Text | Sun Jie Edited | Zhang Qiqi
Source | Chief Digital Officer
In recent years, with the popularization and development of technologies such as IOT and 5G, as well as the advancement of the process of digital intelligence, the amount of data obtained by enterprises is growing exponentially. In order to solve the collection, processing and analysis of massive data, and to precipitate data assets to better face business problems and challenges such as marketing and operation, the concept of data middle office came into being. This concept was proposed by Alibaba in 2015 according to its own business needs, and then as major Internet companies have proposed the adjustment of the middle office strategy and organizational structure, various middle office concepts such as technology middle office, organizational middle office, and data middle office have gushed out. This article focuses on the data middle office.
Sun Jie, CTO of Hongxue Technology
Why build a data middle office?
First of all, the leadership should recognize the value of the data center, and only by obtaining the trust and support of the leadership can the work of building the data center be carried out smoothly. And data can only reflect its value if it is continuously used. Data center is not a specific product or technical architecture, but a comprehensive solution or operating concept, which is not yet clearly defined in the industry. We believe that the core of the data center is to build a standard data system and the ability to quickly support front-end applications, and drive business innovation and change through data. In order to build a standard data system and achieve the corresponding capabilities, it is necessary to combine products, technologies, data, business and organizational strength to ensure the comprehensive operation and efficient operation of the enterprise, so that the data middle office can play a huge value. Measuring the value of the data middle office can be summarized into five key objectives, namely business data, data assetization, asset service, service business, and business intelligence.
Business data. For Internet companies like Ali and Tencent, the dataization of business was realized earlier. According to the online data of real-time feedback, analyze customer behavior, so as to optimize their own business processes, form a data closed loop, realize the precipitation and value of data in all aspects, and have significant advantages in precision marketing and recommendation engine. However, for traditional entity enterprises, it is difficult for business to achieve complete online and data-based. In many manufacturing processes, relying only on front-line workers or middle-level management to make decisions directly, making decisions based on limited data, data must not be effectively precipitated.
Data assetization. The biggest barrier to the full use of data within an enterprise is the presence of language barriers. Data assetization means the formation of a common "data language" within the company, and each department forms its own corresponding statistical standards for unified analysis purposes, and collects and summarizes the data in real time during the operation process. After the data assetization, data assets will gradually become the strategic assets of enterprises, and enterprises will further own and strengthen the stock and value of data resources, as well as the ability to analyze and mine them, which will greatly enhance the core competitiveness of enterprises.
Asset servitization. After the data is formed as an asset, it is opened to various departments through the data center, releasing the value of the data while discovering problems and optimizing their own production links and processes. In order to solve the problem that business units are reluctant to open up data, we need to establish relevant incentive mechanisms to improve the willingness of business units to actively cooperate and achieve better collaboration. Data middle office needs to have some core capabilities. First, the data source leader should control the quality and value of the data. Second, data applications should be realized from offline to online, from manual to automatic, and from individual soldiers to collaboration. Finally, in order to achieve business control, the abnormalities of business trends can be divided into three dimensions: time, space and data.
Commercialization of services. Any product needs to be promoted and operated, and it is crucial to commercialize data services.
Business intelligence. The business intelligence of most enterprises is still developing, and it has not yet realized true data intelligence, collaborative sharing, knowledge recommendation, etc.
How do I build a data middle office?
Taking a large enterprise as an example, its data middle office construction has gone through three stages, namely external brain construction, technical construction and independent construction.
External brain construction: three-plate axe
At the time of the rise of the data middle office, many business leaders did not understand its value. Canada in the construction of the middle and Taiwan construction investment is large, the benefit is difficult to output, the manufacturer directly provides the construction of the middle office service is not feasible, so began to provide consulting services, for the enterprise to carry out relevant customized course training services, to help Party A continue to build and operate. Party A can purchase products and services at a single time/on an annual basis, purchase products and services with the number of license authorized nodes, and obtain OEM services in the form of original OEM. In the implementation service, the data middle office can be divided into two aspects: data application and data engineering.
Overall, the external brain construction model has a relatively mature methodology and ready-made general products, you pay money today, tomorrow will deploy you well, bring you a full set of technical personnel over, but also give you a lot of customer cases for your reference, for Party A is simply too comfortable. However, the disadvantages of this scheme are that it cannot go deep into the business, the understanding of the industry is not comprehensive, and it is easy to have expectation gaps. If the project has a problem, it is said that the leadership support is not enough, and the first-hand project is actually a way of saying that they shirk their responsibilities, and both the leadership and the executive level must have patience. What the external brain can do, you can have what others have. What the external brain cannot do or is difficult to do is to create the core competitiveness of the enterprise.
Technical construction: fast, quasi, provincial
Compared with the external brain construction, is the technical data center construction too technical? Isn't the outer brain much friendlier? Pure technology construction is often forced by helplessness. There are too many business lines to take care of one by one, so they can only be platformized. But for real users, the senses may not be as obvious. Before, I could run with a horse-drawn carriage, and now I can run the same way with a different car, which is not as magical as advertised outside. Even some special scenes are not as fast as the carriage, because the abstraction of the business will inevitably cause the loss and neglect of some personalized needs. At that time, the business line will pick the bones in the egg and say that the middle office is not successful. Therefore, pure technical construction is often thankless. Because it is my own team, all the code, components, and data are all in control, but there are slight barriers to cross-departmental communication, and there are problems with the integration of technology and business. The cost of pure technology construction is often large, and it takes a lot of effort to form a data team alone, and these people are very expensive. This construction model may ignore the feelings of the real business side, the value of the technology may not be large enough, the technology may be weak, and the successor needs to have a deep technical accumulation. The investment is large, and the results are easily criticized.
Third, a large enterprise data middle office construction experience summary
The data middle office construction methodology, summarized as "12345", includes five aspects: strategy, guarantee, standard, system and steps.
The construction of enterprise data middle office includes the whole process of project initiation, demand analysis and maturity assessment, program review, project implementation, project acceptance and operation.
In order to achieve the strategic goal of a large enterprise, data into the lake is the necessary foundation to achieve "cost reduction and efficiency increase", and the construction of regional data lake is the only way. Based on the data lake and data middle platform, the intelligent entry into the lake, intelligent governance and sharing application of multi-source heterogeneous data are realized, and an intelligent data sharing mode is formed.
Independent construction is truly from the enterprise business itself, which can reflect the core business value. Business in the "sky flying", technology is always a solid base, adhere to independent construction is the last word. At the same time, this is a process of continuous optimization and reconstruction, spending a large cross-departmental communication cost, and carrying out technical polishing for a long time, which requires us to identify the core values, die to the end, and use time and effort to construct an indestructible commercial moat!