Big data is hot word of IT, Hadoop, HBase and other technologies related to big data grow fast too. The analyzers are busy testing the new technology and method of bid data; while many business leaders are adjusting business models to draw more energy of big data. McKinsey & Company calls big data as “the new front of business”, and thinks the influence of big data on business transformation is comparable to the internet over the past 15 years.

In order to take advantage in this trend, business leaders should learn the key steps of maximum data value mining. The business success of big data drive means not only to choose right cloud computing technology or recruit excellent data scientists, but also to build a new method centered on business, connect enterprise data and business strategies, achieve continuous improvement, and finally realize the key target concerning process, profit space, customer satisfaction and so on.
 
According to my experience in big data field, I conclude the seven steps about how to successfully implement big data project for business leaders:

一、Formulate data strateg
You need a powerful data strategy, which can connect and support your business strategy, and can peg with departmental responsibilities at the same time. Make a plan with many segmented targets, and make clear the data ability target to achieve in each step and how these abilities will be used. This is actually equal to some “consumption expectation management”, which enables all departments of enterprise to have a right anticipation of the data ability in different implementation stages, and to know their responsibilities, targets and their business achievements at the same time. Good data strategy can unify the target of you and your department under the whole business strategy.

二、Flexible design 
Big data system is large and complicated, which means it has a bad flexibility. For example, a good BI system will help enterprise to transform, but in return BI system is also required to make changes. Then it needs your system to be able to adapt to the development speed of business fast. The release cycle of some components with 12 to 18 months is acceptable, but it is better to control the cycle of the other components within 3 to 6 months. It is necessary to carefully analyze each component of big data system, in order to assure the design has enough flexibility.

三、Learn delay  
Delay is the challenge of traditional BI system, and big data has further enlarged this issue. Big data solving plan normally uses batch processing structure, and puts reducing delay work backwards. You have to make changes, and limitedly consider the delay issue from the first beginning. Analyze the requirements of the several key usage scenarios of big data system on delay, and take them and business drive into consideration together. Make sure match proper delay for each need, and let these needs to decide your design! Some low delay needs can even need you to meet by bypassing big data system temporarily.

四、Invest on data quality and metadata 
Data quality is the target that all systems are fighting for, so is big data system. Furthermore, big data system need more automation and advanced plans. You have to make sure from the first beginning that data quality is treated as the basic important work, and can obtain enough data source and management support. Secondly, build multiple lines, from main data management (like creating customer account), data collecting (customer interaction record), to metadata (data organization classification for future report and analysis). Thirdly, realize automation of the two processes of data quality assessment and improvement process and data quality measurement and report. Provide your data quality team with necessary workflow tools and other tools able to carry out large scale data analysis.

五、Be good at establishing prototype 
The data amount of most big data systems is large, it is not necessary to call all data in testing phase, generally, you can establish small scale prototype to carry out testing, especially when you establish complex data integration, on-line computing or user interface, establishing prototype can not only enable a testing with a smaller price, but also makes it possible to share it with your users, who can provide valuable advice in surprise.
六、Be good at sampling 

If operated properly, data sampling can save much time for you. There are three points for proper sampling: firstly, establish standard sampling data set, and update regularly, which can help your analyzers save a lot of time to answer questions of salesmen. Secondly, make sure there is at least one master (like a statistician master), who can assure the correct data sampling and reasonable using of sampling results. Finally, it is necessary to let decision makers learn the value and limits of data sampling, so they can make decisions based on sampling data. Correct data sampling can not only improve productivity, but also provide business value without discount.
七、Often get feedback