My product is priced at $3.49 /year after 5 free uses, but I'm just wondering if it could be a better idea to just ask for a coffee or something and leave the thing free. What do you guys think?
It's a bar/restaurant splitting app. I tested every bill-splitting app I could find and kept going back to a spreadsheet. Bill-e works best for that single purpose.
And how can I post the url without being banned xd?
For my first digital product, I wanted to start with something easy, but I didn't realize beforehand how much work it would take to create something I was proud of.
The problem: splitting a bar/restaurant bill among a group of people. I know it's a boring problem to solve, and there are a ton of options out there already (splitwise?). But you know what? I tested every bill-splitting app I could find and kept going back to a spreadsheet. Most of them want you to sign up, type every item by hand, and decide too many things at once.
It's a very simple task, so why isn't there a simple product to do it?
So I built Bill-e, a mobile-first web app, and I did it with the intention of making money from it as a challenge. With that in mind, the challenge was not just building it, but figuring out how to create the best possible solution for that single problem so it's worth paying for. Two key points:
Low cognitive effort and simplicity.
Cognitive effort. From watching a friend interact with the prototype (she literally said after scanning, "What do I do now?"), I realized I had to design a pipeline where only one decision has to be made at a time, and what has to be done has to be obvious. Bill-e is designed in 3 simple steps for that reason: (i) scan and verify, (ii) assign items, and (iii) share results. Also, the UI was carefully thought out for that purpose, and in case the LLM gets the scan wrong, you'll be prompted on how to edit manually.
Simplicity. My goal was to get the job done with the least number of clicks possible. To that end, the scanning and digitalization of the bill is essential. Turns out that solving that single problem is complex because bill formats are infinite, and photo quality varies. I evaluated going with a trained ML solution and/or LLM, and I picked the latter for cost reasons. Then I ran a benchmark against 300 bills, testing 3 different models, and I tried a few prompt formats with many variations each. The result: 94% of scanned bills will be digitized correctly, and you can go directly to assign items after only 2-3 clicks. If the scan fails, it will be due to small differences with the bill, and editing is super easy. Then, item assignment is also pretty straightforward, and it takes 1-3 clicks. Also, sessions can be shared so that other people can assist you with the task.
Paywall experiment: For the sneaky ones, or in case $3.49 for a year of unlimited use is out of your budget, when you hit the paywall after 5 free splits, you will be able to split the cost of Bill-e among the people you went out with. You pay the full price of Bill-e, but then the item "Bill-e" is added to the bill split results and your total is recalculated, so you only pay your share. You end up paying cents for Bill-e.
Bonus: All your bills are saved, the share results option is always available, and a checkbox is included to track who paid.
Does the paywall feel fair or scammy? Find any bugs?