Expense Management POC
Tabscanner was contacted by a large expense management and accounts payable solutions provider. Their application serves over 50,000 small and medium businesses around the world who use their services to gather and submit employee expense reports for internal review and approval.
Their existing system allowed employees to take pictures of their receipts and invoices, fill in the information and submit their report. The new direction was to automate this process for employees whilst also assisting with internal verification.
The provider was already quite far into the project when Tabscanner was contacted, and had been working with a number of large technology consultancy firms on a POC to decide on the most accurate solution. The provider required a 70% unison result on accurate establishment, date and total extraction in order to move forward with integration and a closed beta test.
The current providers had fallen short of these requirements and Tabscanner was highlighted due to tests that the CTO had performed using the Tabscanner open test platform. The results had peaked his interest and Tabscanner was invited to join the POC to evaluate our technology for the use case.
The initial test was performed on a blind batch of 500 UK and US POS receipts which were the 2 largest regions that the provider served. Tabscanner was able to achieve 68% initially which was significantly higher than the other vendors participating in the POC. Whilst this still did not meet the criteria required to move forward, the CTO was sufficiently satisfied enough to perform a final POC on a mix of 1000 new receipts from a wider range of sources.
Final Evaluation
This time, Tabscanner was chosen with 2 other vendors to provide the final POC and evaluation. All vendors were now supplied a larger range of 10,000 example documents from what the provider considered as the largest batch of stores and formats covering a much higher portion of the expenses submitted by their customers.
As a further test on the ability of vendors to refine and improve accuracy over time, they were given only 3 weeks in which to prepare for the final POC. Each vendor had to provide the cloud based API in which the final POC could be tested through and vendors were also able to offer guidance of parameters that could be passed to the API to increase accuracy of needed.
The vendors were also provided with the initial results from the first POC to evaluate their performance and the possible areas that needed improvement. Looking at these results, Tabscanner could see that the field that was most affecting the unison result was the establishment which was sometimes ambiguous due to the local store name or the larger company name. A field which was not clearly defined within the initial POC. After communications with the provider, a clearer and more measurable definition of the establishment name was identified and agreed upon as the criteria for the last evaluation.
After 3 weeks the final test was performed and the following results were obtained. Actual percentage numbers were rounded down.
