Careers at Crowdflower
CrowdFlower’s mission is to empower data scientists to enrich their data.
Lukas Biewald was a data scientist at Powerset, a natural language search technology firm. His least favorite part of the job was collecting and cleaning data – a process that involved tasks such as removing duplicates and identifying miscategorized elements. He decided to build some tools to make his job easier. He then realized he could launch a business off of his creation.
In 2007 he teamed with colleague Chris Van Pelt to found CrowdFlower, a provider of data enrichment and mining services. They initially had difficulty obtaining seed funding, as their offering was in a very rudimentary stage and there were fewer investment resources for tech entrepreneurs at the time. So the two focused on finding paying customers in order to raise capital.
They found a number of clients who later became angel investors – these included Gary Kremen, founder of Match.com, and Travis Kalanick, CEO of Uber. Within 18 months, their software-as-a-service solution was fully ready -- the product utilized a combination of human employees and machine learning for its activities. The company was able to raise $1.2 million in seed funding.
Benefits at Crowdflower
Business model of Crowdflower
CrowdFlower has a mass market business model, with no significant differentiation between customer segments. The company targets its offerings at enterprises, startups, and academic institutions of all industries and sizes that need to have their data collected and cleaned.
CrowdFlower offers six primary value propositions: convenience, accessibility, customization, risk reduction, performance, and brand/status.
The company offers convenience by simplifying clients’ operations. It combines work from human employees (whom it calls “humans-in-the-loop”) and machine learning in a single platform that can complete search relevance, sentiment analysis, and business data classification.
The company creates accessibility by providing a wide variety of options. Data submitted by customers for enrichment can be in the form of images, text, audio, or video, and can come from public sources or mission-critical systems of record. In addition, clients have used its solution for a broad range of applications, ranging from the analysis of Twitter sentiment for a brand to the routing of text messages to natural disaster aid workers to the identification of drug-resistant TB cells.
The company enables customization by allowing clients to personalize what their deliverable will look like. Customers can select from its large template library and customize their design settings.
The company reduces risk through high standards. The solution directs data cleaning/collection tasks to the appropriate employees based on skill set. It then tests the workers against known answers hidden within the job. The accuracy at which the workers answers the questions is used to determine the system’s level of trust in them. If that level is high, they can continue working on the task; however, if the level drops low enough, they are removed from the job and their work is disregarded.
The company has demonstrated strong performance through tangible results. High-profile examples of positive outcomes for customers include the following:
- eBay used CrowdFlower’s solution to improve its internal search relevance, completing product categorization at a rate five times faster than previous with a much higher accuracy rate
- Autodesk used CrowdFlower’s solution to clean its CRM system, increasing the percentage of customer records with complete information from 70% to 85%, translating to more sales leads
- Delectable used CrowdFlower’s solution to enhance customer service, reducing the amount of time it takes for a user to obtain information about an uploaded wine label by 45%
- Skout used CrowdFlower’s solution to increase the efficiency of its image moderation process, reducing wait times for content review to 10 minutes and saving 20% on staffing costs
- Artimys used CrowdFlower’s solution to enhance its text models’ detection abilities, improving precision by 5.2x, F1 score by 4.2x, and recall by 25%
The company has established a strong brand due to its performance. It bills itself as the largest data cleanup platform in terms of volume. It employs millions of workers from over 208 countries, and they have completed over 1.5 billion judgments for customers. Its clients include many members of the Fortune 500, such as IBM, The Home Depot, Intuit, Cisco, Google, Microsoft, Yahoo!, and Facebook. Lastly, it has won a number of honors, including “Best in Show“ at FinovateFall2014.
CrowdFlower’s main channel is its direct sales team. The company promotes its offering through its website, social media pages, and participation in conferences.
It also hosts the Rich Data Summit, a one-day annual event it describes as “the leading conference focused on turning big data into rich, meaningful data“. The conference features data scientists, investors, and prominent speakers.
CrowdFlower’s customer relationship is primarily of a personal assistance nature. Employees complete tasks for clients and interact with them if necessary. The company also offers phone and e-mail customer support.
CrowdFlower’s business model entails maintaining its SaaS platform and providing problem-solving services for customers.
CrowdFlower maintains the following types of partnerships:
- Channel Partners – Firms that provide the company with access to workers (called “contributors”) for data enrichment/mining tasks. These channels can be specialized in nature, providing contributors who are fluent in a given language such as Russian or Hindi, or who come from a specific country. Specific partners include Neobux, ClixSense, Hiving, Netpoint, Listia, instaGC, Boizu, and Wannads.
- NDA Channel Partners – Channel partners that have signed a Non-Disclosure Agreement (NDA) with CrowdFlower and provide contributors for jobs requiring a high level of privacy (e.g., involving sensitive or personally-identifiable data). These partners manage their workers in dedicated, monitored facilities. Specific partners are iMerit, IndiVillage, and Daproim Africa.
CrowdFlower’s main resources are its proprietary SaaS platform and its workforce of over a million contributors who use the platform to serve clients.
It depends on its engineering employees to maintain and update the platform, its sales staff to promote its solution, and its customer service staff to provide support.
Lastly, as a startup it has relied heavily on funding from outside parties, raising $38 million from 20 investors as of June 2016.
CrowdFlower has a value-driven structure, aiming to provide a premium proposition through significant personal service.
Its biggest cost driver is likely sales/marketing, a fixed cost. Other major drivers are in the areas of cost of services, a variable cost, and customer support/operations, a fixed expense.
CrowdFlower has one revenue stream: revenues from fees customers pay for a license to its SaaS platform. The company offers three license plans: Trial, Pro, Pro + AI, and Data for Everyone.
info: Robin earned a Master’s degree in Engineering from Cambridge University and a Master’s degree in Business Administration from Stanford University. He previously served as Vice President & General Manager, Strategic Consumer Industries at Marketo.
info: Lukas earned a B.S. in Mathematics and an M.S. in Computer Science at Stanford University. He previously served as a Senior Scientist and Manager at Powerset, an Engineering Manager at Yahoo, and an Engineer at Enablelearning.
info: Chris earned a Bachelor’s degree in Computer Science at Hope College. He previously served as a Technical Product Manager at Powerset, and has also held roles as a computer scientist, web engineer, and studio artist.
info: Cameron previously served as VP of Technology and Ad Operations in the Core Audience division of Hearst Corporation, as VP of Engineering at 33Across, as CTO and VP of Engineering at Tynt Multimedia, and as a Director of Systems Engineering at Avocent.
1.402.000+ job opportunitiesStart here!