LinkedIn's big Data new gameplay

In the information developed today, I believe that anyone with a little interest in technology will not be unfamiliar with the word big data. Although most people do not necessarily know the exact meaning of it, they will still be heartened by the changes it may bring to the world. That's what makes big data fascinating.

LinkedIn

As a person who has worked in related fields before the advent of large data nouns (my postgraduate program is to build a three-dimensional climate model with the fastest parallel computer in the United States, to imitate the physical processes and chemical reactions of the atmosphere, I am pleased to see that the big data is becoming known and has shown great power. At the same time, it is disappointing that big data is increasingly being used in marketing rather than creating what is truly valuable to business and human society. That's why I started to call myself a big data advocate and performer on LinkedIn. At some of the recent events and conferences, I have started to introduce myself with this title.

What I want to talk about today is an analytical architecture called EOI (Empower/optimize/innovate, that is, power, optimization, innovation), a way for LinkedIn's Business Analytics team to use large data tools to continually drive business value. I'll elaborate on this analysis architecture below:

E: Helping information sharing

At this level, the industry's usual practice is to specialize analysis of issues raised by business partners, such as "How much money did we make last year, last month or last week", and "what are the main reasons for the big drop in the core business performance index". This may be the definition of what most people mean by the word analysis. Indeed, this analysis is important for business development because it helps managers make decisions based on such data, at least to enable them to consider using data for decision-making.

Many analysis teams spend a lot of time solving such problems. But as productivity continues to improve, the problem is that analysts are tired of repeating similar analyses over and over again. One way to avoid boredom is to use as much technical means as possible to simplify the analysis process and automate the analysis steps, such as automating data cleansing and automating data format conversions. So they can make time to do more interesting things--discover more insights into the nature of things and provide a reference for business partners.

A typical case is our team's launch of an internal analytics website called "Merlin." The site is built for the LinkedIn sales team, whose function is to generate conclusive information automatically, and team members can quickly share this information with their customers after a single key search. Every day, thousands of staff from the sales team use the site, fully self-service access to data, indicators, reports, charts and so on. The project was selected by the company's management as one of the top ten most reformed cases in 2011, and won the "Impact Award" awarded by the company's international sales department, as a result of huge support for the ground sales business and a significant economic return for the former.

O: Optimizing Business Performance

This area includes more advanced analytical work, such as in-depth analysis based on business assumptions, marketing positioning, and the establishment of a user orientation model to answer questions like "What happens if we do this," and "what is the best result?" Although this kind of analysis usually takes longer to execute, it will bring more rewards to the business. More importantly, since the knowledge base is almost always built from E, the "Help" link, analysts can better understand the nature of the data and combine it effectively with the actual business requirements.

What is often happening in the industry is that when an analysis team wants to skip the "Help" link and go directly to O, the "optimization" link, it often encounters a lack of data infrastructure and basic business knowledge, and ultimately must go back to tamping the basics before moving on to the next step. A typical example of "optimization" is the propensity model we have built for LinkedIn's advanced account business. In this model, we use user identity, user behavior and social graph data to differentiate the people attribute from the email marketing behavior. The model has become the core driving force for LinkedIn's largest online business at the marketing level.

I: Exploring Innovative models

In Silicon Valley, everyone gets excited about "innovation". LinkedIn's Analytics team has a lot of innovation. We firmly believe that the ultimate criterion for measuring the level of innovation in the team is the size of its core business impact. When evaluating the development potential of an innovation or risk project, we will focus on its potential business impact over the next 1-3 years, with its revenue, profit, user stickiness, and traffic growth being the main indicators. We also need to ensure that high-impact business activities can leverage the research results of our projects to quickly verify that our analytical solutions are viable in the marketplace, rather than innovate for innovation.

A recent example is the corporate user interest index, which we have established with the company's marketing department, to sort out the expected business customers who are likely to become LinkedIn. The key to this innovation is the possibility of translating the company into a LinkedIn client, combined with a weighted internal personal-level score and the influence of the decision-makers in the business-to-business sales process. The establishment of the system has been widely adopted by our ground sales team, which has effectively improved the customer conversion rate, and has been helpful to the increase of sales and work efficiency.

One of the questions I've often been asked since advocating this analytical architecture is what the resource mix is like under a reasonable EOI framework. In fact, based on the analysis of team work progress and the company's development phase, the cost of the EOI on the analysis of resources are also different, so you can draw a rough configuration curve. The key is that at any level of E, O, I, you should at least invest in the resources that will make a real difference and design a ratio of configurations that you think will work best for your current business growth. Overall, based on my discussions with many analyst colleagues in the industry, a reasonable proportion of resource allocations is generally e>o>i, especially for companies with double-digit business growth rates in more than two digits.

(Responsible editor: Mengyishan)