When sharing a workspace, you share more than just the amenities. You share in the good and the bad moments of everyone else. You get excited by the wins of your coworkers – small or big. Recently, we celebrated a big win for one of our members – Sigma News Ventures.

They are one of the FinTech teams in our space and are developing an app called Newsful. And they just signed a contract with Thomson-Reuters to launch it on their trading terminals in Q3/17, which secures their first distribution channel.



So, what does Newsful do? For one, it has digested every piece of company news from more than 50,000 media sources, today and every day for the last decade. Each story is logged, time-stamped and classified by business events into a uniquely powerful system. The business events are linked to simultaneous movements in the companies’ respective stock price to create models of what drives price movements in the stock markets.

To dig deeper into the idea, technology and team behind the software that unlocks new value from corporate news, I reached out to the founder Christofer Solheim. He is a Norwegian technology entrepreneur that founded Commetric and CommEq - both large shareholders in Sigma News Ventures. Prior to Commetric, Mr. Solheim worked in various roles at Hill & Knowlton Public Relations spanning 10 years, including leading H&K’s Business Intelligence unit and providing executive counsel to Paul Taaffe, Global Chief Executive Officer of Hill & Knowlton.


Christofer Solheim


How did you come up with the idea for Newsful and what were the first steps you took to start the project and get the right people on board?

Many years ago, during a dive in the Philippines, I watched helplessly as water leaked into the transparent casing of my brand-new camera. Since it’s dangerous to bounce straight back up to the surface, I tried to think of other things. Then came the first ideas for the methodology. The technology was built in Sofia, by the brilliant team of developers, mathematicians and computational linguists.


Talking about the right people, could you please introduce the team that is working on the project.

The team is fairly small but varied. Julia and Gergana develop our Business News classification engine. Ivan, Iva, and Stefan form our small team of developers who are working on the web application itself. Alex manages data processing and Eti does customer relationship/support. Everybody is backed by Polina and Mihail, who are the administrative, organizational and financial unit.



What technology stack is Newsful currently using?

Newsful is a web app built on ASP.NET.

Behind the minimalistic graphics is a fairly complex heterogeneous data processing system. We utilize a combination of Elastic, MongoDB and MySQL databases to store our Big Data: an archive of trillions of stock price records, billions of articles from tens of thousands of sources spanning more than ten years.

For data processing, we mainly use Python and R. Every day our systems harvest the business news, then enrich each text with metadata from our proprietary Business Event Classification engine. This is then merged with financial information from the US stock markets. After preprocessing is finished in our datacenter in Sofia, data are synchronised with Amazon’s cloud to feed Newsful.


I heard that it took more than 100 man-years to develop the NLP (Natural Language Processing) and Machine Learning methodologies and technologies. That’s quite an impressive number. Could you please share what other types of resources were invested in the project?

“I can make you duck!” Consider the possible meanings of that sentence. In a war movie set, you will think of cheating a bullet, in a Harry Potter book animal transformations are quite normal, but it could also simply be a promise of a delicious dinner. Natural (human) languages are full of ambiguity. We work to massively reduce that, to make computer models that understand crisply and quickly what is going on in the business world. Our specific focus is the automatic detection of key developments in the circumstances of publicly traded companies and achieving that simply by having computers automatically read and interpret news coverage.

There are three approaches:

  1. Employ specialist linguists to develop NLP rules to intelligently interpret the sentences: We do that.
  2. Use machine learning to auto-classify. We do that. 
  3. Employ humans to manually classify. We do that, too.

If we have #1 and #2, why on earth do #3?

The human classifications are both used to test the precision/recall of the NLP rules, and as a learning set for the machine learning.

How did we know what kind of news matters? Two ways. Firstly, we have trillions of stock price records the time of the news so we can see what moves the market. Secondly and importantly, Hill and Knowlton (one of the world’s largest PR agencies) worked with us to create the taxonomy of business events of the events. The stuff that the key stakeholders really care about, good and bad.


Tell us more about the contract with Thomson-Reuters. What does it mean and how it will help you develop and distribute Newsful?

It’s a revenue sharing deal to sell our technology to their clients. Thomson Reuter’s flagship platform is called Eikon, which is used by more than 125,000 trading professionals. The business model is that they will license on a per seat, per month basis. Thomson Reuters will take the responsibility for billing and contracting, and they retain a percentage of revenues.

Thomson Reuters employs 45,000 people, and we will need to spend a good deal of time with different teams to create awareness of our solutions. 

It took fifteen months to secure the TR contract, with significant resource investments by both sides. We are certainly going to invest in maximizing the channel potential.

That said, the TR distribution agreement will also raise our profile more generally and establish credentials we can gradually leverage for direct sales and sales through other partners. 


What about your customers?

Actually, after many years in the lab and using our technology for trading, the Thomson Reuters distribution deal will be the first time we offer the resources to clients. Initially, our ideal clients would be banks, fund managers, maybe brokers. And, of course, the publicly traded companies themselves.


To celebrate the partnership with Thomson-Reuters you invited all Puzl CowOrKing coworkers on a party. Tell me more about your drive to join a coworking space.

We are a startup. We have a relatively good idea of where we are going and what needs to be done to get there. However, we believe we can always benefit from additional inspiration, good contacts, and an energetic environment. When we came across Puzl, we were immediately attracted by the vibe, the community and, of course, the interior design.

Coming from a classical office, it took the team some time to get used to the constant buzz of a coworking ecosystem. But now we enjoy it quite a bit. And, of course, we enjoy the gorgeous view of the Vitosha Mountain.


Good news is even better when shared! We all wish all the best to Christofer and Sigma’s amazing team!