Introduction: there is no need to say more about the importance of data today. All enterprises are aware of the importance of data and hope to use data to drive business development. However, many enterprise information managers still have some misunderstandings about data intelligence and data-driven, which will make the data utilization of enterprises fall into the abyss.
Trap 1: the application has not been built yet, so data utilization is not considered
When we talk with some enterprise information managers that we should consider the use of data as soon as possible and make overall planning for data, we often hear such a sentence.
“I haven’t started my business yet, and I don’t think about data utilization.”
This sentence represents a large part of enterprises’ cognition of data utilization, that is, data utilization starts with data first, and data is stored in the database after application construction, so first build the application, and then consider how to use the data after data is available in the database.
Sounds like the logic is absolutely right.
But in fact, this is the primary misunderstanding of data utilization in many enterprises: “first build applications, then consider data utilization”.
If we use this way of thinking to build, after a year, often the enterprise will immediately put forward new problems, “the data between multiple application systems is not connected, not aligned, inconsistent, and the data can not be used.”.
This misunderstanding is that we do not fully understand the two natures of data utilization
First, data is objective and does not depend on whether you build an application or not
An enterprise, as long as the business is running, even if it does not build any system, its data are generated in real time, but you do not collect it.
Data is the constituent atom of business in the digital world. Business processes and behaviors will produce all kinds of data at any time, rather than having to build and apply these data. For example, when a courier receives an express order, the data such as sender, recipient, goods category, delivery place, delivery place, means of transport type and distance have been generated, which will drive the direction of the express delivery. With or without the support of information system, it only changes whether the means of recording and transferring these data is a piece of paper or a network. These data exist objectively and will not change because of the information system itself.
We should recognize that data is the projection model of business in the digital world, it is the mirror image of business, and it is an objective existence.
As long as there is business, there will be corresponding data. The application just collects the data into the storage device through the software.
Second, the planning of data utilization should be earlier than the construction of application and process
Before we build a house, we need to do the overall design and plan out all kinds of use scenarios of a building. Only in this way can we avoid a house that cannot be entered.
Now, every enterprise realizes that data is the core asset of the enterprise, and application is the tool to collect and use these assets. In order to make full use of the data after data collection, each enterprise must complete the data utilization planning before the application and process planning.
This includes the planning and design of the enterprise’s data asset directory, the planning and storage of the enterprise’s data utilization scenarios, and the demand planning of the technical platform for processing and analyzing these data.
Data first, when the system has not been built, it has done a good job in the blueprint planning of data and completed the panorama of data distribution of each application system, so that enterprises can avoid the existence of data islands.
So, if you haven’t built an app yet, congratulations. This is the best opportunity to plan data and use blueprints. Let’s get started.
Trap 2: there is no big data, so data utilization is not considered
“We have very little data now, so we can only call it small data, so we can’t talk about data utilization”, which is also a typical misunderstanding of data utilization.
The first time I heard this is in B2B2C retail enterprises. It is true that the traditional brands, whose main channel is through distributors, often do not establish their own e-commerce system, so the behavior data of the final consumers are not available. What they have is the data of sell in, and the data of sell in is often small in quantity and has few dimensions, so its utilization value is limited.
However, what the enterprise is doing now is to establish various contacts with terminal consumers and customers through small programs and applications, so as to obtain all kinds of data. Single data is small data, with small amount and few dimensions. However, when all these points are connected together, they constitute a rich and diverse panorama of user data.
The business leader of this enterprise firmly believes that in the digital era, whoever has more data scenarios can have a stronger competitive advantage.
This example fully shows that maybe now your business model determines that you don’t have rich data, but you still need to obtain users’ and consumers’ data through various application innovation channels. What applications should be built, what data should be acquired, and how can these single point data be interconnected to form data scenario value?
This is the need to have data planning before building an application, outline a data scene map, so as to build a small and medium-sized application along this map.
Third, the use of data is to do data analysis and mining, transaction application system does not use data technology
In the past, application systems were divided into OLTP and OLAP, online transaction system and online analysis system. Therefore, we often see that the application itself is a transactional software. According to the traditional architecture, that is OLTP system, so we often do not use some OLAP technology.
However, great changes have taken place in the current situation.
In terms of the car dispatching system, according to the traditional division, this is a typical trading system, which creates orders and distributes drivers. However, if you want to support tens of thousands of orders per second, it is impossible to use manual allocation. This scheduling system needs to have the ability of real-time data analysis, and the price determination and route planning need to refer to the historical data analysis results. In this way, this typical transaction application is data driven, and its bottom and core are actually batch data analysis and real-time data processing.
In the future, all applications will be like this, that is, OLAP is supporting every decision and behavior of OLTP system, thus becoming an intelligent application.
Data technology is gradually reconstructing all traditional process applications to make them data-driven systems and become more intelligent.
Trap 4: the most important thing is algorithm, so software engineering companies can’t do data science projects
When it comes to data projects, the first thing many people think of is algorithm models. It seems that only those who do research, algorithms and artificial intelligence do data.
Therefore, there is a kind of view that the information industry is divided into algorithms and software, and only algorithms do artificial intelligence and data.
This is a typical misunderstanding, which separates algorithm from software engineering. Just like not long ago, a long-term cooperative customer denied us an opportunity with an inherent impression that “stwerk is not an AI man”. This is a misunderstanding of AI applications.
Let’s use the following graph to show the relationship between algorithm and artificial intelligence (Data Science).
The bottom layer of artificial intelligence is composed of various algorithms, but at present, the commonly used algorithms used by all people in the industry are open, and it is academic research institutions that really research and produce these algorithms.
Artificial intelligence is divided into two fields, one is the frontier research field, the other is the application field. As an enterprise engaged in industrial production and commercial operation, what it needs is the latter. The most important thing of the latter is to use software engineering ability to apply suitable algorithms to valuable scenarios, so as to enable business.
On top of algorithms, the application of artificial intelligence is more important to have sufficient high-quality data sets, and the engineering ability to develop algorithms and data into intelligent software with good user experience.
Therefore, in addition to the ability of tuning and invoking open algorithms and code, excellent AI enterprises have the ability of business innovation and software engineering.
Summary and Enlightenment
By analyzing the four pitfalls of data intelligence one by one, we can draw the following enlightenment:
1、 Data planning should give priority to the construction of business system, build a coherent and consistent data panorama, and avoid data islands between applications
2、 After building the data panorama, build one by one small application along the map to collect and fill the data, so as to build their own data assets
3、 All applications will be enabled by data technology and become data-driven intelligent applications
4、 The most important application of AI in business is the ability of scene innovation and software engineering
This article is the original content of Alibaba cloud and cannot be reproduced without permission