IT people will inevitably have bad projects in their careers. Some projects are bad when you join them, some are bad when you start from scratch.

Some are rotten from inside to outside, think the project is huge surface bright, actually jump in to find a “tar pit”; Some of it is not rotten, but the problem is obvious, a “dangerous house” that is falling down.

BI, as an important branch of IT projects, will naturally have similar problems, such as the experience of the following brother.

Li ge is only more than 30 years old, he has been mixed in the Internet, holding a try attitude applied for a well-known clothing company BI manager, pleasantly surprised to be employed. After officially taking up the post, I found that I had inherited a mess. The BI system that my predecessor had spent more than a year making seemed to meet all the needs of the leader, but in fact it was a crumbling “dangerous house”.

The system is quite complete, ERP, MES, CRM should have. Reports are also quite rich, from business personalized demand reports to the fixed use of various departments to query reports, summary reports.

Familiar with a few weeks, however, li3 ge found BI system launched more than six months, the original design of the report to browse and use a handful, even the general requirements of business analysis and monthly performance evaluation must be obtained from the BI didn’t realize these two, still need to export data from various systems business department to calculate statistics on its own.

In addition, the business often reported some differences between the business system and the BI data in the same dimension of the report, and all departments argued red-faced in the meeting, leading to a serious decline in the trust of BI data.

However, the BI team led by myself, with 7 or 8 people complaining about the bad demands raised every day, constantly taking the numbers to make reports, and the demands were arranged after a month, so the data integration, screening, classification and analysis all had to stay in each business system, and they could only rely on manual typing of hundreds of thousands of lines of SQL inefficiently.

Now Li Ge reaction, why they can so smoothly “pick up success”, the original is a pile of mess:

  • Data warehouse is built, but the underlying data are not well handled, index caliber is not on, resulting in the final report data quality in a mess…

  • The original idea of the business around the operation and performance of the analysis report planning, but the actual implementation of the business is still around the individual needs to demand, and 90% of the report needs business needs sorted out, just the report needs, now there are more than 1000 statements in the library…

  • In addition, not to mention real-time, the original morning to read the report, ETL and other related calculations, the report only a week, the following people still have to deceive the demand side, server performance is not enough, the database system memory is too small…

  • Statements are also, using an open source tool, simple will be, but when it comes to extremely complex multi-topic, complex statistics will be blind, business requirements are sometimes weird, must be completely irrelevant to the whole thing into a report, you have to implement. Tools can not implement, finally back to write SQL…

Realize that the moment to run away is too late, but from scratch to overturn the company’s entire BI planning is not easy, the Internet to do BI that set of difficult to apply, Li Ge was in an impasse.

The solution is to diagnose the problem first. Brother Li spent more than 10 days visiting and investigating various business departments, the management and the grassroots level of the business. Although he collected a lot of questions and jokes, the value and confidence of BI remained in the company. He discussed with his team mates and CIO for many times and decided to change the current implementation idea and determine the next optimization plan.

1. First, stop the technical work, reorganize the requirements, and focus on data quality management

** The first is master data governance. ** That is to say, what master data will be used in the business management process? How is this master data generated, distributed, and which dimensions are marked to form the derived master data? Then, a master data center library was separately built in BI, and the master data of the business system was extracted and stored according to classification principles, and the master data consistency verification program and master data distribution log table were developed.

** The second is to sort out indicators and establish an indicator system. ** Defines the business indicators used in each analysis process, establishes evaluation criteria and calculation methods, and presents the business management logic in a more intuitive way. The data fluctuations in the sales link can be presented intuitively. Through the presentation of indicators, the problems in which parts of the business can be traced.

** The third is standards that govern the entry points for data generation and the exit points for data values. ** Make clear the operating standards for all data input and establish the interface specifications between each system and BI. Almost all data generated in the company’s current business activities will be put into the data warehouse, and the BI system will uniformly extract and process data.

In addition, all data requirements for monthly operation analysis, performance assessment and daily management analysis submitted by all business departments were evaluated and analyzed, and corresponding data models were built. All application data were required to be taken from BI system, and data consistency and uniqueness could be guaranteed only with the standardization of entrance and exit.

After completing the above three actions, the company simply compiled the data management system, defined the master data generation, index system structure and algorithm, data input and output standards.

2. Secondly, build confidence and prioritize the rigid requirements in the analysis requirements, that is, the real high-priority requirements

** Do demand research first. Demand is so important! The previous survey was too extensive, and many problems were exposed after the launch. This time, based on the previous survey, we made supplementary indicators and analysis points and confirmed the caliber.

By means of questionnaire and interview, managers of various business departments were surveyed, and their requirements for statements and indicators of concern were collected.

For example, through the introduction of balanced Scorecard, the financial department hopes to obtain their usual analysis methods and concerns to form an analysis system. Finally, the financial department also cards out a value chain based on cash flow. In addition, in the follow-up interview process, the requirements should be refined to the granularity of number caliber, dimension and measurement level as far as possible, and confirmation should be made.

** A BI project restart requires leverage. ** This lever is to solve or even miss the rigid requirements in the user enterprise analysis requirements, that is, the real high priority requirements.

In the process of investigation, the financial department was the most active. On the one hand, there were many demands for analysis, and on the other hand, the financial department also faced the pressure of high-level questioning. And finance, as the core department of the company, is also the best teammate for IT to establish a brand. Therefore,, I carried out “in-depth cooperation” with the financial department, sorted out the analysis model, and developed a number of analytical statements such as financial operation.

In addition, to get the top, managers pay the most attention to the company’s business indicators, cockpit design is an important means. Managers can quickly read their KPI indicators and the changes of business indicators they are concerned about through the combined report of cockpit and key assessment indicators, because what each management position should be concerned about is clearly sorted out in the system.

3. Launched a mature BI reporting tool, solved the problem of manual reporting, and reorganized the reporting system

Reports are the final presentation of BI projects, and the most important concern is the efficiency of report generation and the speed of response to demands. Open source tools are free but demanding.

To understand a pile of English documents, can flexibly think of interface development or SQL processing for the needs that can not be processed, but can deal with a small number of people, after all, 7 or 8 levels of uneven, low-level repetitive work makes the loss of staff, with a large cost. The ability to respond to reporting needs needs to be attached to the tool.

There is no need to spare this link, but the commercial report sails soft FineReport resolutely. In addition, the report analysis system, cockpit, and data rights and other issues need a BI decision platform to carry, acting as the company’s business data portal.

In terms of production, finereport’s aggregation report production mode basically solves most of the complex report development, demand adjustment only need to modify the report template;

In addition, the filling process can be linked to share or mounted to the decision-making platform for processing, the basic pass manual Excel. The working efficiency of the whole operation layer has been improved a lot. All users are in the same channel, using the same data source to report, and there is no need to temporarily process some chaotic reports like in the past.

In addition, different data analysis systems have been planned around different business topics before, which can be divided from low to high in terms of level: business basic report (fill in and query) — business analysis report (for business decision-making layer) — operation management decision cockpit.

Compressed more than 1000 reports to more than 200 report templates, all developed with FineReport, and then mounted to FineReport’s data decision-making platform, forming the company’s data portal, and with the platform itself to provide a series of functions such as data reporting, process approval, authority management, formed a complete set of systems.

The final results

So far, the company’s business analysis reports and KPI assessment data values are provided by BI. The daily usage frequency of BI system is second only to that of the core business system, and the daily browsing frequency of BI system on the platform is more than 3,000 (more than 1,000 in the company). The management concept of the company has also undergone profound changes. It is no longer expressed in qualitative language from top to bottom, but has formed a habit of speaking with data.