubt 134 people is a critical mass yet.
I'm still of the opinion that this business does not need money, but time, although at the moment a marketing budget does sound nice.
Today I lauched a funnel experiment. I realized that the sites that let users and teams share code are actually in a much better position to offer what SugarKubes does than I am.
So I built a very simple tool to organize and save multi-file modules. Somewhere in between a gist (one file) and a repo (kind of annoying when you have just a few files of code).
it's called Code Clip and I just put it up today at https://codeclip.dev.
I'm especially interested to see if this can be used as a top line funnel to sugarkubes and if there's any traction, I'd like to add a "sell this component" button in the app to more easily allow people to upload cool projects to SugarKubes.
We've reached 100 users and 7 sellers!
Truthfully for the amount of work it took to get here I'm bearish on the ability to reach a critical velocity. I'm sticking with it because I don't see why this shouldn't be an option for many people who are in between an open source project and a full on company. There are millions of people and many millions of tiny projects that would be amazingly valuable if they were packed up on their own and sold on SugarKubes.
This final push to 100 was achieved by posting a freelance job on upwork to build products for SugarKubes. I was hoping to attract some more users and it seemed to work!
Next up is posting on 100 github repos to see if I can get some more visibility that way.
The question is, who will be the 101st user?!
...because ml is hard. We sell well documented, easy to use machine learning models packaged as docker containers.