The First Step Is the Hardest: The Path to Advanced Analytics
There's a lot of buzz about advanced analytics. To go beyond the hype and benefit from its potential, managers and practitioners alike need to retool their thinking.
- By Troy Hiltbrand
- February 1, 2016
As people hear the buzz about advanced analytics, they get very excited about the potential of this new silver bullet and how it will revolutionize their businesses. As the conversation reaches a fever pitch, terms such as artificial intelligence, machine learning, and computer vision get added to the mix, and businesses get almost giddy with excitement over what that could mean for them. Then reality sets in and they find themselves in the position of wondering where to start in unearthing this treasure trove of insight that is hiding within their data, just waiting to be discovered and exploited to take their organizations to the next level.
To go beyond the hype of advanced analytics, managers and practitioners alike need to retool their thinking. They can get to their ultimate target of implementing higher-order analytics that will change the face of their business, but they have to first start small.
With advanced analytics, there are two major types of operations: categorization and prediction. Categorization takes an unknown instance of data and groups it with other instances within known categories that share similar characteristics. Prediction estimates a likely value based on other known values.
The challenge is to break these complex business questions down into smaller, less complex questions that can be answered utilizing these two operations. It is similar to the way a computer works. At the lowest level, a computer only understands ones and zeros. From this simplistic building block, some of the most complex machines known to the human race have been constructed that have forever changed our lives.
For a business to effectively implement advanced analytics, it must be able to formulate its business questions so they can be answered with categorization or prediction.
Categorization: A common categorization question asks, Will something or won’t something? This becomes a yes/no problem where analytics can be applied. Once the question is formed, example data with a known answer is used to develop a model. That model is then applied to data with an unknown answer to predict the most likely answer for that data. When the problem is broken down into multiple, simple yes/no questions (e.g., Will the customer buy again? Is this a fraudulent order? Is this a defect?), analytics can predict with a certain level of confidence whether the answer is yes or no.
Estimation: The second class of question deals with estimation. These questions may relate to time-series data but are more often about data that occurs in a pattern. Identifying that pattern allows the business to anticipate the next value. Estimation questions are a useful way to anticipate direction and to get a sense of what the value might be at an unknown point.
To solve a business’s most vexing challenges, advanced analytics is first about breaking down complex problems into smaller problems that can be solved using categorization or estimation.
The target is to change your mindset. Instead of getting discouraged because the big questions can’t be quickly and easily answered, break down the big problems into smaller problems that can be solved. Once those smaller problems are resolved, they come back together to solve the business-altering challenges.
Troy Hiltbrand is the chief digital officer at Kyäni where he is responsible for digital strategy and transformation. You can reach the author at email@example.com.