Recurring complaints about shadow removal, border detection accuracy, and export flow friction.

Most common failure modes in scanning physical documents.

Competitive scan quality benchmarking against iOS native camera and other scanner apps.

List view exposes document name, date, page count, and context actions at a glance — making large repositories navigable without opening each document.

The existing flat structure couldn't support how power users actually organized their documents. Nested folders + easy document moving replaced the workaround of redundant root-level stacks.

Repositioning all capture controls to the bottom thumb zone made the core action accessible one-handed. Research confirmed this matched real capture behavior in the field.

Tested transparent frames that showed the document beneath guides. User testing showed an 18% quality drop — users zoomed in to "help" the detector, causing focus issues. Reverted to opaque guides with A4/Letter aspect ratios preserved.

Keeping edit controls open while swiping between pages turned a repetitive 10-step cycle into a single continuous session. Batch operations became practical for the first time.

Introduced clear before/after transitions and explicit apply/discard controls. Users could see what was changing and confidently commit — or undo. Support tickets about lost originals dropped significantly.

Instead of buying a dataset, we engaged our 500+ beta users to tag 4,000+ documents in one week. This gave us real-world data across document types our users actually scanned, not generic stock samples.

92% accuracy on document type detection (receipts, business cards, contracts, forms). The optimal color mode — B&W, color, grayscale, photo — is selected automatically. Every scan improves without the user having to think about it.

An adjustable corrector with ML-powered border smoothing. Removes defects or sensitive information and reconstructs the background behind them — so the scan looks untouched rather than redacted.

The most common scan artifact — fingers in the frame — is detected and removed automatically. The background is restored without any user action. Both features shipped with dedicated GTM onboarding screens at launch.

Search moved from buried to prominent — accessible from anywhere in the repository. Not a layout tweak: it became the primary navigation method for users with large document collections.

Users can now search by the text inside their scans. Chose an OCR engine with strong multilingual support — Scanner Pro has users in 60+ countries and searches needed to work across alphabets and scripts.

First-mover advantage on new platform capabilities signals an active, evolving product — exactly the opposite of the "abandoned" narrative we were fighting. Designed across all three widget sizes and both light and dark appearances.

One tap from the home screen to the camera — skipping the app entirely for the most common use case. Recent documents surfaced in the large widget for quick retrieval without opening the app.