FinVision
Document analysis using OCR technology
Client
Yellowtail Conclusion
Period
August - March 2025
Tools
Figma
Project type
Sales pitch
Project Description
To design and develop a robust and secure document analysis tool leveraging Optical Character Recognition (OCR) technology. This solution will enable efficient extraction and analysis of textual data from various document formats prevalent within the financial sector.
Introduction
Optical Character Recognition (OCR) technology serves as a crucial bridge between the visual realm of images and the digital world of text. At its core, OCR is a sophisticated process that empowers computers to identify and interpret textual characters present within raster images, such as scanned documents, photographs, or even screenshots. Instead of treating these images as mere collections of pixels, OCR algorithms analyze the patterns and structures within them to recognize individual letters, numbers, and symbols, effectively translating them into a format that a computer can understand, process, and manipulate as standard text data.
​
The "recognition of characters" is the core of OCR technology. During this phase, the OCR engine analyzes individual characters isolated from the image. It compares their visual features (shapes, lines, curves) against a database of known characters and fonts. Modern OCR often uses feature extraction and machine learning models trained on vast datasets to identify these features and predict the most likely character, enabling it to accurately recognize a wide variety of fonts and text styles, even with some degradation.​​
Briefing
The project aims to develop a secure OCR-powered document analysis tool for financial professionals, enabling efficient text extraction and intelligent data analysis from various document formats. The deliverable for this project is an interactive prototype meticulously crafted to demonstrate the practical application and workflow optimization potential of this technology. The prototype will be visually aligned with the Yellowtail design style, ensuring a professional and recognizable appearance suitable for presentation purposes.
My responsibilities
Set up requirements
Understanding the technology, translate it to product requirements and integrating them effectively.
Deliverables
Create wireframes, (hi-fi) designs and prototypes for stakeholders presentation.
Cooperation
Through a process of close interaction and shared expertise among various specialists, the final product has been realized.
Component library
Creating a custom made component library with fonts, colors, button (states) and many other UI components.
Prototype video
The video illustrates a scenario featuring Dirk Landzaat (fictional), owner of Banket Bakker Landsaat B.V., who needs to share annual financial figures with his advisor. Traditionally a manual and error-prone process, FinVision streamlines this workflow significantly. Instead of manual data entry, Dirk can now simply upload the relevant document. Subsequently, integrated OCR technology automatically analyzes the document, extracting key financial data and presenting it in a dedicated component alongside the document. This allows Dirk to focus on verification, a task made effortless by the synchronized highlighting of amounts both in the data component and the document upon cursor hover. This intuitive feature eliminates the need for manual searching. In cases where the system requires confirmation, Dirk retains the option to manually verify or adjust the identified data. Once reviewed and confirmed, a clear call-to-action (CTA) at the bottom initiates the automatic transfer of all data to his online environment.
Functionalities
Chapter recognition
Following the analysis, the system intelligently identifies headings and subsequently groups the associated amounts under their respective categories.
Furthermore, it provides a clear overview of the analysis outcome, indicating the total number of amounts detected, the quantity successfully analyzed, and whether any amounts require user confirmation.
Synchronized highlighting
The process of verifying extracted financial data is significantly simplified and expedited by the intelligent implementation of synchronized highlighting.
As the user moves their cursor over a specific amount displayed within the dedicated data component, the corresponding numerical value is simultaneously and intuitively highlighted within the source document itself.
When in doubt..
Although the OCR technology operates with almost 99% accuracy, there remains a possibility of incomplete analysis for certain amounts.
To address this, the system empowers the user to manually verify the correctness of the flagged amount. Should a discrepancy exist, the user can readily correct the value.
