Premium Version & Survey Results

Jakob, April 26, 2024

TL;DR

• Why is there a premium version now? 👑

  To support the app long term. It was only free before because I worked without pay for more than 2 years, burned through all my savings, and friends and family helped out with some loans.

  Generating revenue will also help upgrade the servers, allowing for bigger neural networks to improve accuracy and performance. Thanks to premium, I can now finally enable you to take unlimited “pro” images that identify multiple items at once.

• How did the survey go? What’s planned for the future?

  I ran a survey through the app to learn what you think about how the app should be monetized and how it should be improved. 665 users responded! 🙏

  I’m very excited that most users understand why monetization is necessary and appreciate the app’s value. I’ll use your feedback to prioritize feature updates.

Premium Version & Survey Results

After more than 2 years of working on this app, I need to figure out how it can make money to ensure its future. This will also help upgrade the servers, allowing for bigger neural networks to improve accuracy and performance.

That’s why a few weeks ago, I ran a survey to learn what you think about how the app should be monetized. I also asked what can be improved and which features you’d like to see.

Here, I want to share what I’ve learned, outline future plans for the app, and discuss monetization.

Thanks for responding to the survey!

665 people responded to the survey. The effort you put in was almost overwhelming. So many good ideas, thoughts, and insightful answers! You wrote 37,000 words. (This is nearly half of “Harry Potter and the Sorcerer’s Stone,” which has 77,000 words.) 🤯

Thanks for participating in the survey!

I’ve read every single word and categorized the responses. This allows us to see which features and issues were important to most people so we can prioritize accordingly.

There were many interesting insights, too many to describe all of them here. I’ll try to keep this post at a reasonable length.

Feature Ideas & Improvement Ideas

I asked “What is the most frustrating thing in the app?”.

Here are the 3 most common answers:

• After taking an image, it is very tedious to take the next image because of “Are you sure you want to go back”.
  • I understand. There are too many unnecessary clicks for most users. I am asking “are you sure” because it is very annoying to accidentally click back and lose your progress after identifying 10 items on the image. I wanted to avoid that. This will be solved by letting you view the most recent scans, so that you can always go back if you want to.
• Database updates should be more frequent.
  • Before the beginning of 2024 the updates took way too long. Ideally, you’d like to log new sets immediately after they come out or after the next app update. The database updates have gotten more frequent in 2024, and soon you should never have to wait more than a couple of weeks.
• It is annoying to have a limit of only 3 Pro images per day.
  • Identifying items requires neural networks that are too big to run on a phone. So they need to run on a server. But beefy servers are expensive. With premium I’ll be able to afford that so that you can take unlimited pro images!

Here are the most common feature ideas:

• Import Collections from XML

• Better Set Management

  For example, if you want to build a set and are searching for parts of it, simply check off parts that you’ve found. Better handle figures inside sets

• Show percentage parts owned for a set

• Color Detection

These closely match the features discussed on the feedback forum. Most users haven’t checked the forum yet. I encourage you to use it and upvote posts to help me prioritize. I’ve written more about it here.

I also loved your answers to the question “What would need to change to make it worth $100 / month? (as a thought experiment)”. There is so much cool stuff to build! These ideas are more advanced, but with enough time and possibly team members, I’d love to tackle them:

• Pan across the collection and see in real-time/AR mode.
• Make the app a marketplace. Be able to buy and sell items.
• Improve marketplace integration. For example, take an image and immediately create a listing on eBay.
• Take a picture of LEGO pieces, identify each one, and then show what you can build.
• Build a sorting robot

I am very happy that only a small number of people encounter app crashes or bugs. If you come across anything, please let me know immediately at [email protected].

It’s been a long journey

As a child, LEGO was my favorite toy. I guess I am not alone if you’re reading this :) But as I got older, my interests shifted to RTS games like Age of Empires and StarCraft 2. And after high school, I became fascinated by programming and started studying computer science. During my master’s degree, I really got into computer vision and deep learning.

On Christmas Eve 2020, my nephew was building a LEGO set and we talked about how cool it would be to reuse the LEGO in our basement. It would be incredible to know which sets are hidden there, how to put together the minifigures, and how much our collection is worth. However, trying to look things up online to figure out which minifigures and parts we owned was extremely tedious. It seemed to be perfect for an image recognition app. The idea started to take shape.

Around that time I was doing an internship on a computer vision research team at Google and I enjoyed it as well. So when I finished my master’s degree at the end of 2021, I had a decision to make: work in a big tech company or pursue a little app idea.

I made the choice to put my time into a project that helps people enjoy their hobby more. And I’ve been on it ever since.

It was a lot more complicated and painful than I anticipated.

The first 10 months were rough, working alone on a simple toy app that didn’t have any users. Back then, the app could only identify torsos. It took a long time to make it available on the app stores. At some point, I started texting individual BrickLink sellers via Discord to see if they wanted to test the app. Some feedback trickled in. Then, in October 2022, I shared the app on Reddit and it went viral! 25,000 upvotes. The comments were extremely positive and encouraging. This was 1.5 years ago, and since then, there have been many updates and improvements.

Sep 12, 2018This is the first LEGO-related identification image on my phone. More than five years ago, I looked at 3D LEGO models to train with generated data. At that time, I was dreaming about a sorting robot.

Apr 18, 2021To identify pieces, I wanted to get images from many different perspectives, so I placed LEGO pieces on a turntable and took a video.

Aug 18, 2021I started with just identifying Minifigures. My entire family helped to sort, take training images and beta test. Special thanks to my Mom, my siblings and my girlfriend at the time 🙏❤️

But from then onwards people started using the app and I quickly figured out that running an app costs quite a bit of money.

Here are the main expenses:

• High-performance GPU servers for neural network inference. These are necessary for accurate and fast detection results.
• Sending countless images to your phone. If you search for an item and then scroll down a bit you’re quickly requesting 100s of images. This adds up in hosting and data transfer fees.
• A reliable database so you can back up and access your collection on any device.
• Various ongoing costs such as Apple’s annual developer fee, hosting services, domain registration, legal upkeep, company fees, backup storage, software tools and other developer necessities.

It was only free before because I worked without pay for more than 2 years, burned through all my savings, and friends and family helped out with some loans.

Premium Version 👑

It’s a bit scary to ask for money. That’s why I’ve put it off for a long time.

But in the survey I asked you what you’d be willing to pay and I was extremely excited to see that the app is clearly valuable to you. And most respondents agree that the main difference between the paid and the free version should be a daily image identification limit.

If you’re using the app occasionally to identify items, check how to put together a Minifigure, view prices or set inventory and store your collection, it will stay free.

But if you’re taking more images and storing more collections, then you’ll need to use premium.

This makes sense. To identify an item we need to use a neural net. It is too big to run on phones so we need to pay for beefy GPU servers and they are costly.

• Why a subscription instead of a one-time payment?

  There are two main reasons: first, I have ongoing monthly costs based on the number of users and they need to be covered. Second, you can be sure that I’ll work hard to make the app the best it can be and support it long-term because it is in my best interest.

The Future Is Bright

I am extremely optimistic about the future of the app. There is so much cool stuff and so much room for improvement!

The difficult technical issues in the backend have been worked out, and it’s running smoothly now. So, I won’t need to spend months working on backend changes that are not visible in the app. The subscription logic is finished now as well, and everything is up and running.

I cannot wait to go back to improving the app. I hope to make some quick and meaningful updates to the user interface and then work on other useful features based on your feedback and prioritization!

I would love to welcome you to the premium version if you want to support my work! For that, you would simply need to click the “👑” button on the home screen.

Keep building and scanning!

Jakob

See you next time!