DIR 2 7 2020 scalehub.pdf
Document Imaging Report
Business Trends on Converting Paper Processes to Electronic Format
February 7, 2020
ScaleHub Explores New Frontiers of Capture
The document capture industry is evolving rapidly. A few years ago, Harvey Spencer Associates first defined where the market is heading as "Capture 2.0," which includes characteristics like multiple types of input utilization of cloud services, and the application of intelligence for better understanding of content. ScaleHub, a German-based capture outsourcing specialist that has a U.S. office in San Francisco, seems to be fully embracing these concepts. Already, known for its use of crowdsourcing to drive AI-based document capture, ScaleHub is expanding its service to include capture for computer vision applications.
"We have taken our company into a new dimension by expanding our interpretation of capture and what we bring to the market," explained Dan Dubiner, CTO of ScaleHub. "For many years, capture has been considered something that is related to operational data. You use capture software or a service bureau to extract information from documents. But, what happens if you need data related to computer vision?
"We have already seen some BPOs transition into capture that serves other needs. In particular, we've seen them take on training for AI applications. We have a set of services that allows us to create 2D bounding boxes, polygons, semantic segmentations, and everything else that users needed to train computer vision and NLP type technologies."
Scaling the capture business
Since 2016, ScaleHub has been honing its crowdsourcing capabilities in the document capture space. It has enterprise customers like School and Student Services by Community Brands, for which it processes millions of tax documents each year. "We will be able to take all the experience we have with enterprises, in terms of customer service and scalability, and transition it into a new world," explained Torsten Malchow, an experienced capture sales executive who recently joined ScaleHub as its first Chief Revenue Officer.
Malchow is the brother of Olaf Malchow, founder and CEO of ScaleHub. Previously, Olaf founded foxray, an enterprise class capture software ISV that focused on the BPO space. Torsten served as foxray's Chief Sales Officer, prior to its being acquired by ReadSoft in 2012. Torsten then spent four years at ReadSoft, before it was acquired by Lexmark, which later acquired Kofax, where he spent just over a year. Torsten then worked as ABBYY's head of global enterprise sales until the end of 2019.
"After my departure from ABBYY, one of the first calls I made was to my old colleagues from foxray," said Torsten Malchow (Ralf Gobel, the COO at ScaleHub, also spent more than 10 years at foxray). "ScaleHub obviously has sold some projects, but there wasn't anyone focused just on doing sales. I recognized that ScaleHub's solutions were ready to go to market and that the only missing piece was sales power and visibility. We decided it made sense for me to join to develop and execute a global sales strategy in targeted industries and regions."
Those regions are North America and Europe, with the ability to leverage relationships into global deals. Even since his days at foxray, BPOs and shared services centers have been on Malchow's primary list of customers and that should continue at ScaleHub. School and Student Services, for example, utilizes ScaleHub to process tax documents that parents submit when applying for aid at private schools. Peak volumes can reach hundreds of thousands of documents per day during the school application season, and ScaleHub can turn around data from submitted tax documents in 24-48 hours-a significant improvement over what SSS was getting previously.
ScaleHub utilizes a combination of intelligent document capture software for automated extraction and secure crowdsourcing for verification and exceptions to drive high accuracy and fast turnaround times. "It's our combination of artificial and human intelligence that differentiates us from traditional capture vendors," said Malchow. "I have worked at so many of these vendors over the years that I know the limitations of their products, especially in the current market.
"Buying behavior has changed dramatically during the past few years. Previously, most capture projects were decided upon by IT. Now, decisions are more or less being made by business users who have no clue about data entry and are not interested in doing it. If you work with any of the leading capture ISVs, if their software is good, it will deliver about 80% extraction rates and then the customer has to do something with the images to get the rest of the data."
Malchow also views ScaleHub's subscription-based licensing as an advantage. "Almost 80-90% of customers I talk to are asking for pay-per-use models and are not interested in buying perpetual software licenses," he said.
ScaleHub continues to increase its certifications around security, recently achieving a level of HIPAA compliance as well as ISO/IEC 27001:2013 certification for data security. It has also developed a pair of vertical solutions, currently being marketed primarily in Europe. One is for documentation used by pharmacies and the other is for bills of lading related to the transportation industry.
Making mTurk work
What separates ScaleHub from many traditional BPOs is its use of Amazon's Mechanical Turk (mTurk) service for crowdsourcing. This concept was pioneered by Dubiner when he founded FasterAP in 2013 and the technology has been merged into ScaleHub. "Depending on what the customers want, we have other options we can use," said Dubiner. "But, we are very happy with mTurk, we are committed to it, and we are one of Amazon's largest partners for mTurk." (ScaleHub does not currently utilize Amazon Textract for data extraction.)
A huge opportunity
Malchow noted that he recently read that graphics processor specialist Nvidia is currently employing 1,500 people dedicated to computer vision training. "The data scientists at computer vision companies have high salaries, so they are not going to be responsible for data labeling," he said. "And you need high-quality and accurate data to enable tasks like a car driving on its own. A one percent error rate is not acceptable due to the risks involved. Drones also require high-quality data sets to be able to navigate effectively. We are also looking at smart city initiatives. Being able to put a data labeling application out to the crowd represents a huge opportunity for future growth."
Dubiner noted that a Scale AI, a San Francisco-based company that offers to help build AI systems by using "a combination of machine learning and human insight to label and annotate" data, recently received a $100 million Series C round of funding that values the company at more than $1 billion. Perhaps we should start calling Capture 2.0, Capture ² due to some of the exponential valuations it is leading to.
"We are not abandoning the document capture market," stressed Dubiner. "We are just extending the spectrum of what we are covering. Closer to documents, we can also utilize our platform to extend into areas like NLP for transcribing classification data."
"I think companies like ScaleHub will help change the market," concluded Malchow.