Photo courtesy of Depositphotos
Opinions expressed by Digital Journal contributors are their own.
Data analytics has been transformative for effective business operations, especially since generative AI went mainstream in late 2022. Real-time insights drawn from enormous datasets, which include both sensitive internal proprietary data, and external public data, are becoming a prerequisite for business success, leading to the popularity of embedded analytics.
“Embedded analytics allows an analytics platform or product to be added inside a host application,” explains Avi Perez, CTO of Pyramid Analytics. “The platform can be quite sophisticated, involves a huge number of capabilities and functions, and ultimately allows the user to visualize or ask questions through a point-and-click interface while working within the host application.”
When done right, embedded analytics allows users to tap into advanced decision intelligence capabilities in the context of the apps they’re already using, rather than manually importing data and switching contexts to a dedicated analytics app. This takes the transformative ability of data analytics and spreads it all around the organization and its various work environments.
However, Perez notes that since GenAI came along, the art of implementing embedded analytics has changed dramatically.
True embedded analytics delivers important benefits
Why is it so important to offer rich analytics capabilities within other apps, and how does AI make that more accessible? The fact is, without embedded analytics, there’s too much friction around access to insights.
Perez gives the example of a sales leader who’s looking at their CRM and wants to see a visualization of leads closed by month by sales rep. They want the data right in front of them in an easy-to-read chart, not to have to leave the platform and log into a different one. “That’s a super non-functional way of doing things,” Perez points out. “What makes it more complicated is if the CRM data needs to be migrated out to the BI tool, there’s another hop and another step and another issue, especially for that user.”
This screen-switching can be irritating even for people who are skilled at data science, and most line-of-business users don’t have those skills. “The users of embedded analytics tend to be very non-technical, casual users who have no details or idea about the data inside of those artifacts,” observes Perez, adding that they are easily overwhelmed by the process of accessing insights. As a result, a lot of the potential value of data remains untapped.
This is where new methods of embedded analytics come into play. They can automatically pull the right data visualizations into the business app that someone’s working on, without the user needing to switch screens or wrangle with raw information. “When you’re inside an application and you’re using it, it’d be very, very useful if you have the matching analytical components appearing side by side with whatever you’re looking at, at the same time,” Perez explains.
What’s more, the inclusion of GenAI removes roadblocks for users who don’t have data science competence or even much data literacy. Users can ask queries in natural language and receive clear visualizations in response, without having to import or export data or work out which visualizations to request. This makes them more likely to use data insights, driving better decision-making, more accurate forecasts, and optimized operations.
But there’s a reason why this utopia is difficult to achieve. There are many bumps in the road and obstacles that need to be overcome.
Traditional embedded analytics needed an upgrade
Older embedded analytics solutions used what Perez refers to as “watered-down” versions of full-blown analytics tools. It’s difficult to fully integrate them into other business apps because they are often slow, heavy, and difficult to scale. They tend to drag down the performance of whichever platform they’re embedded into.
This inability to truly embed analytics into other apps also created a barrier for the vast majority of line-of-business users, who don’t have any data science expertise. According to Perez, the marriage of GenAI and embedded analytics brings a new range of possibilities. Users can ask queries in natural language and receive the data insights they need without friction or fuss.
But he’s not skating over the challenges along the way. GenAI-powered embedded analytics can be an extremely complex undertaking. Many technical difficulties still need to be overcome, and other processes need more refinement, before we see the full potential of embedded analytics.
But the results, in Perez’s opinion, are well worth it.
The promise is difficult to realize
The technical barriers to true embedded analytics are high. Perez explains that business solutions are increasingly arriving as single-page apps using frameworks like Angular and React. If you want your analytics capabilities to integrate smoothly, they need to be lightweight, efficient, and scalable. Incorporating GenAI adds another challenge, he adds, because it needs to be “super simple, super easy, and come without any baggage.”
Security is a perpetual worry. You need to enable access to all the data your analytics tools require, which generally includes proprietary data, while also protecting your sensitive data and critical business systems. Some GenAI tools are open to the outside world, which creates more vulnerabilities and increases the security risks. You also need to account for the fact that different users might have different access permissions.
Performance is another concern, especially when you’re talking about customer-facing apps. Perez warns that “when it’s an in-house tool, you can control the client environment. When it’s customer facing, you don’t know what kind of a device they’re using, what kind of a browser they’re using, and suddenly scaling and performance is a lot more of an issue.”
Customer-facing solutions also add to the security headache, because it makes the landscape far less transparent and more threatening. “In-house tools are much easier to develop than customer-facing tools, because you own all the domains,” says Perez. “Security is much easier. All the stringencies around cross site scripting are lesser. The minute it becomes a customer-facing app, it’s exposed to the external internet, and suddenly these problems go up a few notches.”
AI helps embedded analytics to cross the next frontier
Despite all the many obstacles that lie along the path, Perez is confident that AI will make a real difference to embedded analytics. “The market is getting more and more into customer-facing embedded analytics applications,” he says. With demand rising and talented teams like those at Pyramid Analytics working on the challenges, developers are finding it easier to implement true embedded analytics.
