What you need to know:

• Commbank says dashboards are the past. While useful for analysts, they don’t help make quick business decisions. The bank is moving towards AI-powered natural language processing for faster insights.
• Fixed dashboards can help track KPIs, Chief Analytics Officer Dr Andrew McMullen says, but non-analysts need to ask adhoc questions in real time.
• Lionel Kho, Vice President of Data and Analytics at Dentsu’s Merkle, says marketers and agencies should look at intuitive search engines to find answers to complex data questions.
• Kho says to outperform competitors, businesses need to outperform them in data analysis.
• Leon Bombotas, founder of Incites.ai, says dashboards are slowing growth, and AI generated insights – a news feed of analysis – are paying dividends.

AI-robot

Australia’s biggest bank says it is using artificial intelligence (AI) and machine learning (ML) to make faster, fact-based business decisions and detect systemic customer issues, moving away from “static” dashboards that require analysts to interpret.

Dr Andrew McMullen, Commonwealth Bank of Australia’s Chief Analytics Officer, says the analytics industry will increasingly move towards natural language processing (NLP) and AI for adhoc analysis, with dashboards relegated to measuring targets and operational data.

While there is skills shortage in the data engineering industry, AI and ML are “democratising” advanced analytics, meaning science skills are less necessary to work with data.

“To support fact-based decision making we require fast, adhoc insights," McMullen says. "Traditionally this has meant analysts generated a large number of fixed dashboards, with the expectation that others monitor and decipher the information provided." 

“What’s more, once generated, all of these dashboards require ongoing maintenance and updates, and they can be duplicative, are static and often become orphaned.”

Traditional insights used to be generated by analysts querying the data directly. But this, McMullen says, required a reasonable level of skill to do safely and effectively.

“We are starting to see AI that will assist business users in detecting the 'unknowns' that they previously had hoped to find by staring at a dashboard. Additionally, the AI will be able to tell the user what the likely drivers are and what should be investigated further,” he says.

“Going forward NLP and AI will likely replace adhoc questions and analysis, while dashboards will be used to track KPIs. We see dashboards as being best for tracking metrics, but not so good at detecting new insights. 

“We also think there will be a trend towards self-service as non-analysts will be able to ask and answer most of their own adhoc questions, leaving analysts to do actual analytics.”

Commbank last month announced it would hire 50 engineers a month for a year across various disciplines to meet demand and improve the quality of data.

“Quality data is often the biggest challenge, but once that it solved, translating business questions into how you ask those questions of your data and then interpret the answers is one of the biggest challenges,” McMullen says.

“This is where AI will change the game.”

How AI is helping publishers

The shift towards automated insights is not limited to any one sector. Publishers are using ‘news feed’-style data analytics for their content to make decisions in real time.

Leon Bombotas is the founder of automated data analytics firm Incites.ai, which has worked with Medibank, Australian Community Media and Village Roadshow. He says an AI-generated news feed is about shortening the gap between data and action.

“Machines can identify those statistically meaningful events - those moments worth acting on - much faster than the human eye can. It’s not an outright substitution, it’s augmentation. It’s about using the machines to make us more effective at the roles we’re doing,” he said.

“It’s game-over for dashboards.”

Christine Anderson, the Director of Content and Audience for McPherson Media Group – which has worked with Incites.ai, says the company had a 44 per cent increase in average weekly registrations in February 2020 compared to November 2019 after adopting AI-generated insights for stories.

MMG publishes more than a dozen regional publications and, without increasing the number of stories it published, saw a threefold increase in email opt-ins over the same time.

“Newsrooms are notoriously time poor and we found that a focus on data could become overwhelming for our editorial staff and the value of the data in front of them could be unclear,” Anderson says.

“Ultimately, it’s resulted in making data more approachable for our newsroom, giving us more time to focus on writing for our audiences, rather than trying to explain and decipher individual data points and turn it into something actionable.”

In April this year, Gartner published research that called for more “data interpreters”, concluding there is a “lack of influence and impact of analytics on decision making”. At the same time, marketing analytics teams are expected to keep growing, Gartner says.

“In search of greater efficiencies and attributable value, marketing leaders are increasingly in-sourcing and centralizing analytics,” Gartner’s analyst wrote.

There is, however, a growing body of thought that businesses should invest in “augmented analytics”, which can replace “self-service analytics” – dashboards – in informing quick, data-driven decisions.

These “data stories” can be automatically generated as AI scans incoming data.

Dashboards not necessarily dead – they’re just evolving

The past year of Covid has accelerated the volume and velocity of data that businesses are using to get a competitive edge, Lionel Kho, Vice President of Data and Analytics at Dentsu’s Merkle, says.

Kho was recently named the country’s top analytics leader by the Institute of Analytics Professionals of Australia (IAPA). He said the rate at which people produce data has accelerated enormously and a key focus should be developing new ways to quickly learn from it.

“AI reduces the time to process and exploit large volumes of data. This is traditionally a workload undertaken by data analysts. There are some outstanding humans out there, but generally speaking, humans process one thing at a time,” he says.

“We don’t have the brute compute of the cloud.”

The shift, increasingly, is towards predictive business decision-making, rather than retrospective. Dashboards are less useful than they were – rather, leaders are moving towards search engines that are intuitive. 

“What has happened before? How did that campaign perform? As a business becomes more sophisticated in its use of data and analytics, it starts to implement machine learning within its workflows, so the nature of the questions themselves changes. They become predictive. Instead of referring to last year’s performance, we’re predicting next year’s, and identifying what initiatives are required to bridge the gap or reach a target,” he says.  

“This is about search for above category yield. If you want to outperform your competitor, one way to do that is to make better use of data than they do.  

“Dashboards won’t die, but they will evolve. The fundamental purpose of a dashboard is to communicate information visually. So long as those decision makers need to be provided information to make decisions, dashboards are not going to disappear.”