Over the past couple of years, the fake news issue seems to have been on everyone's tongue. Overwhelmingly ubiquitous, it affects our lives in different ways — from pumping & dumping and crashing stock market prices to shaping our purchasing habits.
We spoke to Uri Gruenbaum, the CEO of TipRanks, a company ranking financial analysts, to find out the personal story behind his business venture and discover the truth in a world of fake news.
What does TipRanks do for the crypto market?
What we’re doing in the crypto market is we launched a website called CoinWatch. We tried to duplicate what we did with TipRanks but for the crypto market. … We wanted to help investors understand what is the trend really about. If bitcoin is going up why is it going up? Because nobody knows why it so volatile and there are so many investors that lost money.
We basically took the concept of Coinmarketcap [...] and we added on top of that NLP-based tools... We kind of copied what they did, because users were already acquainted with same user experience and on top of that we gave them the option to see what is the news analysis of bitcoin or ethereum today. So we analyzed 150 financial news websites. In the crypto there are thousands of websites because the industry is still young and there are so many people opening their own blogs and it's really impossible to track them all. So we took the big ones like investing.com or CoinDesk and others. And what we’re doing at phase 1 is measuring sentiment in each article.
So, CoinDesk goes out with an article “Three Reasons Why You Should Buy XRP” and we classify it as a bullish article on XRP and we'll associate it with Coindesk, see who published this. We are not yet at the point where we actually see who’s behind each article and rank them, but we do plan on getting to that point probably in the next two quarters. What we did for the stock markets — where you can see every expert out there that provides an opinion, you can see his track record. The idea was to do the exact same thing for the cryptomarket.
But the cryptomarket is so noisy because there are not just a few big websites enabling experts, everybody has a website in crypto, everybody knows why you should buy ripple or dash or anything.
You describe TipRanks as a company that advises on whether a guy can be trusted or not. And are there services for “bad” guys that help them mask themselves as “good” ones?
One of the problems that companies like us have when we try to bring transparency is that every day we’re getting many emails from people on our platform — underperforming experts like analysts and financial bloggers saying “take us out of your platform, we don’t want to be there.” From our point of view, we are not an endorsement company. We’re not here to say he is good, we’re here to say they are bad. We’re trying to protect the everyday investor — they are who we care about.
Many times we get analysts asking us to either remove them or explain why they are very good. Going back to your question, there are many services to help them look better than they are … so the biggest system in the U.S. to rank analysts is called II (Institutional Investors) and they will rank these analysts not based on how they perform but based on where they work.
So if you work at Goldman Sachs, you’re going to be ranked higher than if you work at a Tier 3 company. Or they’ll be ranked on how much money they earn... And basically, they are ranked on so many things that no one should care about. All I care about is can I trust this guy or not? So II is a good example of a company that isn’t deliberately promoting bad analysts. But they are giving them certificates that are based on me meaningless measures.
You founded Tipranks in 2012, how has the company changed since the beginning of the crypto era?
We started working on the crypto site in November 2017 and then we launched it in February this year. When we launched it, the crypto went from all going up to completely crashing. So the CoinWatch website didn’t change, the only few things that we've changed since the launch is adding localization and a few other tools.
One of our tools is Smart Portfolio, which we developed together with Nasdaq. If you go to the Portfolio you can make a simple watchlist of your stocks, you can get that watchlist analyzed and compared to other investors to see how volatile you are, and how diversified you are. People who actually have a portfolio of cryptocurrencies can compare their portfolios to many other portfolios.
We are basically creating a crowd-wisdom solution because nobody really understands crypto. So we’re trying to create a community of many different experts so that it will be completely transparent. Instead of just following a guy on the news, you'll follow a guy with a good track record. This is part of our future involvement for CoinWatch — creating a Smart Portfolio for cryptocurrencies
How do you address the fake news issue?
Our goal is not to catch the bad news, our goal is to allow you to analyze the value of experts. It’s very difficult for us to find out why the expert has a bad performance, whether he doesn't know what he's doing, or it's just bad luck or he is trying to do a scam by promoting crypto coins that are probably going to fall. It’s very difficult for us to distinguish between those things and we feel that it’s not our job to do it, it’s the regulators' job. We're helping investors to evaluate news sources and experts out there. We help the investors to get advice from the best people and we are very transparent and objective about it without being affiliated with any institution. So we’re not sure if the news is fake or not but we are sure if the person who is writing the news is good or not, historically.
What prompted you to start this business, where did the idea come from?
I studied computer science in the university and worked in high-tech companies as a software engineer and seven years ago I went online and read the article where the analyst gave 5 reasons to invest in TaTa Motos, an Indian company. I thought that would be a good way to invest my yearly bonus — so I invested all of it in it. Then, half a year later, my wife logged into our bank account and saw that we had a stock that’s 55% down. She asked why we had this stock and what I knew about this Indian company. I told her that I followed that analyst who was on the front page of a very big website and I assumed that he knew what he was talking about.
So what I did, I started googling his name and created a spreadsheet of over 50 recommendations he gave over the years and then I went on charts to see what happened with the companies that he was recommending to invest in. And I saw that he was wrong half of the times, even when the market was going up. So when the market is going up it’s very difficult to be wrong half of the times. I imagined there were services that ranked these experts and I was shocked to find out there were no such services.
Your services are clearly in demand. How will you deal with the companies who'd want their reputation to be cleared?
There are companies whose only product is to remove you from Google. What they do is kind of reverse SEO — they can write so many different things about you that all the bad news will disappear. I believe that theindustry will keep growing but I think that the big players in the fake news industry are basically the social media, it's all the Facebook and Twitter because that's how such news is distributed. As you’re not going to find fake news on CNN, you're going to find biased opinions, but it's not necessarily fake news.
Sothe only way to really fight that is to regulate the media. Give them fines for distributing fake news, which is something that's going to happen with social media in Europe. There was a law that was approved in the E.U. something like a month ago: now the E.U. can ask Facebook to remove offensive content and fake news and if Facebook doesn't comply within 15 minutes there are going to be high fines. And that was a big step in the direction of removing fake news and cleaning the media.
Could you explain the mechanism of algorithms that Tipranks uses in its work?
We have two types of algorithms. The first one is a rule-based algorithm, which means I'm going to create thousands of different rules. And every time I read an article that is published, I'm going to search for very specific patterns in this article and if they fall under one of my rules — great, I have the information. Many times articles are so complicated and unstructured that there is an infinite amount of rules that need to apply. Which means it is impossible to analyze this. In this case, we need to use machine learning, which is more statistical, to read the article and see what it is similar to.
When it comes to financial bloggers, they have grammar issues, they have typos, sometimes they are sarcastic, they have nicknames for stocks, it is so difficult to understand what they are saying, it is impossible to find two blogs that have a similar structure. For that, we need a statistical algorithm which is why we use machine learning. First of all, we took 10,000 financial articles and we gave interns the task to read those articles and classify them and asked them to classify these article, whether they are bullish, meaning they are promoting the company, or they are bearish, promoting you to sell the company or they are just informative.
We're looking for persona; opinions in order to determine if it's good news or bad news, or if it's a recommendation or not. So when we bring machine learning, what it does it reads and tries to find what is the closest thing to this article - is it good news, bad news or recommendation. So we either can work with patterns for accurate results or we can machine learning that spans a wider portion of content.
What are your company's plans for the future?
First of all, we are signing agreements with financial institutions that integrate our products. We are also trying to cover more geographical areas to find new markets, we're working in the U.S. and Canada and now we are onboarding more and more information for the European market which is very big. Also, TipRanks has a lot of unique datasets, there's news sentiment, crowd wisdom and we're trying to add more and more data sets. The issue with adding more datasets is that you're adding more value but you're also adding more confusion. So one of the things we'll be launching in the next few weeks is kind of an aggregated score of all of it.