Ingrid Pappel, a member of the research group and an associate professor of e-governance technologies at TalTech, explained that the cooperation between TalTech and Eurora started with the desire to create an automated solution that would help e-commerce providers calculate the tax amount at the time of purchase, prepare a pre-filled customs declaration after paying for the goods, and digitally send it to the customs office of the destination country.

Pappel described that on the one hand, there was a desire to automate the process to simplify it and save people time and effort, and on the other hand, there was the new VAT threshold law, which was due to come into force in 2021.

Faced with a complex problem that required an algorithmic solution involving artificial intelligence and data analysis, Eurora needed to find a research organisation to help explore the complexities of this field.

The need for automation stemmed from a 130-fold increase in the number of declarations

Pappel explained that, according to a study carried out in 2016, around 70% of e-commerce shipments went through traditional mail channels, with 65% of the shipments not paying VAT, resulting in a total loss of around €1 billion in VAT revenue for the state.

In addition, under the VAT rules at the time, consignments imported from the EU were fully exempt from VAT if their intrinsic value was below the €22 threshold (the rate depended on the amount set by each EU Member State; in Estonia, it was €22), which opened up the possibility of fraud – for example, expensive items could be discounted on import documents.

As of 1 July 2021, new EU regulations came into force to address these issues, but at the same time, a requirement to charge VAT on every shipment was imposed. Based on the 2018 statistics of the Estonian Tax and Customs Board, such a change was estimated to result in a rise in the number of customs declarations of postal items submitted to the Tax and Customs Board, which was predicted to increase from 23,000 to 3 million, or grow about 130 times.

Amendments to the same regulations created a possibility to separate the physical and fiscal flows of shipments, which helped to speed up the customs clearance process and paved the way for exploiting the potential of automation through IT and machine learning.

Pappel explained that the AI will be able to assess whether the description of the goods displayed in the online shop and the commodity code assigned to it match the textual description when preparing an electronic customs declaration. If they match, the parcel will pass through the green corridor at the customs office. If not, the parcel will be inspected upon arrival to see if the seller has been honest or tried to cheat.

For more information, click HERE.

According to Ingrid Pappel, the main role of TalTech in the project was the development of new algorithmic solutions for predicting commodity codes as accurately as possible and the analysis of specific requirements, which became the basis for the electronic declaration.

Difficulties are meant to be overcome

Together, they created an AI-based international e-commerce compliance platform that can automatically process up to 5,000 requests per second with up to 96% accuracy.

‘Of course, changes were made to the original plan during the project, as the current software development process is flexible and has to adapt to requirements that may change at any time,’ says Pappel.

As the activities of TalTech were very specific and involved the analysis of extremely concrete data and training a model based on this data, the research team only had difficulties with obtaining data that was as accurate as possible as well as with the necessary level of quality.

For example, some product groups did not have data with a sufficiently high level of quality that could be used as a basis for model training. There were also some situations where obtaining data from certain target groups was difficult. In addition, there was a certain complexity related to the architecture of the whole platform because many parties are involved in the solution, both in terms of their interests as well as the services they provide; in addition, the technical capabilities of the parties for exchanging data were at very different levels of maturity.

Pappel said that it is not just businesses that will benefit from the solution – shoppers will too. She explained that if a third-country online store integrates with the solution created by TalTech and Eurora, the necessary taxes can be calculated and paid immediately without the need for the local customs office to contact the customer.

The VAT paid on the goods will be recovered according to the amounts shown on the electronic declaration under a separate scheme set out in the EU regulation (IOSS, or import one stop shop).

‘This reduces the hassle of paying for the same thing several times and the work load of customs authorities of charging a service fee for each parcel,’ says Pappel.

A springboard thanks to a successful investment round

In the spring of 2022, Eurora Solutions raised USD 40 million in a Series A funding round. The new funds were used for the further development of the product and for making strategic purchases for expansion, particularly to the United States, the United Kingdom, and the Middle East.

Eurora has more than 200 paying customers worldwide and their B2B platform can be used by online shops, logistics and postal companies, as well as tax and customs authorities.

Ingrid Pappel was pleased to report that the development activities of Eurora continue to be fruitful and that the company is now up and running and expanding rapidly.

TalTech still has to document the main achievements academically, and therefore, the activities of Eurora are still monitored and communication is ongoing.

‘As a university, we get a lot of problem-solving tasks for our post-graduate students and many PhD students are also involved in this topic. In addition, there are certain ideas for new research calls and grants, so it is worthwhile to follow our news – some problems still need solutions and have to be explored further,’ Pappel points out.

Kaie Hansson, Chief Innovation Officer of Eurora, adds that Eurora has been running an innovation and development centre in Tartu since autumn 2022, focusing on applied research in machine learning and on testing scientific solutions.

Dozens of data scientists are working on new solutions at the innovation centre. The centre plans to grow and bring on board another 80 researchers.

‘This is the start of a great collaboration and the success story of Eurora is only just beginning. Among other things, Eurora and TalTech are in the process of discussing the involvement of doctoral students from industrial fields in the applied research projects of the innovation centre,’ Hansson said, describing future plans.