Virtual Analyst for Adobe Analytics Helps Uncover Customer Insights

Adobe's new virtual analyst uses AI and machine learning to help companies uncover information on customer interactions and trends previously hidden or that required far greater time by data scientists to make accessible.

Adobe

With companies accumulating customer data at an unprecedented rate, it’s getting harder to find and analyze the key trends and insights that could impact what products to make, where and when to sell them, and a raft of other development, distribution and marketing decisions. Adobe said it has a solution that will make these insights far more accessible to product and line of business managers without going through the lengthier process of getting data scientists to research and produce the reports. It is a service that data analysts and scientists will also want to leverage as it promises to do the digital digging needed for customer insights much easier, the company said.

Designed for Adobe Analytics, the new virtual analyst, announced on Sept. 24, surfaces customer insights without the user having to ask for them or design complicated queries to get the desired insights. More importantly, it can deliver insights that traditional analytical queries won’t find.  

“For the first time we are baking machine learning and AI [artificial intelligence] directly into the interface of Adobe Analytics,” said John Bates, director of product management for Adobe Analytics, in a press briefing.

“The amount of data enterprises are collecting is staggering—literally a mountain of data—yet the amount of what’s collected for analysis is a small sliver of that,” he added. “What is happening with the rest of all that data, and what risks can be found there that brands may not be aware of?”

The virtual analyst uses deep learning models to tap all the data points connected to customer interactions, from how long a consumer spends on a website to the consumer’s movements between an app and the web. Adobe said insights from this kind of data, what it calls “unknown-unknowns,” have never been available to brands before because they didn’t know where to look or what questions to ask to find them.

“Marketers and analysts typically look for the ‘known knowns’ like, ‘Yup, our home page still gets the most traffic of all our pages,’” said Bates. And while you might have a system that alerts when, for example, shopping cart abandonment is up 20 percent, you don’t get real insights as to why.

A Netflix for Marketers?

“It’s almost like a Netflix kind of recommendation for marketers. It gives them a signal to insights related to their customer data they wouldn’t otherwise get,” said Bates.

One early test customer, a travel company, told Adobe it would have needed to hire a hundred more people to get the kind of insights the virtual analyst gives them.  

As with other AI systems, Adobe’s virtual analyst “learns” over time. Adobe said the virtual analyst takes into consideration the preferences and consumption patterns of users to deliver more intuitive and relevant insights. And by analyzing the behavior of other users within the company, it can find people with similar use cases to make better decisions.

Almost three years in development, the virtual analyst is built on several Adobe Analytics solutions such as Anomaly Detection, where the system looks for statistically significant deviations in data. Another is Contribution Analysis, which identifies the factors that contribute to anomalies.  The virtual analyst also leverages Sensei, Adobe’s AI and machine learning framework.

“In the past, you had to manually create the metrics for what you would analyze,  but the virtual analyst automatically creates and analyzes all the metrics, potentially thousands of them” said Bates. “And this is done to the individual’s preferences and pushed out without them necessarily even asking for them.”

He also noted that Adobe is working on additional capabilities, calling the new release “a stepping stone to the broader vision we have.” You can expect to see, for example, additional features down the road like proscribed next actions based on insights the system uncovers from email campaigns and A/B testing.

David Needle

David Needle

Based in Silicon Valley, veteran technology reporter David Needle covers mobile, bi g data, and social media among other topics. He was formerly News Editor at Infoworld, Editor of Computer Currents...