How to implement augmented analytics: 3 important caveats

How to implement augmented analytics: 3 important caveats

With augmented analytics, keep in mind that users might not be data literate. Find out what else you need to know about using augmented analytics to pull insights from big data.

Digital marketing campaign data analytics report with metrics and key performance indicators (KPI) on information dashboard for advertisement strategy on internet, business person in office

Image: iStockphoto/NicoElNino

Conversations about big data and analytics have emphasized the importance of leveraging data for the past decade. What hasn’t been discussed as often is the need to leverage the ability of people to understand data and apply this understanding to the business.

The need to democratize data usage and understanding beyond traditional dashboards and reports has been a major driver of augmented analytics, which Gartner defines as “the use of machine learning (ML) and natural language processing (NLP) to enhance data analytics, data sharing, and business intelligence.”

SEE: Cheat sheet: Data management (free PDF) (TechRepublic)

What is augmented analytics?

More about Big Data

Instead of waiting for a data scientist or an IT specialist to assemble complex data models and algorithms to query data, an end user without a formal background in data science can submit a request in a natural language like English, which a software engine translates into machine-understandable code. The code automatically creates the required analytics models to detect patterns, trends, and anomalies in the data to answer the request.

For purposes of pattern and trend recognition, machine learning is employed. ML discovers repetitive patterns—or anomalies in patterns—in the data, which in turn lead to business insights. Once a trend or pattern is discovered, a predictive software engine performs a root cause analysis to identify the most probable factors causing the trend. 

At the end of the process, the insights uncovered in this series of automated processes are converted back into a natural language such as English and delivered to the user. This enables the user to act on the information.

What are examples of augmented analytics in action?

An agricultural producer looks at historical harvest and sales trends for strawberries, which typically have an early harvest in Mexico and a later-season harvest in California. What the producer overlooks and the augmented analytics system picks up is an anomaly in the growing season trends data. The system looks further, trying to find the root cause of the anomaly, and discovers that temperatures have steadily risen due to climate change. Now the growing seasons for strawberries in Mexico and California are beginning to overlap. This creates a projected oversupply of strawberries in the market, which threatens to depress prices and squeeze profits. 

A human resources manager wants to learn why employees leave the company. She sees that in some cases  there are retirements, and in other cases employees find better opportunities. In the past, she might have been satisfied to write a report, but the augmented analytics tool she is using, which includes machine learning, also sees a pattern of employees leaving in greater numbers from the company’s Atlanta and San Diego offices. In a root cause analysis, the system deduces that there have been excessive management and organizational changes at both offices. The system concludes that the excessive change may have driven some employees to leave, prompting the HR manager to consider whether too much change is creating an unstable and uncomfortable work environment.

SEE: 10 ways data and analytics will impact businesses (TechRepublic)

What should you know when implementing augmented analytics?

The promise of augmented analytics is to eliminate longer lead times to insight for the end business. This is possible because end users can now query data in a natural language like English, and a system can then go to work with machine learning and self-developed algorithms to provide new insights. It uses data patterns that it discerns can augment what users have already asked for.

The process isn’t flawless,  but neither is the process of algorithm development and data modeling that data scientists use. Companies should consider adding augmented analytics to their data query strategies, but with some caveats.

Augmented analytics are only as good as the people who use them. Many end users are not data literate. Data literacy is, “understanding what data mean, including how to read graphs and charts appropriately, draw correct conclusions from data, and recognize when data are being used in misleading or inappropriate ways.” These skills typically aren’t asked for in the job requirements of production supervisors, customer service managers, or sales executives. 

Augmented analytics should be a carefully orchestrated addition to existing data science and analytics applications. This is because data literacy is likely to be underdeveloped in most organizations. When augmented analytics is used, data scientists and IT data analysts should be heavily engaged in the process of implementing it.

The vendor you choose for your augmented analytics is important. If the vendor doesn’t have a road map on how it’s going to further develop the product, or have a support and training system robust enough to impart data literacy and tools competence to citizen data analysts, it should probably be avoided. 

Also see

Source of Article