Non-IT projects with large quality data

With ERP, MES, SPC and other systems as the representative of the application of modern IT technology for the enterprise's production, quality and operation Management has brought great help, not only that, these systems in the course of operation, but also for the enterprise accumulated another valuable wealth: data. In the wide application of large data today, how can enterprises fully tap the data implied by the important information, to help enterprises improve products, services and even management of various dimensions of quality level?

Large quality data is not just about it

In the process of investigating a large number of enterprises, it is found that many people think that the data collection, storage, acquisition, analysis and presentation of the various links is an IT problem: data collection can be measured with the computer software implementation, Oracle, SQL Specialized databases such as server provide us with strong data storage capabilities, and we can easily get the information we need from massive amounts of data through an interface program, which can help us analyze data, build models, and show analysis results with very useful icons.

However, why many IT systems have been quite sophisticated enterprises have not been able to get help from the quality of large data technology?

What data is really needed to collect and store in terms of data acquisition? Product yield data do we need to collect? What are some of the variables that we need to collect, such as pressure, temperature, weather, time, etc.? We should use the product qualified/unqualified as the standard of product quality evaluation, Or should there be a better measure? For the data to be collected, the way to effectively reduce errors and avoid errors ...

In terms of data acquisition and collation, what data should we analyze to help solve practical problems? We often inevitably have missing values, outliers, or outliers in our data, how can we identify the authenticity of the data, especially these special values, and how to properly fill, Correction or removal of data that has been prepared for subsequent analysis?

The plays are also data analysis. When we analyze a data for the first time and have no previous experience of the data, we should use that analytic method to find the secret in the data most effectively. Hypothesis testing, variance analysis, simple/generalized linear model, cluster analysis ... With so many analytical methods, how do we make the right choices based on the specific circumstances, and how do we find the best balance in lack of fit and over-fitting? How to transform the engineering problem into the data analysis problem, and how to restore the data analysis result to the actual engineering application environment? What is the best way to help others understand what we have learned from our analysis?

Obviously, this is far beyond the scope of the IT system problem. The application of large data can be our production of all aspects of life, but in terms of quality management, although statistical quality management in the United States has long been a mature method of quality improvement, the more sophisticated process requirements in the industry, the data collection and analysis requirements are higher, but on the micro-Mai consulting research, The connotation of large quality data is more extensive, the methodology is more diverse, and the value that can bring to the enterprise is much bigger.

"Large quality data" can be said to be a quantitative decision thinking, industry quality management experience, a reasonable industrial data collection plan, professional industrial analysis of the statistics (including but not limited to statistical) method of the composition of the solution! China's quality management because it has not experienced the real statistical quality management stage, if the management of fine in the process of flexible use of "large quality data" and "quantitative decision-making" ideas and methodologies, often can play a multiplier effect.

Generally, large quality data needs to be based on the characteristics of each line itself, develop targeted data collection plans, including indicators development, variable selection, data structure design, sample size and effectiveness evaluation, experimental structure design, etc. we need modern it technology to extract, clean and collate data that will help solve our problems, Sometimes we need to integrate the after-sale service, warranty and even customer satisfaction data with research and development or production data to evaluate the impact of the actual improvements we are going to have on the aftermarket and the customer. On this basis, we can explore the data to find the quality of ascension, Product design or customer loyalty has important clues, exploratory data analysis can help us to complete the work well; the important factors of key quality indexes often need to be locked after we have repeatedly tempered and analyzed the quality data. The next step is to develop a workable quality improvement program than to predict the quality level ... "Data can sometimes lie" and we have to be screened.

If the quality data (including research and development, production, after-sale, reliability, customer satisfaction, etc.) is compared to the golden sands, IT system can be considered as a packing container, but we also need to have the idea and method of how to effectively Amoy gold.