"Reprint" 6 tips for good big data

http://www.36dsj.com/archives/40815

In this article, three big data operators from different companies share their experience of using big data. The three are Andy Hill, CEO of Luzzi of Viacom (Viacom), Olly Downs of Globys Company and Dunnhumby, a well-known market consultant.

Tip One: aim to be clear

Even if a company has more data, it does not necessarily mean that it will succeed in business. Only if you really know how to use big data and understand what the company can achieve with big data, the company will ultimately be able to truly succeed. In the development process of the company often will also face a lot of choices, but also only the goal set clear, to narrow the scope of focus on the development of the choice. Enterprises should always keep a clear mind and move towards their own goals, which will help the company to carry on the long-term good operation.

However, Luzzi also said that sometimes, the use of too sophisticated data analysis tools can often lead to a lot of problems, but if we can analyze a large amount of data to get the final result, then do not doubt, you do, at least the direction is certainly right.

Secret Two: to distinguish between "forest" and "tree"

Now, businesses can do things that they have not been able to do before. For many companies, more data is available for analysis, and tools and methods that can be used to analyze data are more advanced and convenient than ever before. Companies are fully capable of analyzing and processing the vast amounts of data they collect, which may be a good thing for businesses, but sometimes the data is too fragmented.

' Many companies now tend to collect high-precision data, because the more accurate the data, the better it is for analyzing audiences and making it easier for companies to adjust their strategies and products, ' said Olly down, Globys company. However, companies often need to spend a lot of time to deal with a large number of data, the results may not be satisfactory, so, in the analysis of data processing, sometimes there is no need to stick to a "tree" to grow into what kind of, and should pay attention to this piece of "forest", to understand the small take big.

tip Three: Do a good job of team coordination

In the world of big data, the most valuable and useful data are often scarce. Finding truly valuable data is as difficult as looking for a needle in a haystack. Therefore, in order to find these valuable data, enterprises should work together in a concerted effort to constantly maintain effective communication and collaboration.

For example, in order to better utilize the data to analyze the actual operation of the company, data experts should understand the strategic goals set by the company's decision-makers for the company. In turn, corporate decision makers should also know what the results of the analysis the company's data team will ultimately bring to the company.

With big data as a tool, Luzzi says, he can build models to help companies make business decisions. As the company's decision makers understand the overall operation and business environment of the company, when the decision maker sees the results of the analysis, it is certain to see some places that he cannot see. But at the same time, policymakers will not know what method he is using to produce the data and results.

Olly Downs also said that the company's data team and departments and management should maintain good communication and communication, so that the company can operate well and efficiently, effective coordination needs to be achieved through effective communication. A business intelligence team established a model for predicting the company's churn rate, and the operations team thought the model was "interesting" because of the lack of effective communication, but the company considered the model meaningless.

"If your company hires a data-research group that claims to have built an effective model, the company's other departments think the model is ineffective because there is a lack of communication between them." "Downs said.

tip four: Using machines instead of artificial

Machine learning refers to the computer simulation or realization of human learning behavior, in order to acquire new knowledge or skills, so as to improve their own functions. Machine learning is faster and learns larger than manual learning, and a company can discover new problems faster by machine learning.

For example, in order to study the consumption behavior patterns of specific consumers, enterprises can study how to conduct research and analysis on individual consumers, choose different models according to different consumers, and then follow up on consumer behavior. The Company believes that the analysis time required for a single consumer is 10 minutes, at this rate, the company has a total of 2.6 million consumers need to investigate the tracking, need 416,000 analysts, work 10 hours a day, observe a single consumer 8 times a day to produce results. Obviously, it takes a lot of time and effort if you really pass a manual analysis.

Therefore, downs that, if the enterprise has a large number of data to be analyzed and processed, the best way is to let the machine instead of manual, machine learning speed, can be in a short period of time to analyze a large number of data, so that the required analysis time will be greatly shortened. Costs can also be significantly reduced compared to manual analysis.

tip five: Treat data with care

Sometimes, the enterprise is unable to obtain the data, also cannot use the data to solve the problem. Even if the company gets some data, it is often unclear whether the data will ultimately solve their problems.

At this point, the Luzzi of the Viacom Group suggests that it is better to ask the data team for advice on whether a data is valid and whether it can help the company solve the problem.

Dunnhumby Company will make statistics and analysis of consumer data, so that the company can understand what data is useful, and how much value of the data. If the company can collect valuable data, the relevant problems encountered in the actual operation of the company can be effectively resolved. Andy Hill, CEO of Dunnhumby, a well-known market consultant, said that companies should not only understand what the data collected could solve, but also what problems could not be solved by the data. If there are some problems that cannot be solved, the company will need to continue to collect data from other dimensions to supplement it.

Sometimes, important data can be ignored. For example, when downs a traffic model for his former club, it is widely accepted that weather is the most important factor in predicting traffic conditions. Later studies showed that the most influential area of traffic was the school hours of the local schools. When the students are out of school, traffic jams are particularly serious.

' From the outset, we didn't foresee the conclusion, so we should be careful with the data and the data will tell you exactly what you want, ' downs said. Sometimes, the data can tell you that it will surprise you.

tip Six: To avoid the wrong conclusion

Because of the human subjective factors and irrelevant data interference, sometimes the conclusion is often wrong.

"Don't let irrelevant data affect the entire outcome, a significant portion of the data is not important, and these unrelated ' trees ' often do not represent the entire ' forest '. "If the wrong data is used, the conclusions are often wrong," Luzzi said. ”

Errors in data selection can affect how people deal with the problem and how people perceive the data and results. Bad data choices can affect the company making relevant decisions.

Andy Hill, CEO of Dunnhumby, said: "To eliminate data errors, you need to pinpoint specific target populations that are often able to accurately answer the questions you need to solve." ”

"Reprint" 6 tips for good big data