Successful BI Implementations
Read this paper for an overview and examples of how big data is used in specific areas, such as supply chain management, risk management, and logistics of business in industry. One of the biggest issues for analysts with big data is knowing how to separate the valuable data from that which does not help answer their requirements. Consider the times, even in school, when you cannot find the right information. Sometimes, narrowing search terms can be difficult if you are unfamiliar with the topic. Sometimes, people describe intelligence as 'connecting the dots', but it is rarely simple, like a 'paint-by-numbers' art project. The dots are not just lying around waiting to be connected. More appropriately, it has been described as filtering out the right radio signals from the fray in a huge city. You have to be carefully tuned to your requirements, as these guidelines keep you on track to finding the right data to answer the questions you need to focus on rather than following rabbit holes and finding yourself in the weeds, awash in signals. What are some examples where you have had to make decisions and were concerned about the quality of the data you used to make those decisions? What did you do to 'connect the dots'?
Conclusions
Hurdles
Despite of all the big data benefits for businesses, applying big data has not been accepted at the management level of many companies yet; this may be because the cost of using big data requires a high initial investment. Management support has a crucial role in the successful implementation of a big data analysis system. An initial cost-and-benefit assessment of big data in terms of how it will be utilized long-term is a very difficult task. It is not easy to determine applying big data analysis is beneficial for businesses that make fewer than a certain number of transactions every day. In some companies, such as Amazon, Walmart, Google, etc., traditional systems cannot be used to analyze the data because of how enormous the volume is. However, in some other companies, the V characteristics of big data are more questionable, and there is no rule of thumb to help tell the manager that the available data is suitable for establishing a big data analytics framework.
One of the problems with big data applications is knowing how, where, and by which means to collect useful data. Another issue has to do with inadvertently separating valuable information out of the available data. The analyst should know the information that he/she wants to exclude from the data; additionally, the available data should be able to answer the analyst's questions. Yet another problem regards finding the methods by which one can provide the most accurate answer while still using a reasonable amount of time and financial cost. There is an increasing demand for employees who are qualified to analyze big data as companies respond to the rapid pace of technology developments. It is also a challenge to create trust between data analysts and the managers. Most systems originally resist change, so it is vital to have the higher-level managers' support in order to use the results of data analysis to change a system.
A computers' Central Processing Unit (CPU) could be an obstacle for big data analysis because of the underdeveloped capability of traditional computers to store and effectively process a big data set. On the other hand, not all of the available raw data is complete and consistent; therefore, effective cleaning and integration methods are required to make the dataset ready for analysis.
The 5th "V" added to the definition of big data refers to the value that can be obtained from a big data analysis. Unfortunately, whenever value comes in, hacking can start to crop up as well. Information security can be a hurdle to applying big data analysis in companies. A huge data volume increases the probability of having confidential and valuable information in the system, and this may increase data vulnerability and the chance of cybercriminals.
Another issue can be selecting the appropriate type of decision-making data for most of a system. More explicitly, not all of the available data in a system is used for making each and every decision. These decisions are based on the knowledge and experience of the data analyst that determine the part of a dataset that should be used. Moreover, it is an unfortunate fact that the available big data may not necessarily be created by the target population. For example, there is a huge volume of information on Twitter, but not everyone in a community will have Twitter accounts. Thus, there is a part of community which creates a lot of data, while the other part is not involved in creating any of the available information in a dataset. This fact continuously emphasizes statistical uncertainties such as biased statistics.
Applying the results of a non-real-time data analysis can lead to a significant difference between the analysis for both the historical data and the real-time data; for example, the initial assumption of the forecasting analysis that "the future follows the past" would not be true. As an additional example, when the data shows the proficiency of a transportation path in terms of cost and time (many companies have access to this data), other companies may start to use this same path. By increasing the demand for the mentioned path, it may lose its attractiveness both in terms of cost and time.
Another old and common challenge is sharing data between the echelons of a supply chain, or even various departments of a plant. It is challenging to ensure that all the stakeholders who share data receive some benefits from this cooperation. Moreover, it is vital to have a supportive information technology department which provides both the hardware and software requirements for working with big data. Data analysts who can work with big data should be hired, or knowledgeable instructors should be easy to contact so that they can educate the big data analysts for the plant.
Advantages
Big data analytics can help companies better understand their business needs. Many companies plan their business growth model according to a demand boom, which may be the result of business growth in general. However, they should consider that a change in market growth may leave them with several empty warehouses and idle manufacturing plants. Big data enables these firms to predict the market direction and plan development strategies based on this analytical information.
Using big data enables companies to simulate a digital model of an entire manufacturing process. Collected data from the customers can be used to improve the marketing and sales processes. Clustering the customers into various groups, providing service according to the needs of each group, and using customers' data to target where/when to advertise these new products are the other applications of big data in marketing.
The importance of big data analytics in improving customer loyalty is not negligible. Most customers would be loyal to the company which provides them a high-quality service the first time. Companies can use big data analysis to predict customers' needs and satisfy those needs in order to make them loyal customers.
Optimizing activities which are out of the normal boundaries of the company - such as supplier selection or technological adjustments - require a high level of information-sharing between the stakeholders in a supply chain. Sharing information has always been an obstacle in supply chains, but big data technological developments can help simplify and speed up this process.
From a logistical point of view, big data can be applied in order to forecast delivery times or optimize delivery routing by using traffic, weather, and drivers' information. Another possible application is to use the products' information in making inventory management and sales decisions.
Directions for future studies
Both the literature and a review of companies' experience in the area of using big data analytics in manufacturing shows that the number of applied case studies are more than the number of theoretical publications. It is likely that researchers will develop more novel applications for big data in manufacturing systems, such as developing methods that can obtain high-quality solutions using less time and money.
Big data has been widely used for predictive studies in the literature, but there are not many prediction error measurement studies in big data. More precisely, beyond simply the quality of the input data, the accuracy of big data analysis is significantly affected by the quality of the model used to analyze the data. We still have a way to go regarding developing measures which can determine the accuracy of a big data analysis method.
The current studies regarding big data applications in supply chain management are mostly theoretical and conceptual, and there is a noticeable shortage of studies of analytical models. Moreover, the existing analytical models mostly study big data applications in modeling sustainability. Therefore, there is still a gap in the application of big data regarding optimizing operations (such as logistics and procurement) in a supply chain.
There are some study directions in big data that can significantly improve the performance of logistics systems: 1- Developing an efficient collaboration among all the decision makers, transporters, retailers, and door-to-door delivery service providers; and 2- Applying cloud-based services in smart transportation systems and integrating them in an online planning framework in order to provide a connection between vehicles, traffic managers, and the final customers.
Finally, a good research topic to follow would be to find out if analyzing a sample of a big data set could give us the same quality results as the basic big data analytics. Design of experiment methods can be used for categorizing big data into different groups in order to save time regarding analysis.