Augmented Analytics: the Future of Data Analytics

Picture this: Your manager sends you the sales data for the last 12 months and asks you to prepare a report on the sales trend during this period. He also wants you to analyze why sales dipped or rose in a particular period. If you not a data analyst, chances are you find the task daunting.

You would probably wish the computer to churn out the report within minutes with minimum intervention from your end.

That is exactly what augmented analytics does for you.

In 2017, Gartner published a report which defines augmented analytics as “an approach that automates insights using machine learning and natural-language generation.”

Gartner has cited augmented analytics as “the next wave of disruption in the data and analytics market that leaders should plan to adopt.”

While this definition may be clear to data science experts, it would be sheer jargon to many. So, let us understand what augmented analytics is all about and how it is superior to traditional analytics.

Recall the above scenario. Let us use traditional analytics to solve the above problem. Traditional approach would involve the following steps:

1) Gathering data from different sources

2) Cleaning and labelling the data so that it can be used for analysis

3) Carrying out analysis

4) Gathering relevant insights

5) Communicating the insights to the concerned person (s) and devising your strategy accordingly.

These steps may sound simple, but in actual practice they are complex and too many problems creep in at each step.

First, gathering data from multiple sources is difficult and time-consuming. Take the above example where you need insights on sales figures. You may need to communicate with heads of different departments like finance, operations, administration, HR and IT to gather relevant information.

Then, to get an accurate picture of the marketplace, you may need to purchase data from one or more sources.

But the problem does not end here. All the data that you have gathered may not be in a usable format. It may have errors, it may not be properly labelled. There may be some duplicate entries.

You will need to spend considerable time in cleaning data.

As a matter of fact, this is the most time-consuming step in the entire process.

In fact, data scientists spend around 80% of their time on routine tasks such as cleaning and labelling data.

Besides, there is so much data that you may miss out on some important information while carrying out analysis and end up using only a small portion of the data for actual analysis. You may not understand all the data gathered from various sources. This would be a waste of your efforts.

In actual practice, data scientists don’t use more than 10% of the data.

Once the data has been analyzed and insights gathered, the results need to be communicated to the concerned person or department. You have to ensure that the results are in a form understandable by most.

In addition to this, analytics is an emerging field and competent data scientists are difficult to find and expensive to hire. That is why most of the small-and-medium-sized organizations cannot extract maximum value from the data available with them.

There is a dearth of data scientists in the marketplace. In 2018, the demand for data scientists in India grew by over 400% while supply grew by 19% only.

Read: India’s demand for data scientists grows over 400%

Augmented Analytics: A Viable Alternative

Augmented analytics automates the process of data analysis and minimizes, or in some cases, eliminates the need to hire data scientists.

Augmented analytics tools automatically go through a company’s data, clean it, analyze it and draw useful insights using advanced machine learning and natural-language generation.

Let us understand how augmented analytics makes things easy for you.

Augmented analytics has three main components: machine learning, natural language generation and automated learning.

 Machine Learning

Figure 2: Augmented analytics uses machine learning solutions for faster, more insightful results

Machine learning is a field of artificial intelligence based on algorithms that can learn from the data without depending on rule-based programming.

In simple words, this means you do not give explicit instructions to a machine, but program a machine with data. The machine learning algorithm finds patterns in your data and turns those patterns into instructions.

To understand this better, let us take an example.

You feed a machine with a dataset containing different images of cats and dogs, labelled as such. The machine goes through this dataset and identifies patterns between images of cats and dogs. Based on these patterns, the machine builds an algorithm that identifies (classifies) an image as cat or dog.

This algorithm can now be applied to other datasets without labels. With time, the machine is fed with more data, and the algorithm uses this data to improve its accuracy in identifying images.

Machine learning algorithms can process much faster than humans.

In real life, machine learning has been applied in areas ranging from facial recognition, speech recognition, statistical arbitrage and even medical diagnosis.

When it comes to data analysis, machine learning can build algorithms to detect patterns in your datasets and churn out actionable insights within minutes.

Read: The simplest explanation of machine learning you’ll ever read

Natural-Language Generation

Natural-Language Generation (NLG) is a process that translates the findings of a machine into words that a layman can understand.

NLG is a crucial aspect of augmented analytics-it allows even a non-technical person to understand the output of an analysis.

For example, if your system finds that the sales of product A in your company declined by 25% in the last quarter, NLG will enable the system to communicate this to you: “Sales of product A declined by 25% in the last quarter.”

Some augmented analytics platforms have NLG integrated with their search functions that allow a user to ask questions in plain English and get answers in form of visualizations. Thus, NLG has the potential to facilitate a two-way conversation between man and machine.

NLG not only makes it easy to understand the findings of an analysis, but also prompts a user to dig deeper and ask why something happened.

In the context of the above example, you may be curious to find out “Why sales of product A declined by 25%?” or “How has the sales of product A changed over the last 4 quarters?”

It is important to note that every augmented analytics platform has its limitations when it comes to answering questions involving such in-depth analysis.

Read: The Ultimate Guide to Natural Language Generation

Automating Insights

Augmented analytics combines the power of machine learning and natural language generation to automate the labor-intensive process of data analysis and communicate insights to the concerned person.

These automated insights can help you assess your company’s performance, identify growth opportunities and/or threats and thus create a holistic picture of the health of your organization.

You can use these insights to create a solid business strategy.

Figure 3: Augmented analytics helps create a holistic picture of your organization’s performance

You can go beyond routine reporting and dig deeper to find out, for instance, “Why sales dipped in quarter 3?” or “Why product A did not sell in north?”

Machine learning capabilities in augmented analytics empower the user to go beyond the insights offered by routine visualization tools.

You can discover valuable patterns, trends and correlations and take steps to gain a competitive edge.

What are the benefits of using an augmented analytics platform?

It offers deeper, more valuable insights: As discussed above, augmented analytics can go beyond the routine reporting and dig deeper to find the root cause for a phenomenon. It can study several datasets in combination and reveal hitherto unknown correlations.

For example, if as a marketing executive you are trying to assess why sales plummeted in a particular region, you may discover some variables strongly correlated with buyer behavior that you ignored in routine analysis. You can study buyer behavior in conjunction with economic variables and industry trends to create a holistic picture.

It gives faster results: Augmented analytics platforms allow you to ask questions and get answers in a jiffy.

For example, if your boss needs insights on consumer behavior, he can immediately turn to augmented analytics platform and get answers on any pertinent question without the intervention of any analyst. He does not need to wait for weeks or months for the analyst to generate the report and explain to him the results.

It helps utilize resources better: Augmented analytics automates many steps involved in data analysis. As a result, the data analyst need not spend hours in just cleaning and preparing the data. He can free himself from repetitive tasks and focus on more complex aspects of analysis.

Augmented Analytics Is Evolving

Augmented analytics is evolving slowly but steadily. Of late, several BI and analytics platforms with advanced machine learning capabilities have been developed.

For instance, Yellowfin recently released Yellowfin 8.0.1, an augmented analytics platform which is an enhanced version of Yellowfin 8.0.

Yellowfin 8.0.1 comprises Yellowfin Signals, an automated insights discovery product, and Yellowfin Stories, a data storytelling product.

Yellowfin Signals has the ability to customize insights according to the role of the user so that a user gets only relevant information. In the most recent version of the product, a plugin framework has been provided with Yellowfin Stories. This framework allows users to embed live and interactive reports from other BI and non-BI platforms like Tableau, Power BI and Qlik into a story.

Suggested Reading: Yellowfin releases version 8.0.1 with new Signal governance and Tableau, Qlik, and Power BI report integrations

Sisense Labs has developed an analytics chatbot Sisense Boto. The chatbot combines AI and BI capabilities. Sisense Boto can be used by anyone having messaging apps like Facebook Messenger, Skype or Slack. All a user needs to do is to add the bot to their contact list. The user can then drag and drop any csv file containing their data into Sisense Boto and allow it to work.

The chatbot runs advanced analysis in the background where it detects hidden patterns and anomalies and carries out in-depth analysis. The user can share the insights generated through analysis with anyone through his chat interface.

Einstein Discovery from Salesforce is another noteworthy example in this regard.

BI and analytics are growing by leaps and bounds and augmented analytics is taking it to a whole new level. We, therefore, expect to see some path-breaking innovations in this space any time soon.