Well, I guess this is becoming a series! See part 1 & part 2 here.
For over a year, I've been chronicling my experiences launching a tech startup, without writing any code. As Kollecto grows, I get more sophisticated in how I build. Thus far, my posts have unintentionally been helpful to lots of non-technical founders, so I wanted to share another update on how exactly I'm making this work.
In May, Kollecto raised $100k from 500 Startups & was invited to join their accelerator program in California. We entered the program having written zero code. At the time, we had an invisible app that was run by stringing together a handful of existing (logic-based) software tools.
The first part of this post is about how I built that app. By the time we left 500, we had our first real web app, plus a cool art matching algorithm—all built without any code. The second part of this post is about the current app & how we're building today!
Part experiment & part necessity.
The experiment: In the past year I've had a good number of people share that my process has helped them launch their own startup ideas. That really fuels me to see how far I can take this! But on top of it all, the process has also been super empowering for me.
To let users sign up & create a "taste profile."
To send data from Typeform to Google Spreadsheets.
Using Zapier, I sent each user's data into a spreadsheet.
The raw data speadsheet was referenced by a different taste profiles spreadsheet which put each user into a taste bucket using an "if statement."
For example, if the user had a budget of $100-$300 and was mainly collecting, and was looking for known artists & wanted street art, then the spreadsheet would put that user into a taste bucket called, Street B.
Each week, Art Advisors found art for each taste category. Then, over the weekend, we would have a freelancer on Upwork upload all the art into SurveyGizmo.
Most surveys have show/hide logic. If you notice, all the tools I use to build without code have heavy logic capabilities: that's the key to making this work. Each artwork was assigned to be shown to a certain taste group by the Upwork virtual assistant.
When we emailed users with a link to their art, we put the user's info in the URL. We used Campaign Monitor (a Mailchimp alternative) to dynamically create urls for each person. Like this: art.kollecto.com?profile=StreetBB&name=Taraemail@example.com
SurveyGizmo could read the info in the URL & show the right artwork to the right user.
In June, everything I knew about building without code changed. I know that sounds super dramatic, but it's true!
To be honest, we hit an unbearable place with Kollecto's invisible app. Stringing together a bunch of tools worked well until we hit ~1500 monthly active users and then it fell apart. The main issue was that we were still partially relying on human labor.
This was a good and bad thing. Good because human art advisors could read between the lines to understand the tastes of our users. Bad because the human process was much slower & also, each human advisor was limited by his/her own tastes & biases.
But the main problem was speed. There was a period of time where it took us weeks to get art recommendations out to our new users. So it was clearly time to update our process & technology.
By July I stumbled upon AirDev, an agency that builds apps for people using a new platform, called Bubble—which lets you program without writing code.
PS: If you need an MVP built really fast, talk to these AirDev guys! They're good!!!!
I had heard of Bubble before and had been meaning to use it. AirDev was hosting a free 2-weekend workshop on how to use Bubble.The course was intended for college students at Berkely & Stanford, but they let me sit in on the class.
After the workshop I hired the AirDev team to consult me when I got stuck learning/using Bubble. This isn't their normal business model—they usually build the app for you! But built it myself & did a few consultation sessions with them to learn some of the more complicated tricks in Bubble.
Chose Art Type:
Set a budget:
Answer some questions about your taste, based on the genres you selected:
Get art recommendations, rate them & we'll get smarter about your taste over time.
If you can understand an "if this, then that" statement, then you can program in Bubble.
Translation: When the art genre/ category button is clicked, then add that category to the user's list of Art Categories Selected.
Translation: When the art genre/ category button is clicked & the user has already selected that category, then remove that category to the user's list of Art Categories Selected. There's lots of logic like this in Kollecto's app, including a really complicated thread which is our matching algorithm.
Man! There's never been anything perfect about the apps I've built. For example, when we launched v4 in Bubble, the UX was terrible. This kinda stuff happened:
And this stuff happened too:
Lol, I'm pretty sure you're not supposed to share your negative support tickets, but I'm just trying to share the real deal. Now, we have a lot of the kinks worked out!
We're still working out issues with speed & we're still iterating on the matching algorithm. But we're getting good feedback!
People ask me all the time, for a list of entreprenuers who also build without code. Until this summer, I pretty much thought I was alone. But I've met some cool people through the Bubble community that are doing cool things!
For example, I'll leave you with this crazy project by the AirDev team who are doing some cool things while building without code! Check out notrealtwitter.com.