First see, then believe. That is the opinion of Cees-Jan Smit, Director Accounting and Reporting at clothing label G-Star. The result of the trial by fire to which he subjected Pipple left no doubt: Pipple makes more efficient.
Flexible and substantive from the very beginning
A flexible attitude and confidence in quality ensured that Cees-Jan chose Pipple. ‘I also spoke to other agencies that can broadly do the same thing. They gave a sales pitch and preferred to lock everything into SLAs. Pipple doesn’t. He thought along with us and almost immediately hooked up a knowledgeable subject matter expert. One that, given pipple’s size and flat organisation, would remain our expert throughout the process.’
Automated invoice classification
Together with Pipple, Cees-Jan formulated a test case. ‘Our finance department receives 40,000 emails from suppliers every year containing a PDF. That attachment can be anything: a letter, an invoice, a reminder, a question. Employees open the emails and attachments, review them, drag them to another inbox that then reads our ERP system, after which a person checks whether everything has been read and retyped correctly. Every day, six or seven employees spend more than an hour on this. If we automate that process, it will save us one FTE.’ The test case would have been successful if Pipple’s software processed more than 70% of the mails correctly.
Pipple may have won Cees-Jan’s trust with her pitch, but in order for the test case to succeed, she also had to conquer the heart of the IT department. Cees-Jan: ‘IT departments are never so fond of small initiatives from the business. Afraid that an FTE less with us will cost them an FTE extra for the management. My colleagues were therefore critical of whether what we wanted was really necessary.’ Pipple was able to counter their objections with substantive knowledge, until IT was convinced that the new application would put minimal burden on them.
Agreed success rate far exceeded
For a month, every incoming supplier email was processed twice: by an employee and by Pipple’s algorithm. The outcome? More than 90% was processed well automatically. Such a result would give the average citizen courage. Nevertheless, Cees-Jan thought it was too early to fly the flag. “A test is just a test. In real life, things can turn out differently. I’ve seen that often enough.’ It was only when the performance remained high in production that Cees-Jan was satisfied. Although, as a right-minded perfectionist, he wants to increase the figure even further. ‘That will certainly work, because the application is self-learning. By analysing the manual corrections, the filter criteria are tightened up a bit each time.’
I also see all kinds of opportunities at a strategic level.
The results achieved encourage Cees-Jan to philosophise about future projects. ‘With invoice analytics we can detect suspected errors. And if we use data over longer periods of time, we can even have an early warning system built to prevent outages.’ He sees that Pipples’ potential goes beyond practical applications. ‘Data science is still the party of Finance for us, but I also see all kinds of opportunities at a strategic level.’
