Computer Vision Drives 20% User Growth for Rubik's Cube App
21Moves had a Rubik's Cube solver app with real traction — but a fragile computer vision scanner was holding users back. We rebuilt the pipeline from the ground up, lifting scanner success from 60% to 80%.
A scanner that worked — until real-world conditions broke it
21Moves had built a Rubik's Cube solver app with solid fundamentals, but the computer vision scanner struggled the moment conditions deviated from ideal. Glare, shadows, and colour ambiguity were silently failing 40% of users every day.
A redesigned pipeline — built for the real world, not ideal conditions
Gradient Insight systematically mapped every failure mode in the existing pipeline, then rebuilt it to handle the full spectrum of real-world scanning conditions — going beyond the brief to also shape product strategy with the 21Moves team.
Full analysis of the existing image processing pipeline to locate every failure mode
Redesigned pipeline with robust handling for glare, shadows, and colour ambiguity
Precision colour discrimination logic eliminating red/orange confusion across lighting conditions
Adaptive geometry handling for non-standard cube configurations and sticker sizes
Improved end-to-end reliability across a wider range of real-world scanning environments
Strategic product input beyond the brief — shaping the roadmap for scanner capabilities
A measurable lift in users, ratings, and market position
The improvements didn't just fix technical bugs — they unlocked a growth driver. With a scanner that works reliably across all conditions, 21Moves converted more users and solidified their position as a top-rated cube app.
"They helped improve core aspects of our computer vision pipeline. Not only did they improve key functionality like glare detection, they provided a great sounding board for determining overall strategy with the product moving forward."
Similar challenge?
We fix what's quietly failing in your AI pipeline — and help you think through what comes next.
Book a free discovery call"They helped improve core aspects of our computer vision pipeline. Not only did they improve key functionality like glare detection, they provided a great sounding board for determining overall strategy with the product moving forward. We were able to go from 60% of users successfully using our computer vision scanner, to 80%!"
Your pipeline, your users — built to actually work
We don't patch symptoms. We analyse your AI pipeline end-to-end, find the real failure modes, and fix them — then help you think through what comes next.