One of the best parts of my work is interviewing the founder teams of IDP startups. There I find the most innovative ideas. As a former founder and member of several startup teams, I also carry a deep appreciation for the daunting challenges they face. Making payroll, bringing a tech idea to MVP, closing those first deals, raising money, building market awareness out of nothing, etc.
However, there is one challenge no startup should ever have to face: just when the product is ready to launch, a devastating war breaks out in the country.
This was (and is still) the incredible challenge confronting RaccoonDoc, an IDP startup from the Ukraine. I first noticed the product because of its unique name. Today I spoke with the founder Igor Svitelskyy. The first question was the one I thought you all wanted to ask too: why Raccoon? Turns out this is his son’s favorite animal. The logo is quite cool.
RaccoonDoc appeared in 2021 out of Innora, a Kyiv-based system integrator specializing in RPA solutions for enterprises and a UiPath partner. Innora found that some 50% of RPA projects also needed document processing and data extraction. They tried to implement several existing text recognition products such as Tesseract, ABBYY, Microsoft, Google, AWS, and others, but could not achieve the high level of accuracy required to support straight-through processing for the RPA. Much of the problem with the existing OCR products was due to differences with the Ukrainian language, which Igor knew his team could fix.
RaccoonDoc is similar to many other startups who grew out of a frustrated system integrator’s experience with the limitations of existing products, and thought why not build our own? Igor said RaccoonDoc is highly focused on the accuracy of the data extraction; makes sense, as we know how bad or incomplete data can completely derail process automation.
The RaccoonDoc offering is 100% cloud native and marketed as “IDPaaS” (IDP as a service). It is built on the latest machine learning (ML), computervision, and neural network technologies. Customers upload their docs into the RaccoonDoc cloud where the RaccoonDoc team finetunes the ML models to fit, the server performs the data extraction, and RaccoonDoc’s inhouse Human-in-the-Loop (HITL) team reviews the results and guarantees 100% accuracy for returned data. RaccoonDoc also provides a browser UI for the customer’s team to take part in training the models and correcting low confidence results. For most straight-forward semi-structured documents such as invoices, 10-20 samples are needed to finetune the ML models. More complex, multinational forms and invoices may require up to 100 samples. No matter, because the RaccoonDoc team does not charge extra for the fine-tuning.
Despite the circumstances, RaccoonDoc has managed to close its first enterprise customers earlier this year and has some big deals in the pipeline. The company plans to expand into Europe, but understandably the war has put that on hold for now.
As I said at the start, I have the utmost appreciation for all the people who start software companies and then persevere through the tough times. But also doing it in a war zone? Wow – extra respect goes to RaccoonDoc.
To Igor and the team, I think I speak for the startup community around the world: we all applaud your innovation and your courage, and we wish for you and your families to have peace and security once again. And may your IDP business prosper.