While Meltwater's existing business continues unaffected — and will likely benefit from the new data and AI extensions added into the platform — the company has made a big bet on its pivot to a platform strategy. It has made no less than seven acquisitions in the space of less than a year to build out the new capabilities, funded with the help of $60 million in debt finance raised early last year from Silicon Valley Bank and Vector Capital. The closely-held private company, which claims a customer base of 30,000 and is said to book annual revenues in excess of $300 million, has never taken on external equity since its foundation in 2001 in Norway with a $15k grant from the country's government.
Meltwater opens up its Fairhair.ai platform
The Fairhair.ai proposition to enterprises is a ready-made platform for harnessing insights from historic and real-time market signals found online, across a diverse range of sources such as news outlets, social media, consumer review sites and corporate filings, drawing on source material in 80 languages and more than 100 countries. As Meltwater's founder and CEO Jorn Lyseggen explains:
Historically, online data has been hard to track and analyze in a systematic and rigorous way because of its sheer scale, plethora of data types, and its multitude of languages. Fairhair.ai addresses many of these challenges and helps companies analyze the noisy and messy web to better understand their competitive landscape.
Thanks to Meltwater's nearly two decades of experience gathering these various types of online data for clients, Fairhair.ai is much more than a rich dataset — comprising more than 1.2 trillion documents from 10 million sources and growing by 600 million documents per day. Because its clients mostly want to capture every mention of a specific company or brand for PR or marketing purposes, the platform has always been specifically tuned to identify businesses and products, and to understand the contexts in which they're mentioned.
Mapping knowledge within a business dataset
This history means that its value lies in the metadata that Meltwater has built around the dataset, including the 12 million business entities identified within it, and a sophisticated knowledge graph that maps relationships. The flurry of acquisitions in the run-up to the launch have both extended the dataset and contributed the final gamechanging ingredient to unlock all this value — a library of ready-made AI models that can be configured to provide tailored insights.
Probably the most seminal acquisition was Wrapidity, a spin-out from the University of Oxford, whose engineering team formed the core of Meltwater's AI facility in London after the deal closed in February last year. Other acquisitions last year include Canadian competitor Infomart, Hong Kong-based social media analytics startup Klarity, San Francisco-based data science startup Cosmify and AI-based news search startup Algo.
In March this year, Meltwater acquired DataSift, a London-based startup that had raised some $72 million in funding and which specializes in analyzing topic data trends across social media. The seventh acquisition, in April, was Sysomos, a ten-year-old social analytics company based in Toronto.
Helping data scientists get results
The company is also working with leading academics in data science to refine its AI models and tools. Partnering up on Fairhair.ai appeals to academia not only for the huge and unique dataset at Meltwater's disposal, but also because a lot of the routine groundwork required to analyze the data has already been done by the platform's built-in knowledge graph and AI tooling. Cutting out this preparatory "data wrangling" means that researchers can spend their time more productively, as Eric Nyberg, director of the Master of Computational Data Science program at Carnegie Mellon University's School of Computer Science, explains;
The biggest challenge for current students is how to explore the large space of data, features and models available for developing a particular analytic in order to find an optimal or acceptable solution before they run out of time or computing resources.
Nyberg previously worked with IBM Research to develop Watson for the Jeopardy! Challenge and has helped shape development of the Fairhair.ai platform as a founding member of its scientific advisory board, alongside Regina Barzilay, Georg Gottlob, and Jure Leskovec, all considered pioneers in their fields.
Pushing the boundaries of knowledge graphs
Several projects at the University of Oxford, where Gottlob is professor of informatics and a fellow at St John’s College, will use the Fairhair.ai platform to further their research. It will provide a real-world testbed for two projects that aim to challenge the growing phenomenon of fake news and information online, while these and other projects will refine techniques for intelligent analysis at scale. The focus is on "pushing the boundaries of knowledge graphs," says Gottlob:
Together, we have a shared interest to explore what we believe is the future of AI — the combination of machine learning and logical reasoning — and significantly advance the greater data science ecosystem.
Enterprise developers are also delving into what the platform can offer to help them build what Meltwater is calling Outside Insight applications, drawing on its externally sourced data streams. CTO Aditya Jami cites examples of applications in four different fields:
In finance and investment, data analysis can pick out signals that might impact the performance of specific companies, or can identify correlations that help identify potential high-growth companies for investment. Jami explains:
We convert the firehose of data into a firehose of business signals. From these signals, they want to distil these down into market-moving signals.
In sales, the Fairhair.ai dataset can act as a rich source of proactive intelligence that helps salespeople target prospects more effectively, based on signals that indicate when and why a company is most likely to buy specific products.
Data science solving business problems
In marketing, the ability to monitor shifts in sentiment over time can help inform positioning, messaging and new product development. As well as social media such as Twitter and Facebook, the Fairhair.ai platform also gathers data from reviews, blog postings and other sources. By analyzing such a broad mix of data, brands can identify what's specially praised about their products versus the general market expectation, or establish what new features are most likely to be welcomed.
And in corporate communications, Meltwater's home turf, clients can create their own custom analysis or add data from their own sources that Meltwater doesn't collect. As far as Jami's concerned, there are no limits to what customers choose to do with the platform:
We would really like to push the data science to solve really hard problems. This is also helping us understand what is the right interface we can offer to help the data scientists.
We want other people to build real business solutions to real business problems where we help them build the tools ... If someone built a better Meltwater on top of that platform, that would be a great validation.
We want to alleviate the pain of a new company or a data scientist that wants to build a real Outside Insight application to solve a business problem.
What Meltwater is doing with Fairhair.ai is a very important step on the path towards the industrialization of AI. Instead of just exposing AI tools and services, or providing a raw dataset in the cloud, the company provides a carefully architected combination of tools and data, already structured with an intensively curated knowledge graph and entity classification.
That makes it the AI equivalent of what is known in the cloud world as platform-as-a-service — it provides not only the bare infrastructure, but also the tooling and data model that's common to any application built on top of the platform. PaaS has been a very important enabler of application creativity in the cloud arena and so it would not be a surprise to see the same effect in the field of AI. I'm tempted to call this a K-PaaS, or Knowledge-Platform-as-a-Service.
Of course this is still very early days for Fairhair.ai — the full platform is currently only open to early adopters among Meltwater's enterprise customers and the academic teams that have helped shape its initial form, although visitors to the Fairhair.ai website can try out some demo functions — but the venture promises to be very interesting, both for the contribution it can make in advancing data science research and for the potential uses enterprise developers can put it to. We'll keep in touch to watch its further evolution.