Beyond the Dashboard: How AI Changes the Way We Measure Business
The dashboard is nearly 20 years old, and it’s time for an overhaul. As one industry observer said, “Using a dashboard is like viewing your business through a straw.”
Although dashboards will never disappear, they will be radically transformed by artificial intelligence (AI). Rather than dispense aggregated information about a handful of top-level indicators, modern dashboards will surface relevant insights about those metrics from millions of records in real-time. They will also automatically issue alerts, generate forecasts, analyze root causes, recommend fixes, prescribe next steps, and suggest ways to optimize processes. Moreover, dashboards will become untethered—users will speak, dashboards will reply, and users will get instant answers, wherever they are, via their mobile phones. In short, dashboards will become more intelligent, predictive, timely, and conversational.
The Yin and Yang of Dashboards
Dashboards are the perfect interface for managing a business: they blend both strategic and operational perspectives in a single, compact display. Dashboards resonate deeply with business users, and I don’t see how AI or any other technology would change this.
Strategically, dashboards show business users how their performance compares to plan and their peers across key indicators of success aligned with corporate goals. This is a critical strategic activity: every company needs to distill its strategy into key performance indicators (KPIs) tailored to every individual in the organization. This is why dashboards will never disappear. It’s also why dashboards (and their scorecard cousins) are so challenging to construct. It is not easy to translate and personalize strategy into metrics that everyone understands and the business can monitor.
Operationally, traditional dashboards have many deficiencies. One design principle for a dashboard (which I promulgated in my 2005 book, “Performance Dashboards: Measuring, Managing, and Monitoring Your Business) is “three clicks to any data.” This means users should be able to click three times in a dashboard to get to the root cause of any top-level KPI. A well-designed dashboard is a moderately sized data sandbox that consists of about 10 key performance indicators, 20 dimensions, and several hierarchies. This results in roughly 1,000 metrics.
In theory, with well-designed drill-down paths and simple navigation, business users can find the root cause of any anomaly displayed in a dashboard. In reality, executives click once or twice before they call an analyst, while managers may click two to four times before doing the same. More than likely, the dashboard, with its limited data and prescribed drill paths, won’t reveal root causes. In fact, most dashboards, because they aggregate data at the top level, are more likely to hide anomalies than expose them. Consequently, most dashboards users remain oblivious to the real action happening under the hood and in the trenches of their organizations until it's too late.
AI to the Rescue
What organizations need is a way to analyze every single combination of metric, dimension, and hierarchy for every KPI they want to monitor, not just a select few. However, a single KPI, such as revenue or profit, could spawn hundreds of thousands of metrics when you consider all the metrics, dimensions, and hierarchies that comprise the KPI. (See figure 1.) Imagine the horsepower and time required to drive such computation.
Figure 1. Deep KPI Analytics
Time-Series Analytics. Although it is great to do a point-in-time analysis of these metrics against a target and prior period performance, it is better to track them every day, hour, or minute and develop a time-series performance model. Such a model would make it easy to identify anomalies and trends, like seasonality, as well as forecast future performance. Anodot offers a cloud-based service that does exactly this. (See figure 2.)
Figure 2. Time-Series Analytics
Courtesy Anodot
Many BI companies see the potential of AI and have jumped on the bandwagon. Most today generate point-and-click automated insights that surface significant trends, anomalies, and clusters in the day, usually for a highly constrained data set, such as a chart or dashboard. The trick is to do this at scale and in real time. Most BI vendors don’t have the processing power to do that, let alone run it continuously in the background for multiple KPIs simultaneously.
With automated insights, the dashboard becomes a jumping off point for obtaining deep insights about business processes. These insights might pop up above or below a KPI, or upon a click; or they might be encoded in text via a natural language generation tool that displays or speaks a deep analysis of the dashboard KPIs. (See a roundup of “AI for BI” capabilities in our latest report.)
Figure 3 shows a predictive dashboard from FinancialForce, which sells customer-centric ERP software to Salesforce.com customers. FinancialForce applies Salesforce's Einstein AI engine to sales and financial data to generate rich, action-oriented views of customers. Its dashboards display color-coded indicators of customer health, and with a single click, an analysis of under-performing health indicators along with recommendations for improvement.
Figure 3. Predictive Dashboard from Financial Force
Deep Insights
ThoughtSpot. One vendor that is at the forefront of AI for BI is ThoughtSpot, a relative BI newcomer founded in 2012 that brought Google search technology to the BI world. Its relational search engine enables users to type queries into a search box and generate charts and tables, which they can pin to a dashboard in one click and share with others. ThoughtSpot is the clear BI leader in natural language queries.
ThoughtSpot began rolling out AI capabilities last year, well ahead of the pack. But it’s market timing is not what’s interesting; rather it’s the scale and performance of its automated insights. To generate the performance required to support relational search, ThoughtSpot built its own massively parallel, in-memory, columnar calculation engine, called Falcon. This turbo-charged engine fuels SpotIQ, ThoughtSpot’s AI-powered insight engine. (See figure 4.)
Figure 4. Automated Insights from ThoughtSpot
In its 10-minute “AI for BI” demo on Eckerson Group’s YouTube channel, ThoughtSpot generated 23 “insights” by analyzing 16.8 million records in 15.8 seconds. Using regression and classification algorithms, the tool performed 72,000+ drills and pared down 22,000 insights into 23 relevant ones based on its search index and relevancy ratings. It displays the insights in human-readable form using natural language generation. (See figure 5.)
Figure 5. Automated Mining
Self-Driving Analytics. At its annual conference in November, ThoughtSpot announced the next step in its AI march. Called “Self-Driving Analytics”, this yet-to-be-released feature stitches together SpotIQ snapshots to create a time-series analytic. For example, Self-Driving Data might take daily snapshots of “sales by store and region” for several months. Once it learns the normal distribution of data, it can then send alerts when it detects values outside the normal range. Ideally, these show up as natural language alerts embedded in a dashboard or chart.
Anodot, a startup focused on anomaly detection, already supports time-series analytics at scale. It offers a cloud-based, multi-tenant service that currently supports 75+ customers, including Microsoft, Tesla, Lyft, and Upworks. (See our podcast interview with Rich Galan of the Rubicon Project, an Anodot customer.) Today, Anodot uses machine learning to detect anomalies and send alerts. In the future, it plans to use algorithms to identify the causes of anomalies, recommend fixes, forecast performance, and suggest ways to optimize processes. If it fulfills this vision, Anodot will enable companies to automate operational processes in high-volume, non-volatile environments.
The Future
Although the idea of autonomous decisions may conjure up images of faceless robots replacing humans—the reality is far different. AI will augment humans, making them more productive and allowing them to focus on value-added activities, not mindless, repetitive work. Besides, humans will need set up, configure, and tune algorithms, handle exceptions and adjust for data and model drift.
And people will still need their dashboards, including operational workers. After all, they will still need to track their performance against corporate goals and plans. But dashboards of the future will be used for more than just a KPI-display vehicle. Rather, they will serve as real-time collection points for relevant, timely insights and recommendations. They will collect, consolidate, and display AI-generated alerts and insights so workers can better execute and optimize business processes and strategy.
With AI-powered dashboards, workers will no longer look at their business through a straw. Next-generation dashboards will be as powerful for generating business insights as the Hubble telescope is for finding remote galaxies, quasars, and black holes. Let the journey begin!