Big data consulting requires making sense out of petabytes of seemingly unrelated data. Your job is to assimilate data sets, including structured data stored in the enterprise database and unstructured data, such as emails, web content, or social media conversations. Using Hadoop or NoSQL to create a big data analytics framework, you distill this information into intelligence that provides insight for making business-critical decisions. However, before you can engage on any type of big data consulting engagement, you have to know the right questions to ask in order to identify and analyze the right datasets.
One of the biggest reasons cited for failure of big data projects is because organizations don’t know where to start. According to Gartner, more than 64 percent of organizations have invested or plan to invest in big data to date, but only 8 percent have deployed. Twenty-six percent say their biggest challenge is how to get value from big data, 12 percent say they have trouble defining a strategy, and 15 percent don’t even understand what big data is.
Big data consulting helps the customer identify where they can get the most benefit from big data and how to ask the right questions in order to formulate the right use case to extract needed insight. To get started, you need to determine what kind of analytics would be most valuable.
Let’s consider the three basic kinds of big data analytics – descriptive, predictive, and prescriptive:
As you might expect, descriptive analytics are used to isolate what has transpired. Descriptive analytics are used to describe the past, using past performance to predict future outcomes. Some experts say that 80 percent of business analytics are descriptive.
Social analytics, for example, are a popular form of descriptive analysis. Using descriptive analytics you can track social media posts, followers, comments, page views, check-ins pins, and a myriad of metrics. By assimilating that data using descriptive analytics, you can see patterns that will never be revealed by looking at the raw event counters. The outcome could be used to inform a new marketing program, guide product development, or shape customer relations.
Descriptive analytics are the most common type of analysis, and will reveal past performance for sales, production, shipping, or other operations. If you input the right synchronous and asynchronous data sets it should reveal insights that you could not gain using statistical analysis alone.
To make educated forecasts about the future, predictive analysis applies statistical modeling, data mining, and other techniques to study historical data to anticipate future trends. Since no algorithm can actually predict the future, predictive analytics offers possibilities based on degrees of probability.
Predictive analytics take the data you have available and fill in the blanks for best guesses. For example combining historical data from CRM, ERP, and POS systems you can identify patterns. Using those data patterns you can apply statistical models and algorithms to identify relationships between data sets. From there you can forecast purchasing patterns, likely sales trends, customer behavior, or supply chain and inventory requirements.
One of the most common uses for predictive analysis is sentiment analysis. Using plain text and other data sources you can generate a sentiment score, e.g. +1 or -1. While this is not a prediction, it is an indicator of a likely outcome as to a positive or negative response.
As the name implies, prescriptive analytics are used to identify or “prescribe” potential actions. The objective of prescriptive analytics is to quantify the impact of future decisions, instructing current decisions by identifying possible outcomes. Prescriptive analytics should not only reveal what is likely to happen, but why it should happen, providing recommendations so you can make the most of predictions.
These analytics are designed to predict multiple possible futures, assessing potential outcomes based on specific actions. Prescriptive analytics typically rely on historical and transactional data sets, real-time data feeds, and other sources and analysis is done using business rules, algorithms, machine learning, and computational modeling.
Prescriptive analytics also are extremely complex and have not yet been widely applied to business problems. However, some larger organizations are using prescriptive analytics to optimize production and manage inventory and the supply chain, delivering the right products at the right time.
These are the three basic types of analytic models you will apply in any big data project. Part of big data consulting is determining what type of analysis will deliver insight to inform the business question at hand, and then applying the right data sets and analytics to address the question. While big data is still an inexact science, when armed with an understanding of how input leads to outcome, big data consulting can guide customers to a substantial return on their big data investment.