More efficient tax returns based on robust document scanning

German Start-Up App “Taxfix” aims to offer users a convenient and straightforward way to get their tax returns done within a few minutes, giving them an easy-to-use tool that helps them get back the amount of money they are entitled to receive. This digital approach requires flawless technical assistance through a custom UI and auto-cropping.

To support companies such as “Taxfix” to reach their goals, Scanbot offers an easy-to-implement Scanner SDK with various functionalities that can be adjusted to a company’s specific needs. In this case, high-quality scans through excellent document detection, guiding the user to the perfect scanning position were absolutely necessary to replace the previously used expensive and inefficient Deep Learning Model.

With the smooth implementation of the Scanbot Scanner SDK, “Taxfix” was able to save a lot of time regarding app development and to create an enhanced user experience. Processing documents on the user’s device simplified complex backend structures and therefore lowered the costs significantly.

According to Elster, the number of German tax reports submitted online is constantly increasing. In 2017, 22.1 Million documents were submitted electronically; in 2019 the number rose to 23.8. Since interest in handling tax reports online grew, various applications dealing with this matter have been released.

Increasing the usability of a mobile application is an essential factor when adjusting digital workflows. Most users prefer an app with easy-to-use functions and a smooth workflow that does not cause any trouble. Today, customers don’t want to spend their free time on complicated processes within mobile apps, and they demand efficient digital solutions that fulfill their needs within a few minutes.

1. Users are supposed to upload a photo of the regarding document.
2. The application provides information on how to take the photo properly.
3. After analyzing the picture, the app gives either positive feedback or tells the user that the image was “not processable” without further guidance on improving image quality.
4. Once the photo has successfully been uploaded, page detection and a Deep Learning cropping algorithm cut out the actual document, removing unnecessary information from the background.
5. OCR is performed on the document to extract the needed information from the uploaded photo.

This process was prone to mistakes since pictures were often too blurry or contained different background documents, interfering with the Deep Learning cropping algorithm. Additionally, Deep Learning was essential in this case, even though it is expensive regarding computing costs and required runtime. On top of that, improvements assigned to the algorithm also meant changes to the Deep Learning model, consuming additional amounts of time the “Taxfix” developers needed to dedicate to this workflow.

‍To streamline the customer experience, a Scanner SDK with an integrated user guidance feature was crucial. With the Scanbot SDK, Taxfix could take one of the final steps towards an error-free process of scanning documents and extracting text. Snapping a high-quality picture became possible with Scanbot’s intelligent document detection. It includes a guiding system that leads the customer into the perfect scanning position and, therefore, increases the resulting scan’s quality. This guarantees faster results regarding tax returns and finally leads to happier users.

1. The user uploads an image by following the user guide, resulting in a flawless scan.
2. On the backend, OCR is performed on the scan immediately, since the Deep Learning component has become obsolete, saving time and costs.
3. More accurate results can be returned to the user with less time invested in the process.

• Save a tremendous amount of development time
• Increase user satisfaction
• Simplify complex backend structures
• Lower costs