A weekend project, months later
April 21, 2023
In a classic case of "this should only take a weekend to build", months later, I finally feel good about where my latest project Tabbot stands. I should emphasize "latest" side project because I am kind of an expert at building something and then quickly forgetting about it.
But, for the first time, I actually find myself using Tabbot a lot in my own life. It's become a useful tool.
The concept is a SMS based "bot" that is built with the goal of helping you split a big restaurant bill with your friends.
I always want to pick up the bill to score those sweet sweet credit card points, but find it awkward to have to message everyone and ask what they ordered.
I don't want to play debt collector with my friends.
AI is good, why can't we automate this? Take a picture, use OCR (object character recognition) to itemize the receipt, share the generated link, then let people claim the items they ordered on their own.
Native apps exist that can do this sort of thing, but I've learned that trying to get a group of friends (and friends of friends) to download a new app is a losing battle.
My theory: since you would be text a picture your receipt to your group anyway, why not just text it to Tabbot instead?
Tech behind Tabbot
My main hypothesis with Tabbot is that nobody will want to download another app to split a receipt. I am taking a web based approach.
- Twilio - MMS and SMS
- AWS S3 - Image storage
- Serverless - Image processing
- AWS Textract - OCR
- Supabase - Backend as a service
- NextJS - Frontend Framework
- DaisyUI - Styling
I've learned some lessons
I honestly thought this would just take a weekend to build. I had this thought in late 2022. But, as a surprise to nobody, months later I'm still tinkering with it.
Tabbot is not my main focus, so I can only allocate a few hours during the week and some time on the weekends to building. I think I'm finally nearing a useable MVP.
Choose your name carefully
I'm not sure that I nailed it with "Tabbot", however, I originally launched as "Tabbit" and quickly realized my SEO mistake. Try ranking on Google with a brand name that is one letter away from a popular pet animal.
Choose your domain carefully
I've cycled through a number of domains by now. My first domain was Tabbit.xyz. I thought this was innocent enough until I logged in to Google Search Console and saw that my project was flagged as "pure spam".
I had no idea why.
After digging through Google's webmaster documentation for days with no luck, I decided to check Wayback Machine and see if someone had been using Tabbit.xyz prior to me. They were. It was spam. I unintentionally bought a crappy domain that would've tanked my SEO chances for a bit.
Luckily I'm really good at buying domains.
Sometimes it's better to buy than to build
A few years ago I would've blindly jumped into this project with the intention of building everything from scratch.
I would've self hosted a backend and database. I would've spun my own OCR. I would've tried as much as possible on my own.
I'm slowly learning to ignore this impulse. When you're building a house, sometimes it's better to buy a hammer than try to build your own.
Search ads are effective but expensive
I ran some experiments with search ads. I was targeting "receipt splitter app" and similar search terms. It worked out very well. I got a bunch of high intent traffic - these users were searching for my exact solution.
But, I ended up spending way too much money for no return. I have no monetization strategy yet so I was just lighting money on fire.
Hopefully I can organically rank for those search terms instead of purchasing ad spots. However, I'm glad I ran this experiment, it validated that there is a market for this tool.
I'm not totally sure. I'm going to focus on organic SEO, ranking for "receipt splitting apps", etc. Continue to push more features. Refine the user experience. Write about it more. Figure out monetization.
Give Tabbot a try, I would love your feedback!