ɫ����

“Compared to our freelancing network, ɫ����’s platform allows us to iterate quickly with our experimental design and data collection. This is an incredible benefit to us. ”

- Hamish Morgan, VP of Data Science, Brandwatch

The Company

Founded in 2006, Brandwatch offers a digital consumer intelligence platform that delivers actionable insights to marketers, analysts and market researchers. In tracking what consumers say online about a brand or product, Brandwatch can aggregate that data to create meaningful observations for companies wanting to understand how they’re perceived in the market, and detect opportunities for growth and competitive advantage. Brandwatch collects and evaluates public text data from social media (including the full Twitter and Reddit firehose), blogs, forums, reviews, etc. After ingesting this data, Brandwatch enriches it and makes it searchable.

The Challenge

Brandwatch continuously analyzes millions of units of data by applying machine learning techniques, including natural language processing to understand content and sentiment analysis to understand attitudes and perceptions. To annotate the massive amounts of data they collect with the relevant tags, Brandwatch tried using a network of freelance analysts that they managed themselves. Turnaround and overhead were high within the group, however, preventing the data science team from being able to scale or iterate quickly on projects. Brandwatch needed a more reliable data annotation pipeline that would allow them to generate customer insights faster without sacrificing quality.

The Solution

Brandwatch evaluated several data annotation platforms before identifying CrowdFlower (rebranded to Figure Eight, now the ɫ���� Data Annotation Platform) as the most cost-effective option with the best tooling available for their needs. Unlike with other popular platforms, with ɫ����’s platform, Brandwatch didn’t need to construct their own dashboards and QA tools—these were already there as part of the offering.

ɫ����’s platform gave Brandwatch instant access to our global crowd of contributors, replacing the internally managed network of freelancers with access to a consistent stream of annotators. Brandwatch could leverage the ɫ���� crowd to perform sentiment analysis, named entity recognition, and other natural language processing techniques across multiple languages. ɫ����’s platform also included automatic quality checks to ensure consistent accuracy across annotations.

The Results

“With ɫ����, we’re getting trust and reliability. We can have confidence that the things we’re building are functionally right, that they do what we say they do while meeting the metrics our customers care about.”

- Colin Sullivan, Data Science Manager, Brandwatch

With , Brandwatch is able to ingest and analyze over 500 million documents every day. The data science team can experiment quickly using the platform, validating tasks with real-time metrics. The team can also view feedback from annotators, helping them understand right away if instructions are unclear or test questions are flawed.

Quality is key to Brandwatch operations to ensure customers receive the most accurate, valuable insights. Using the ɫ���� QA-integrated platform means the team can be more confident in their analyses, improving the overall customer experience. Brandwatch has also been able to integrate human annotation into the upgrade and release process of their models, incorporating the ɫ���� crowd as part of an automatic QA and release pipeline.

With the ɫ���� platform, Brandwatch is able to iterate at a scale and speed not possible with their previous freelance team. Becoming increasingly agile helps the team deliver precise digital intelligence insights, furthering their clients’ understanding of their customers.