15. Microtask: Extreme Approach to Digital Work, Two-Second Tasks at a Time

During his whole life, Ville Miettinen has been an entrepreneur founding software ventures. A programmer since he was a toddler, Ville is a “hacker at heart” who likes to capture the world in photographs while he travels. This passion led him to becoming a semi-professional photographer and surprisingly to the idea underlying Microtask. As told¹ by Ville Miettinen, the first version of Microtask’s technology platform was inspired by a problem of photo tagging. Miettinen had returned from a trip to South America and realized he would enjoy having the hundreds of thousands of photographs that he shot classified according to their content. Soon, he understood that it was a task that could not be accomplished with existing computer vision and artificial intelligence and that human intelligence was indispensable. Miettinen then gathered a few experts in the area to develop technology and generate business ideas. Hence, Microtask for distributed work was born of an idea for photo tagging travel memories.

¹Interview published in FoundersLY YouTube Channel.

Microtask, a Finnish start-up, is part of an emerging trend in digital services that is changing the nature of work. In 2009, it created a technology service for crowdsourcing and distributed work that combines human labor and computerized work. It breaks down routine work into smaller, short-duration tasks and distributes these to online workers around the world. Therefore, microtasks can be completed anywhere in the world where there is an Internet connection either by freelance or call center workers.

Microtasking is “the crowdsourcing of tiny tasks that can be completed in seconds.” It is a core concept of a segment in the nascent crowdwork industry that seeks to improve the outsourcing of menial tasks, taking advantage of digital communication technologies. A problem that the company tackles is that not only it is expensive to hire internal staff to carry out routine tasks, but also it is resource-consuming to manage outsourcing. Microtask is not alone in trying to solve this problem. Several companies operate in microwork in different areas and with different business models. Currently, Microtask provides on-demand outsourcing solutions for converting paper documents into digital files, a process requiring large amounts of menial and repetitive tasks, such as text recognition and data entry.

In the case of the digitization of documents, OCR systems exhibit inaccuracies that can be corrected only using human intelligence. The correction of these errors consists in verifying that the content of the original document matches its digitized version. This is a typically menial, boring, and repetitive task, which does not require highly specialized labor force or a specific work environment. Still, it can be costly and time-consuming to hire human resources to perform these tasks and audit their performance, but it is also costly to possess inaccurate information.

Microtask developed an on-demand solution to bring human intelligence into these digital processes, by transforming large workflows into standardized and uniform tiny tasks that are carried out in a few seconds. By combining artificial and human intelligence, Microtask’s service makes the digitization of documents more efficient and effective. The company’s core technology can accommodate various kinds of distributed work, such as medical doctors’ prescriptions and consumer research questionnaires.

Microtask was founded by four Finnish entrepreneurs—Ville Miettinen, Harri Holopainen, Otto Chrons, and Panu Wilska. Prior to Microtask, members of its founding team had founded companies of real-time 3D graphics (Hybrid Graphics, which later become part of NVidia, a prominent visual computing company) and wireless technologies and applications (Ionific, which later become part of Sasken Communication Technologies, a communications technology and consultancy company). Microtask operates with a technology-oriented team in Finland and a sales and marketing team in the United States. The company received seed funding from Sunstone Capital and from private angel investors, among them executives and entrepreneurs in various fields of digital technology.

“We’re bringing human intelligence to the cloud, offering human intelligence as a cloud service. Five years down the road, maybe even quicker, a lot of activities that can only be done by humans, and not by machines, are accessible through API’s, 24-7 with near real-time rates. So, if your piece of software needs access to human intelligence—let’s say they need a human to validate something or human face-recognition capabilities or understand speech—you’re just able to tap into that at a fairly low cost. Right now we’re focusing on challenging the existing outsourcing industry; that part is really what is going to be the big change because it allows building completely new kinds of applications, mind-blowing applications that you cannot even think about right now.”

—Ville Miettinen, CEO, interviewed for FounderLY’s YouTube Channel (published on March 13, 2012)

For instance, a client might need to convert into a single digital file the content of a consumer research survey that was conducted by a handful of people interviewing hundreds of random passers-by on the street using paper questionnaires. The paper questionnaires would be converted into digital files using an appropriate scanning and content recognition technology. While the software is able to recognize the handwritten content, there is an accuracy margin that allows for errors in the conversion. For Microtask, a tiny work unit would be verifying that one word converted from handwriting into digital form matches its handwritten original. All these work units, each corresponding to one word, would be sent to thousands of people “microtasking” from anywhere in the world. Another tiny task, for the words that failed the verification test, would be to type the word seen on the screen in handwritten form. Both types of tasks can be performed in a few seconds. This kind of second microtask would correct the errors made by the technology in the conversion. The client would receive only the end result, i.e., the single file with the content of the consumer survey answers in a digital form.

Microtask started when the microwork market was emerging. One of the earliest businesses in microwork platforms is Amazon Web Services’ Mechanical Turk. It is a crowdsourcing marketplace for requesting others to perform human intelligence tasks for which computerized processes are unsuitable. Providers of work select a task among the several listed and complete it for a payment amount set by the requester. Another company is CrowdFlower, a market leading crowdsourcing service specialized building optimized workflows for common tasks. CrowdFlower offers microtasking services for collecting, cleaning, and labeling data, with 5 million contributors that work in an assembly line fashion.

Microtask can handle higher volumes of tasks, has a smaller price per independent unit of labor, allows clients to request a certain turnaround time, and does not allow work providers to choose the following set of microtasks assigned to them. Furthermore, Microtask breaks up tasks into pieces and scatters them across several workers in different places, in order to protect the intellectual property and to secure confidentiality over the content of the whole work assignment. In spite of growing competition within the emerging crowd-labor industry, Microtask appears to focus mostly on services that challenge the existing outsourcing industry as that is where they are able to trigger the major changes.

“We have mashed up the document processing industry with crowdsourcing and turned some traditional business processes on their heads. What the Red Herring judges saw in Microtask is a company that is disrupting a $30 billion industry by breaking it apart into 2-second tasks that can be distributed around the world instantly and delivered back to the customer with full service-level agreements and quality controls.”

—Ville Miettinen, CEO, commenting on Microtask winning a prize awarded by Red Herring, technology business magazine, in 2011

Moreover, Microtask has long-term relationships with labor service providers, which adds to its capabilities for guaranteeing output quality and delivery time. The company contracts work from call centers all over the world where workers perform microtasks in-between calls. Typically, Microtask uses its own worker pool, keeping 10 to 15 percent of labor cost as its revenue, depending on the number of transactions. In addition to this, in cases where clients wish to use their own employees, Microtask sells “seat licenses” to distribute the tasks exclusively to the clients’ workforce.

Standardization and quality assurance address the problem of controlling output quality in outsourced or crowdsourced work, which often requires people to perform additional checks and reviews increasing the amount of resources used. Furthermore, standardization enables a pay-as-you-go pricing structure adapted to each client’s needs and resources, which is calculated based on the turnaround time required, on the amount of documents for processing, and on accuracy. As a result of standardization and automated quality assurance, the outcome of work in Microtask can be traded as a commodity because pricing is based on results rather than on an hourly wage.

Firstly, Microtask Digest is a data extraction service for processing large amounts of scanned material that combines machine intelligence with on-demand human intelligence (digital crowd labor). Digest also analyzes the whole set of documents, performing statistical analysis and building custom dictionaries that help to reconstruct broken words and to determine correct characters in poor quality document images. This means that the computerized process can also learn from the integration of human intelligence, which is an advantage over machine-only OCR software that is unable to read low-resolution documents and images.

Secondly, Microtask Forms is a service to convert paper forms into digital spreadsheets, which targets businesses that frequently use paper forms, such as insurance, recruitment, or medical companies, primarily in the U.S. market. Clients send the forms by e-mail or fax to Microtask and receive the forms as digitized spreadsheets by the end of the following working day. The service is paid per use, costs $0.36 to $0.60 per form and does not require service subscriptions or long-term contracts.

Nevertheless, Microtask’s core technology is a combination of computer vision, machine learning, and artificial intelligence, and has several other potential applications. Ville Miettinen, the CEO, reported² that, at the very start of Microtask, the founding team looked into different applications for the technology platform. The first version of the platform, developed during 2000, resulted in 10 different prototypes for different areas of application. What all those prototypes had in common was that they enabled services in which untrained and unsupervised people could perform labor-intensive tasks, for instance as part of business process outsourcing. Microtask’s founding team became involved in developing the underlying architecture of tasks that combine artificial and human intelligence, and figuring out “how to route tasks to people all over the world with the Internet, how to reach people, how to compensate for their work, how to ensure the quality of the results, and how to manage the whole process.” As reported³ by Ville Miettinen, the first version of the platform was inspired by a problem of phototagging. Eventually, the company targeted text recognition as the first area of application for their platform. Miettinen stated⁴ that, in the future, labor could be outsourced to social games by having people perform tiny repetitive tasks while they play.

²Interview for FounderLY’s YouTube Channel (published on March 13, 2012).

³idem.

⁴https://gigaom.com/2010/10/08/is-microtask-the-future-of-work/.

Microtask created the Mole Chase and the Mole Bridge computer games to easily crowdsource volunteers to help out with the indexing effort. These games consisted of simple captcha solving challenges to be completed in a short period of time. Gamification is present in other crowdsourcing projects. One of the earliest projects was Google Image Labeler, a feature of Google Images in which people helped improve Google’s image search results by competing in labeling random images. Other more sophisticated programs for image and pattern recognition include FoldIt, an online puzzle video game that uses human puzzle-solving intuition to predict the structure of proteins which is crucial to understand how diseases work, and GalaxyZoo, a crowdsourced astronomy project in which people help to classify galaxies according to their shapes.

In Mole Chase, players had to choose whether the real word was the same as the computer-recognized word. In Mole Bridge, players had to type erroneous words found in the previous game, as they appeared on the screen, which helped moles cross a bridge over a river. Successful word typing and recognition led them to the next level. The game was designed to that players cross-checked each other, ensuring quality in recognizing the words before final approval. By the end of DigiTalkoot, more than 100,000 volunteers had completed almost 7 million word-fixing tasks.

Another challenge is the change in the governance and organization of crowd labor workers. Indeed, a new kind of Internet-based labor marketplace is emerging with the proliferation of the use of crowdsourcing. This raises questions regarding employment relationships, the legal status of workers, and social protection. Some say that the boundaries between empowerment and exploitation are blurry. CrowdFlower was sued by workers who claimed that the company had violated the U.S. Fair Labor Standards Act. The lawsuit included charges of not paying a minimum wage or not paying at all for work done. Following the growth of the digital workers’ crowd, these communities began to organize. For instance, researchers at the University of California created an online platform for workers using Mechanical Turk to rate the job posters in order to “report and avoid shady employees.”

Others find microwork digital platforms crucial for increasing economic welfare in developing countries. Professors Francesca Gino and Bradley Staats, writing for Harvard Business Review, use the example of Samasource—a non-profit microwork company—to show that economic development in developing communities can be boosted by business process outsourcing supported by digital crowdwork. The authors claim that connecting new workers, who were previously outside of local job markets, to global supply chains and companies can trigger large-scale change for those communities. Clearly, connecting people to jobs online creates many different challenges for the emerging microwork industry, which will have an impact on Microtask’s current and future business.

Microtask created an innovative solution based on artificial and human intelligence which is able to disrupt the business process outsourcing industry. It consists of a software platform that breaks down routine and tedious work into tiny work units that are distributed around the world to a global workforce working remotely on the Internet. These distributed tasks are standardized and can be completed in a few seconds. An automated quality assurance tool that verifies the quality of output, together with standardization of tasks, allows for this kind of crowdwork to be scalable and offered to clients at a fairly low cost per task, which is calculated according to turnaround time, accuracy required, and amount of work. Microtask allows work to be performed from inexpensive Internet-enabled devices or from where workers live, suggesting future changes in the nature of the workplace and in the opportunities for people to perform paid work.