Augmented analytics involves using machine learning and natural language processing (NLP) to assist with data prep, analysis and reporting. In short, it’s using advanced technology to eliminate the major barriers between humans and the ability to quickly and continually get answers from data. As analyst and journalist Bernard Marr put it, “Without data scientists on staff or available to interpret data and turn the intel into solid business activity, the benefits of data could remain unlocked[sic].” The goal of Augmented Analytics is to change this.
Maybe you have heard the term Staff Augmentation - where external resources are used to fill skill, knowledge or capacity gaps. Augmented Analytics is similar, but technology is used to make your people more efficient and self-sufficient for data analysis.
Augmented Analytics solutions may have a wide variety of features, but theGartner defined three components fundamental key components are:
Machine Learning
Natural Language Generation (NLG) & Natural Language Processing (NLP)
Automation
The three fundamental key components of Augmented Analytics are applied at each step in the end to end analytics workflow to make it possible for a non-technical business user or analyst to perform analysis quickly when they need to - not only when someone is available to work on their request. With Augmented Analytics the analytics workflow is transformed from a very manual and laborious sequential workflow plagued by delays across many teams using many tools to a fast and efficient workflow that can make non-technical business experts self sufficient.
Machine Learning has a foundational role in Augmented Analytics
Machine learning essentially allows algorithms to make ‘decisions’ based on data and probabilities, rather than being explicitly told to take certain actions through programming. In addition to allowing advanced techniques such as predictive analysis, machine learning serves as the basis for many forms of automation and natural language technologies that allow humans and computers to interact with human language.
NLP and NLG allows humans to interact more directly with data, sans gatekeepers
Natural language generation (NLG) translates computer code into human language. So instead of getting a set of numbers the user has to interpret, they might get something like “Sales of men’s shoes in Canada declined by 10 percent in 2020.”
Natural language processing (NLP) is basically the reverse of NLG, and allows human language to be converted into computer code. This allows non-technical users to gain actionable insights from data. Just as important, augmented analytics should involve allowing human beings to communicate with computers in human language.
Most of us are quite familiar with these technologies in our daily interactions with applications like Alexa or Cortana. However, the application of these to the data analytics process is surprisingly recent.
With NLP/NLG, rather than building and executing a SQL query, a user might ask an augmented data analytics platform a question like, “How did sales of men’s shoes in Canada in 2020 compare to 2019?”
This ‘augments’ analytics by providing the ability for a person to have a conversation with a machine without the aid of someone to translate into business accessible and relevant language. ‘Conversation’ is the key operator here--it’s facilitating an exchange of information, where the user asks a question, gets an answer, which then prompts another question, all in human language.
So going back to our example, the user might ask, “What were sales of men’s shoes in 2020 compared to 2019?” When the augmented analytics solution responds with “Sales of men’s shoes declined by 30 percent between 2019 and 2020,” the user might ask a follow up like “In which region were sales in category A the lowest in 2020?” to which the platform might respond, “Canada”.
Automating Insights
Many of the questions a business user might ask, if not most, are not one-off queries. Rather, they’re things that a department will want to know repeatedly. Continuing with our example, as sales in Canada saw a sharp decline, the user might decide that they need to have an ongoing report that shows the month-over-month change in sales of men’s shoes in Canada, broken down by more granular details, like shoe product category, or province.
Augmented analytics would allow this kind of insight to be automated in several ways. For example, the user might request a report that is generated on a regular interval, such as weekly or monthly. Or, they might want such a report to be triggered when there is a significant change in direction of sales in that region. Or, such a report could be expanded to be triggered when any sales region or sales category experiences a significant deviation.
Additionally, the augmented analytics system should allow a way to see how users have interacted with data in the past. If one person asked the question, there’s a good chance that someone else may want to ask the same question. For example, imagine that the person who originally asked the question about 2020 shoe sales quit or was promoted, and left without a comprehensive handoff to their replacement. The new person in their position might have to ask the same question. The ability to, in natural language, type in that query and immediately gain access to all the data and the process that was involved in previously answering the question would be very helpful.
This process can save tremendous time in terms of requesting analytics reports. Furthermore, removing the reliance on due diligence by human beings may allow significant events to be recognized, where they might otherwise go overlooked (such as an unexpected spike in ladies’ shoes in Massachusetts). Such information could be passed on to the marketing department, who could then determine whether this was an anomaly, or part of some recurring pattern that might warrant a special advertising campaign.
Conclusion
Augmented Analytics promises to improve the ability of organizations to derive benefits from data by:
Lifting the burden of more mundane analytics from data science teams, allowing them to focus on more complex problems or developing innovative solutions.
Shortening the data analytics lifecycle by quickly identifying significant events, points of concern and trends
Giving decisionmakers quicker access to more information
For data science teams and business departments alike, much of this boils down to one word: scale. Data scientists/analysts are human beings, and as such can handle a limited number of tasks. Augmented analytics augments their efforts, allowing them to forego more mundane aspects of their jobs such as repeatedly answering questions that aren’t necessarily very complex, yet are still too complex for non-analysts. It also enables them to forgo certain repetitive aspects of data prep. On the business side, it allows departments to more quickly gather information necessary to make data-based decisions, essentially eliminating the need to go through a gatekeeper.
As such, it is an indispensable part of what has been referred to as ‘data democratization’, or the ability for the entire organization to unlock insights from data.