Augmented data reflected in glasses

May 22, 2020

Augmented Analytics: What is It and Why is It Important for Utilities Now?

Barry Ellis

Barry Ellis

Managing Director, Data Science and Analytics

At this time of great uncertainty, analytic capabilities for utilities have never been more important. The situational awareness required to understand the unforeseen unemployment and impact to short-term revenue, emergency focus on employee safety, risk of supply chain disruptions, and projections of COVID-19 all have companies scrambling to incorporate new data into revenue and cost projections.This puts immense time pressure on analysts to quickly explain what is going on, create some sort of insight on the magnitude of changes in certain KPIs and adjust forecasts. These analysts and decision makers have no time for non-value-add activities like data extraction, data preparation and cleansing, and manual dashboard building and/or waiting for a data warehouse to be populated. They need answers NOW!

The good news is that many analytical and data preparation tools are incorporating machine learning into “smart” predictive features that “augment” existing capabilities in order to assist analysts in finding insights more quickly than ever before.

What is augmented analytics really?

Here, I’ll describe some of the most compelling features and capabilities of analytics augmentation.

Automatic predictive model creation to identify red flags

Starting with a tabular set of data, rows and columns, most data sets include at least one important numerical fact like amount or quantity, at least one importing time characteristic, and the remainder is descriptive information about the business event being captured. An automatic predictive model should be able to generate a predictive forecast of future business events. It should also identify the data that influence or have a strong correlation to a numerical fact’s movement up or down over time and those that do not.

Ideally, the tool should automatically perform a comparison of actual data against the predictive model and generate a list of outliers to investigate further. The outliers might highlight process or control issues that need to be secured or data capture/quality problems that require attention. Or they could be indicative that the predictive model is not that great, and a data scientist needs to be engaged to apply more robust data science. The generation of a list of outlier transactions, possibly “needles in the haystack,” saves an enormous amount of time that might typically be spent manually searching with drill-down ad-hoc analysis.

Natural language generation of insights to speed understanding

Modern dashboards should include the ability for a business analyst to provide some explanation of point of interest, concern, or opportunity.  But with a predictive model in hand, it should be able to highlight statistically important segments and generate language descriptions of these insights for critical consideration of decision makers. This capability should be enabled for execution on the fly so as filters are applied, the generated insights can be refactored.  So, alongside a bar chart graph is a generated insight that points out something like “40 percent of revenue lost last quarter was located in the Northeast Region.”  

Additionally, users should be able to ask questions in a natural way like, “Show me revenue last quarter” and the tool should find the appropriate data model, determine the result and select the most suitable visualization for that result. In some tools, this feature is enabled via speech recognition and results displayed on a mobile device.

Why augmented analytics is important for utilities

Fundamental changes in data interaction, speeding the process to create actionable insights

For most utilities, the classic information delivery methods and processes are well established.  I would describe this as a rich set of canned reports run on a periodic basis.  When a variance is evident on paper, then a question is raised to an analyst or accountant to explain what is going on.  That individual then starts a research project which is presented sometime later in a static fashion back to the requestor. This either becomes another stand-alone report or a new KPI on a dashboard, and over time, the whole environment becomes hard to use.

When times are stable, this information delivery and consumption model to support actions over time is adequate. These are not those times! Business analysts need access to real-time information as well as the ability to generate interesting insights and proactively raise awareness of important changes that are happening—or are about to happen—so they can address them quickly. That is why augmenting analytics with automated predictive models and natural language generated insights is so important right now. This is a much-needed step change improvement in the manner and speed in which organizations can create actionable insights.

Reducing bias from the analysis

Human beings have a natural tendency to look for information that confirms their beliefs or perspective of the world. This is a mental model of the way utilities work in various business processes. Others seek information that will please the recipient or confirm a preconceived notion of what is happen.

The great thing about a data-driven insight is that it borne of the data and often reveals some surprising thing you would not be searching for but must consider, understand and perhaps explain. This provides an independent point of view of performance that is completely absent of bias. While the insight may require an analyst to explain the relevance, put it into context or add emphasis based on their own deep understanding of the context, the time required to find this machine-generated insight was greatly reduced.

Future Predictions

Per Gartner, “By 2021, augmented analytics will be a dominant driver of new purchases of analytics and business intelligence, as well as of data science and machine learning platforms, and of embedded analytics.” *

Those of you paying attention have noticed that most BI vendors have enabled predictive algorithms to support the business analyst with data-driven observations about the data being analyzed. SAP Analytics Cloud, Tableau, PowerBI and Looker have some these features in one form or another but few utilities we speak to have adapted to interacting with data in this way.

In Summary

For some, this time of uncertainty has provided an opportunity for reflection and clarity. For others, it has been a blur of chaotic existence. For most of us it has been both, but we all have had to adapt quickly. Perhaps one silver lining is that this pandemic has shown us how to be resilient, react quickly to changes in our lives, and focus on the things that are important. 

As utilities begin to emerge from crisis and transition into resilience and recovery, there is no better time to optimize your reporting, analytics and insights to take advantage of augmented BI. After this time of quickly reacting and re-prioritizing, augmented BI can help provide the clarity you need to move forward.

If you would like to have a discussion about this topic or perhaps see a demonstration, please reach out to us at or connect with me via LinkedIn.

*Gartner, “Worlds Collide as Augmented Analytics Draws Analytics, BI and Data Science Together”, Carlie Idoine, 10 March 2020