pathdoc - stock.adobe.com
Embedded analytics has the potential to enable more widespread use of business intelligence.
Depending on the study, BI adoption within organizations has been stuck between about one-quarter of the workforce and one-third.
One of the main hindrances of widespread analytics use is the analytics tools themselves, which -- despite low-code/no-code and augmented intelligence capabilities aimed at making them easier to use -- are geared toward data experts rather than business users.
Another barrier is the effort it takes to use BI tools.
Even for those with the requisite training, it's burdensome to work in one application, realize the need for data analysis, leave that application to go into another designed specifically for BI, format the results of the analysis so it can be used with the format of the first application and finally blend the work done in the different applications.
Embedded analytics eliminates that unnecessary back and forth, essentially combining applications by embedding an analytics platform into another application so data is available for business users in their familiar work environment.
It also reduces the need to know how to use complex BI platforms by delivering data to business users within their workflows in an easily digestible way.
In fact, there may come a day when analytics is so seamlessly intertwined with not just workplace applications but also everyday life that it becomes like word processing is now, according to Donald Farmer, founder and principal of TreeHive Strategy. He noted that word processing was once a specialized skill and environment but is now so fully integrated with other applications, such as email and text, that it's not even noticed.
In May, Farmer, who was an executive at both Microsoft and Qlik before starting his own consulting firm, co-authored a book with Jim Horbury, a solutions program director at InterWorks, titled Embedded Analytics: Integrating Analysis With the Business Workflow.
Recently, Farmer discussed embedded analytics, including its evolution from the 1990s to now and where it might be headed in the next few years. In addition, he spoke about the relationship between embedded analytics and generative AI and how large language models have the potential to be the driver that makes embedded BI -- and BI use in general -- even more ubiquitous.
Editor's note: This Q&A has been edited for clarity and conciseness.
How do you define embedded analytics?
Donald Farmer: To start, you have to think about what analytics is. Analytics are ways of presenting data so it can be readily understood and one can detect patterns, trends and outliers. It's a way of giving people data so that they can make sense of it.
Embedded analytics is simply the same technology but inside another application. The reason someone would want to do that is because if it's a specialized application, then [without embedded BI], they would have to set aside special time to do analytics. They would have to think, "Now I'm going to do some analytics," and they'd have to open a specialized analytics application.
Instead, a lot of times, they'd rather have the analytics in the application they're working in. Embedded analytics is simply the analytics technology inside another application so the user is always close to the data and they don't have to switch context to work with the data.
Why is embedded analytics exciting? How does it advance BI?
Farmer: It's exciting because people are becoming more and more data literate and, as a result, becoming more and more demanding of data. They want to look at data. They want to understand data. They want to use data. Embedded analytics is a way of bringing that to them in a direct, simple way within their existing business workflow.
Picture someone at a call center. They can see their numbers, see an analysis of how they're doing right there in front of them. A real-world example is at Target. Watch the cashiers. After they finish ringing up your items and they're ready to move on to the next customer, an analytic pops up and tells them how they're doing in terms of how many items they've scanned, how quickly they've scanned them and whether they're on target [to meet their standards]. That's analytics right in the workflow of a cashier.
When did you first become aware of embedded analytics, and what did it look like at its beginning?
Farmer: The earliest embedded analytics were people embedding charts and graphs into applications. Most of the programming languages had data libraries that you could use when building applications and included charts and graphs. That was the earliest form, which has been around since the 1990s. But it was simple back then. In particular, the products didn't have analytic engines. They couldn't do complex calculations or complex aggregations. They could present data; they were purely data presentation capabilities.
Between 2005 and 2010, vendors such as BusinessObjects [acquired by SAP in 2007] did a good job of it and made it part of their business model. There were other companies that had capabilities, but BusinessObjects is the one that made it a core part of its business model. Then products like Qlik and Tableau started to make embedded analytics top-level capabilities of their platforms.
How has embedded analytics evolved in recent years?
Farmer: The breakthrough was the ability to not just embed a graphic or a visualization but to embed connections to an analytic engine so a user could analyze large volumes of data and make that interactive and exploratory. Rather than just being a visualization of something, it's now a genuine analytics experience that is exploratory, interactive and working against all the data it needs to work against.
If BusinessObjects was one of the initial pioneers of embedded analytics, which vendors are the ones that have advanced the technology in recent years?
Farmer: We started to see companies emerge that were focused on embedded analytics. An example of that is Sisense. Logi Analytics is another. So is GoodData. All three of them placed a real emphasis on embedded analytics from the time they were startups and have done a great job of it.
Are those vendors still the top practitioners, or have others caught up and made their own embedded analytics capabilities similarly vibrant?
Farmer: It is a broad spectrum now and difficult to parse who is best. That's a fine-grained discussion, and it would come down to requirements people might have. In general, it's now a well-served market. Everybody has embedded capabilities. It's now standard.
How then do Sisense, Logi, GoodData and other embedded analytics specialists stand out and attract users? Are they still able to differentiate themselves?
Farmer: They can differentiate themselves. They do it with the developer experience. They make the experience simpler and more powerful by being more supportive of developers. They do it by focusing on a different audience.
If you think of Microsoft, Tableau or Qlik, their core market is companies that already have those platforms deployed and want the embedded experience inside of their other applications. That's fine because it gives users a consistent experience across both their specialized analytics tools and their embedded analytics. That can be useful. It can reduce training costs and so forth.
But a lot of companies have complex applications that have complex embedding requirements or specific data or security requirements. Especially for retail or manufacturing applications, the enterprise environment might be complex. That's where specialized vendors can supply the features and developer support that's needed.
What's the relationship between embedded analytics and generative AI? Are they competing technologies, or do they instead enhance one another?
Donald FarmerFounder and principal, TreeHive Strategy
Farmer: Every one of the embedded analytics vendors -- especially the specialized ones -- are looking at generative AI and how to use it. Generative AI turns out to provide a good user experience for describing data to people. If you think of visualization to provide data to people, generative AI can be used almost as a form of explanation.
Generative AI can use natural language to describe data, especially for non-specialists who don't necessarily understand the details of data analysis or the complexity of statistics. That's a great experience for people. It's also a great experience for people who want to create SQL queries and other things like that. You can create an interface that is natural for the user and can generate powerful queries. That's a great use as well.
Generative AI as an interface for an analytic engine is a great experience. It doesn't replace the analytics engine because generative AI is not good at analysis. It doesn't understand enough about the math or data structure to build good analytic processes. But it is a good user experience.
Having had a breakthrough five or so years ago, how will embedded analytics further evolve over the next few years?
Farmer: Thanks to generative AI and the improved experience and accessibility of data through generative AI, we're looking at a new generation of more powerful analytics that will be on the user's pane of glass wherever the user is working. Generative AI has enabled that breakthrough.
For a long time, the bottleneck for embedded analytics was computing power. If analytics are embedded in an application and the application is running on a desktop, what the desktop is capable of is the limit. In-memory analytics and cloud analytics solved that, which means there is almost no limit to the power that's available to do the analysis.
The holdup then became the user experience. People still needed the skills and the insight into data to take advantage of powerful analytics. Generative AI helps that next stage. I think there's an exciting future for embedded analytics using generative AI as the experience. That will become the default experience in the future.
Embedded analytics has evolved from being a tab within an application to a more integrated aspect of an application -- how do you foresee it eventually being presented?
Farmer: It's no longer a special activity. It's now part of a user's workflow. Eventually it will become like word processing.
When I met my wife, she was a word processor. There are still some specialized word processors, but they're rare. Now we do word processing everywhere. It's not just that we write documents. We do word processing in our email; we do word processing on our phones when we write a text. In a sense, word processing has gone away as a specialty, and it's just part of everything we do.
My ambition for analytics is that it just goes away and becomes part of everything we do. Whenever you see numbers, you should be able to analyze them -- your credit card statements, your utility bills, sports scores, your child's school records. That's all data, so we should be able to analyze it and do comparisons and projections. My belief is we'll be able to do that, and embedded analytics is the way in which it happens.
Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.