Gridsum Holding Inc. (NMS:GSUM) is a company that provides data analysis software for multinational and domestic enterprises, and gridsum guosheng qigovernment agencies in China.  The company’s CEO, Guosheng Qi, sat down with Executive Casts at the company’s headquarters in Beijing, China to discuss some of the most important things that will help investors connect with and learn about the company beyond what is usually gleaned from brief post-earnings conference calls.  Below he goes into detail about the Evolution of Gridsum, big data, and product innovation:

  • Gridsum’s Early Beginnings and Evolution
  • What Gridsum Stands for and on Big Data
  • Gridsum’s History of Product Innovation
  • New Industry Opportunities for Big Data and Gridsum

Gridsum’s Early Beginnings and Evolution

I already give you one story about how I get into search engine and data analytics. So, then, I realized, “Oh, everybody on the Internet is developing so fast.” And every company is talking about the impact from the internet, I think like in 2005. So, I think that is maybe my chance.

So, what can I do? I’m familiar with search engines. I’m in the first group of people got impacted by search engine, familiar with the data analytics, I’m very good at looking to logs and figure out what happened with your website, on your website. So, I say, “Why not I start from that?” My early time I started some like a project-based data analytics projects for several website owners, like several big websites back to that time – to help them to build some analytical tools and help them to figure out what is happening and what happened. And then later on I said, “Oh search engine is so popular and there’s Google, was in China and a Baidu just took off.” I said, “I am a familiar with search engine why not we do SEO, search engine optimization for companies?” And search engine optimization relies hugely on data analytics. So, then, we did that.

And after a while in 2007, we realized one thing – that a project-based business could not be scaled. So, if we wanted to create a really great business, great company that you need to have your own products and so we made a very tough decision because we have a limited resource, we have a limited people. I think in 2007 we had like 20 employees and then we stopped most of the projects, project-based projects.

Basically, we killed our own revenue and decided we’re going to make our own web analytics tools because during the services for the early years, we were saying people using Omniture, WebTrends, Core Metrics, all these foreign brands web analytics is solutions. Since China internet market became more and more different than in the Western world, I realized that these tools, these products are not that easy to use in China. I think we have better technology, I think that performance wise we can beat them. So, that’s what we did at the early time and we stopped all the project-based projects and I convinced my father to inject some more money and make sure we won’t die for two years.

So, we built our first product named Web Dissector and Web Dissector is our first purest SaaS, pure cloud-based web analytics solutions and performance is extremely good. So, you can basically run a cube like an overlapping engine which basically provide you a multi-dimensional unlimited drill down functions on that interface back to that time and that you can basically drill down data on the fly and without sampling the data, which is much more advanced. Even from a pure technology wise, it’s much more advanced than all these competitors, but these competitors back to that time, they’re not big data. So, they use one server to process all the data, but we use a global server connected together to process the data. So, in most of the case, if you use these softwares, our competitors, they need to do samplings. They cannot process a full data if your website is huge, your amount, volume of data is big and they cannot process, but we can process. And not only that we provide a multi-dimensional drilldown like you are playing with a [pool] table on the interface. Even today this is also still killing futures because the full amount of website data is increasing and increasing. Everyday it’s increasing and bigger and bigger.

What Gridsum Stands for and on Big Data

I think I’m a very lucky guy. I made some early… When we had our early time, I made some decisions and had some visions of the technology, the future technologies. I am very lucky to bet on the right side. So, Gridsum is actually a project. Personally, I code it, I made in 2003. So, the idea was that, as I mentioned, I was very into data analytics, digital marketing.

I was thinking about, “Oh, what will be the next generation of technology?” I think that once you are going to face like a challenge which everybody is going to face will be the volume of data but increasingly bigger and bigger every day. And the increasing speed of the data will be higher than the more slow. So, what does that mean? It means that if you always update your computer, you always apply the latest CPU, the latest memory, the latest hard disk, you can do that. But if you are using this strategy versus a continuously fast growing volume of data, one day you cannot process all your data on time because the data is increasing speed is higher than the hardware upgrading speed. So, what if this happened? What should you do? I was always thinking about that if you can connect a computer together and put a group of computers together, working even for a single task – that is my idea.

So, we have this idea of Gridsum. So, the “sum” is like adding things together, that’s very simple calculations. Computers designed to add things together. So, “sum” is the most basic calculations…a computer is designed to do. And “Grid” is grid computing. So, back to that time there’s already grid computing, the computers are connected together, but they’re not designed for together for a synced processing, they are together for a more sophisticated, more complex like pipelines, pipelines processing things. But my idea is different. I think it should be, one day, even a “sum” the volume, you add too many digits together and even a “sum” needs to be done by distributed computing framework. So, I created something to try to put computers together to do analytics, the basic concept is Gridsum.

And as you may know, you study a little bit technology industry, you know the real idea behind the big data, the reason why big data becomes big data it is because of paper written by the founder of Google in 2004. He wrote up paper, introduced an algorithm, a concept called map-reduce. Basically, the idea is that if we have a huge map, for example, if we have a huge map to process and you cannot process that on time, you can always divide the map into smaller ones and assign it to a group of connected computers. So, as these computers process each smaller part of the map and then after you process all these smaller maps, you can always put it back. Same about Gridsum, because no matter how you divide a sum, you can always put it back and the result won’t change. So, I was so lucky about, you know, one year earlier than the founder of Google to realize that idea.

So, back to your first question, that’s why our performance is better when we decided to make our first web analytics product – its because we are the very first school of people, even world wide, that realized the importance of distributed computing. […] So, most of the algorithm, the algorithm existed in the world for a long while, the old artificial intelligence, AI algorithm, was there for many decades actually because it’s a simple math. But why you cannot do that? It’s because you don’t have a good distributed computing infrastructure.

So, this is now why a lot people use GPUs because in GPU there’s distributed computing, there is parallel computing. So, you can do things at the same time, to greatly reduce that time consumption for the AI, artificial intelligence, that’s it.

Gridsum’s History of Product Innovation

So, from beginnings when we bet on digital marketing, we bet on web analytics, we launched our first product named Web Dissector in 2009, that was a major success. As I mentioned from all the different angles we’re better than our competitors. So, we quickly replaced our competitors especially from multi-national corporations because multi-national corporations they always have like their half quarter purchase and they bought a license from the software and at last their original office in China, use the software and it’s not that easy to use as I mentioned. So, we go to introduce our software and convince our software is better and yes, it’s not that hard to convince, basically. You have the sales guy spend 20 minutes and people would say “I want to try.” And give them two weeks to a month of trial period and they will decide to buy.

And after that we realize that our technology can actually be used in other sectors, maybe. Because we have proved our technology is superior and we have better performance by our distributed computing framework. So, why not we try some other sectors? Then, we look at the government because China has huge population and the government needs to… There are a lot of things you need to – you need to apply for a driving license, you want to get married – and without internet the government hall is always very crowded, it’s a long queue. But with the internet, then, the government started to build websites to serves their client, citizens, better. It’s a great idea. So, we say, “Why cannot we help the government? We’ll use similar technology.” The government also has their own website owners so then we start this business, we had a joint venture and MOU with the State Information Center and started that which is also pretty good.

And then later on, we say, hey which is the video become very popular and most of the TV stations, very traditional TV stations, stayed on the TV station. In China, all the TV station basically stayed on and they’re very good at offline cable TV. Once everything moving on to the internet, they started their internet TV. I said, “Okay. This is also a new challenge and this is a new opportunities.” Why not utilize our data analytic to help them to figure that once they’re moving from offline to online, that transmission effectively, efficiently and also a very interesting thing is that all the time, the ratings is critically important for TV stations and program makers for ratings, it’s all about ratings. But now as long as your entire platform is moving from offline to online, you have data transactions exchanged from terminals and to your server. So, you don’t really need a rating agency, so you can collect the data as real time without any sampling. Basically, you know your ratings without samplings, that’s pretty amazing. So, we started our new media business and we apply our technology into TV stations and now 80% of the biggest TV stations, our clients using our solutions.

And the latest thing, that I’m mostly excited [about] is recent efforts which we did was legal services sectors. Because once we designed our infrastructure, we plan not only for some structured data, which is like adding things together, like summary of structured data. But we wanted to make sure that this platform can also be used to analyze or process unstructured data. For example, natural language, like text, images, they’re unstructured data. So, we invested very significantly on Chinese natural language processing. And Chinese is one of the most complicated languages in the world. We have no space between two words. So, now like Western language, you guys have space between words so at least the computer can figure out this is a one word and this is another. But in Chinese, computer cannot easily figure it out. So, that is the hard part and we invest hugely on Chinese natural language processing, try to train computers to understand, for example, the legal language. How lawyers and judges when they use their own professional language on legal documents. What will  they speak, and that is some technology we developed for several years and we put it into using the legal services sectors in 2015, started in 2015 and started to commercialize it, monetize it last year and this year we were very successful.

We got contract from Supreme Court and we corporate, we signed MOU with the Supreme Court Press, who was the biggest knowledge based sellers in the entire China judicial system. They’re famous because they’re press and they’re controlled by Supreme Court so they have an all kind of best practices in the judicial system. And we have technology to help the judges and the lawyers, the corporates who can more efficiently review all the documents, all the evidence. And even can help them to automatically draft 80% of verdicts based on the court hearing data

We also incorporate with Tencent because Tencent has one of the best speech-to-text and text-to-speech technologyies, they’re very widely used. So, we incorporate with them, we embedded our expertise in our natural language processing technology from deep understanding of legal language, integrate with their speech-to-text technology and have this small court, part of we call a small court solutions which is already being used in several courts in China. There’s always a person who wrote notes, who is recording all the entire court hearing process and now it’s 95% accurate. So, basically that guy sits here and monitors if the computer made any mistakes and he stops there and he can change it. But 95% of his job is done automatically. And it’s very efficient and we’re very happy that we have some technology which helps the industry become better. So, in the future, we’re planning to apply our technology into even other sectors, more sectors.

New Industry Opportunities for Big Data and Gridsum

Yeah, I think we’re like a pure digital, pure software. So, there’s like digital marketing, it’s a pure digital world. Even the legal things this is all digital documents, right? You protect all the digital data and the data is when we have this, when we collect the data it’s also from a pure digital content. But we are thinking that, for example, so I’m familiar with the industrials because my family was in heavy machinery and in the mining business, right? There’s the machinery mining business. So, there are sensors, for example, in big machines and the sensor was designed for warnings. There is a so high temperatures or high pressures, but we didn’t use that data, we didn’t even collect that data. This is just used for like a warning, alerting. So, I think that is what we call an industrial 4.0.

Actually, we already had one project that’s going on, it’s a trial project with very famous car makers, local car makers. And they gave us a lot of data collected from their car and drivings about the oil consumption data with also the pedal pressures with pedal levels data and with the different other data like if they’re turning or not and if you wanted us to figure out something, use our data mining technology, and they can have a better algorithm on their transmissions. So, we have a trial project on that, it’s a real project and we signed a contract and obviously this is a big sector in the future, especially for […] big manufacturer from very traditional to smart manufacturers.

And similar technology, it’s all similar technology, it’s all about figuring out abnormal patterns automatically like we did for many years in our digital marketing sectors for anti-fraud, anti-click fraud and it can also use to be the industrial you find a lot of abnormal data patterns that must mean something. You hand it to an expert and the experts say, “Oh! This that, this is because of this, it’s because of that.” The computer learns and the computer becomes smarter. So, it’s all about that and people cannot just manually go through all this data and also for the financial service industries, for example, auditing.

So, after we became a public company, we realized there’s so many paperwork that needed to be done and the auditing work is tough. The auditors are here, review everything manually and review the contract, review all the invoices, everything. I was thinking about how this can be done. This is the part of the job, but this is very similar to a lawyer reviewing all contracts, all the documents. That can be done semi-automatically, at least, automations. I think automations is everywhere. A lot can be automated. So, we mastered data analytics technology, we mastered technology which can process both structured and unstructured data efficiently.

The approach will be to find an industry, find a sector who has a very well-educated person and every day or every week their business, their job is dealing with data.  That is a part of their job at least every week and then we have to look at their workflow and say if our technology can help them first automate your workflow and business process and then during the business process and the workflow automation, we’ll collect a lot of data. And we’ll get to know their industry, get to know their business much better. The data we collected is big and large enough to make predictions and then we can move the process from business process and a workflow automation into thinking and decision-making automations, which we had already done on our digital marketing sectors.

So, we collected so much data on how much money they spend today, tomorrow, yesterday and a month ago, a year ago, three years ago, we have all this data. How much money they spend, how much outcome, how much KPI they achieved. And we can plan for tomorrow and next year. They said that, “I want my budget is to look like this.” And will my KPI be realistic? Computers can answer. So, that’s very important. It not only analysed if your history is good or not, but it helps you to make your plans, it’s your future. All people would say, “I really want to reach this KPI level, so, how much money should I spend? And how should I allocate this money to different media channels?” We can definitely help if you come with us for more than three years.  We’ve been proved to be very accurate.

Please view the full interview here. https://geoinvesting.com/gridsum-holding-inc/

 


gsum ceo Qi GuoshengGuosheng Qi
Chairman of the Board, Chief Executive Officer

Guosheng Qi is one of our co-founders and has served as our chief executive officer and chairman of our Board of Directors since our inception. Mr. Qi founded Beijing Gridsum in 2005 when he was a student at Tsinghua University. Mr. Qi holds a bachelor’s degree in computer software from Tsinghua University.  See more about Gridsum here.