Organizations are collecting more and more data. But most of the insights it may contain will remain inaccessible, unless it is purposefully accessed by a data scientist or analyst. Considering the difficulty in hiring data scientists, this leaves companies in a bit of a bind. Furthermore, even companies with robust analytics teams find that it simply takes too long to get the insight they need. To remedy this situation, a new concept has emerged: augmented analytics.
According to Gartner, augmented analytics involves leveraging machine learning and artificial intelligence to expedite the processes of data preparation, insight generation and insight explanation through business intelligence (BI) platforms. In this way, it also augments the capabilities of not just the data scientist or analyst, but also the ‘citizen data scientist’—the non-analyst who seeks understanding through data—by automating many aspects of the analytics process.
Augment Analytics to Free Up Data Scientists’ Time for More Complex Issues
To begin, let’s take a look at a standard strategy, in which an analyst:
Standard Data Analytics Workflow
- Keys in on the performance indicators and related goals
- Gets acquainted with the datasets
- Prepares the inputs for analysis
- Creates user-friendly views and dashboards, and starts navigating them to identify correlations, root causes, and other findings
- Presents findings as advised measures
At the end of this process, actions based on the findings may or may not be taken, based on a few factors, not the least of which is how long it took to generate the findings.
Augmented analytics basically takes the grunt work out of all but the first step of this strategy. When this process is augmented, a professional hops on the platform with a business objective or problem and an appropriate set of KPIs, and examines the system, which then handles the rest and returns with a gameplan.
Augmented analytics offers the ability to remove repetitive and tedious tasks from data science groups’ daily activities so they can work on more difficult problems. It also promises to compress the data analytics lifecycle and empower stakeholders with efficient insights and nuggets. In short, augmented analytics involves revising the total analytics and BI workflow to remove the items that bog down analytics.
If this comes across as something that’s still a little ways off in the future, keep in mind that the three elements that Gartner has identified as being essential to augmented analytics are already a part of the standard toolkit for both business and consumer software: machine learning, natural language processing (NLP) and automation.
To be clear, machine learning is a broad term that describes algorithms that make “choices” based on probabilities rather than hard-coding, and not only permits advanced analytics, but also serves as the foundation for the other two technologies.
What Is Machine Learning?
Machine learning is a term that describes algorithms that make “choices” based on probabilities rather than hard-coding. ML makes natural language processing and automation possible.
Overcoming the Barriers to Insight
NLP translates computer code into human language, allowing humans and computers to communicate without the assistance of a programmer or developer. And it has made its way into everyday life. When we ask our phone to direct us to the grocery store, we’re basically leveraging a form of ‘augmentation’ to source data and make a calculation which is then translated into easy-to-follow directions. All the layers of complexity are ‘abstracted’ for us.
What Is Natural Language Processing (NLP)?
Natural language processing translates computer code into human language, allowing humans and computers to communicate without the assistance of a programmer or developer.
It doesn’t take too big a stretch of the imagination to see how a similar approach may be applied to answering a business question:
A person asks—in plain English—“What were sales of phone cases in June vs. July?” The platform responds with “Sales of phone cases rose by 10 percent between June 30 and July 31.” The person then asks a follow up such as, “In which areas were sales of phone cases the lowest in July?” The platform replies, “Europe.” The person might continue to play with the system, learning that a certain line of phone cases, which sold very well in other regions, experienced a higher than usual number of reports of damage in Europe.
Clearly such capabilities would be a major efficiency booster. But to thoroughly understand the value being delivered here, let’s consider what the user does not have to do.
Tasks that Augmented Analytics Could Automate
- NLP converts the question, “What were phone sales in July vs. June?” into a SQL query.
- That query might involve pulling data from multiple databases in disconnected systems, like a data warehouse, an enterprise database, or even a data lake like Hadoop.
- Without moving any data, a query is then executed simultaneously across all of these platforms to assemble a single table.
- Results from the query, based on the table it generated, are then translated back into the same human language in which the question was asked (English, Spanish, Mandarin, etc.).
This back-and-forth may happen multiple times until the person receives what they believe is actionable intelligence, at which point they may then prescribe a policy or action. In the case of our phone cases example, they might initiate an investigation of the European distributor to see what’s causing the damage. Note, however, that while with human-led analytics this process might take weeks, with the augmented system it can happen in minutes.
Consequently, such decisions can be executed on a much larger scale and with greater speed. If there is a problem with a distributor, it doesn’t have to go on for months, alienating customers. It can be solved immediately.
If the user is a data scientist, they’ve sidestepped having to sift through the company’s deep, sectioned data repositories or deal with the gatekeepers at the various departments. Just as importantly, however, the user might not even be a data scientist or analyst. It might be a business user who doesn’t understand programming, or even advanced statistics. Information that they may have previously waited for months on, at the mercy of an expert, now is at their fingertips.
Augmenting analytic processes in this way can improve your speed and scale by orders of magnitude. It doesn’t mean that you won’t need an analytics team, but it does mean that your analytics team won’t be spending its time on the more mundane questions and processes. Instead, they’ll be focused on what you’re really paying them for—developing predictive models and dealing with highly complex questions.