Business leaders are impatient for data that can yield game-changing insights, and they're well aware that enterprise applications generate rich stores of it. That's pushing more organizations to seek out embedded data analytics tools, a trend experts say will empower more business users. It's a win-win for business users and IT, and the sky is the limit for adding embedded analytics.
But the sky may also be the limit when it comes to the cost, experts warn, in part because more complex applications and analytics capabilities will require rethinking your existing infrastructure. Then again, no one said democratizing data would be cheap.
Today, business intelligence (BI) and analytics solutions still operate as separate, standalone tools. For business users, that means they must leave applications to analyze data. Embedded analytics give business users immediate access to that data from within the application, Gartner notes in its most recent Magic Quadrant for BI and Analytics Platforms.
It's not a new concept, of course. For some time, enterprise apps have embedded basic reporting and BI functions. What's new is that enterprise application vendors now have the technology to offer more complex embedded analytics.
SAP pioneered the way by retooling its popular ERP applications to run on HANA, an in-memory database. Oracle offers a similar coupling for Exadata, its in-memory database. In-memory appliances allow companies to store larger databases and perform more complex analytics, including predictive analytics, all from within the application.
Independent software vendors and companies are following suit by embedding predictive analytics. Generally this means they're either acquiring or building their own analytics tools or partnering with existing BI/analytics vendors to integrate these functions into their core applications or platforms, according to Gartner.
"They are also incorporating more advanced and prescriptive analytics built from statistical functions and algorithms available within the BI platform into analytics applications," the report states. "This will deliver insights to a broader range of analytics users that lack advanced analytics skills."
Embedded analytics are part of a larger trend to give users more access to data when and where they need it, said Mark Shilling and Nitin Mittal, principals at Deloitte Consulting.
"It's analytics at the point of impact, in real-time, to embed intelligence in the processing chain earlier, whereas traditionally it was at the back end of the processing chain, often too late to affect the outcome you want, in the hands of few, in a central corporate department," Shilling said. "Now the trend is to push that earlier in the process lifecycle and really improve the lifecycle for when it's needed most and when you can get the most value out of it."
While this wasn't possible a few years ago, emerging technologies such as in-memory processing and cloud allow vendors to support embedded analytics. While SAP and Oracle rely on in-memory, other vendors are using the cloud to power embedded analytics, sometimes through third-party plug-ins.
"Embedding applications is becoming very much a high market priority," Shilling said. "Different vendors are taking different approaches to that, but definitely it's becoming a key opportunity for our clients."
Embedded Analytics Use Cases
Gartner points out that embedded analytics are usually added to transactional and operational systems. Shilling and Mittal point to two popular uses for embedded analytics:
- Customer applications. Embedded analytics allows marketing and sales to run sentiment analysis and micro-segmenting on customer data. That information is then used to offer new products, upsell the customer or better target marketing spend.
- Supply chains. B2B vendors are also adding embedded analytics to optimize networks, supply chains, logistics and pricing.
"Almost anywhere you look on the processing chains, there's huge opportunities to exploit the information more aggressively," Shilling said.
As with all new technologies, business and IT leaders should be aware of "fine print" issues before they invest. The ability to add embedded analytics anywhere can be a pitfall for organizations, Mittal pointed out.
"The bottom line is that CIOs need to be able to articulate a business case, the business case needs to be quantifiable, attributable and measurable, and it needs to be anchored in either cost savings or value generated or some other measure that is meaningful to the organization," Mittal said. "If it is not, then yes, more often than not, you would end up basically overspending in this particular area."
More significantly, CIOs must address some foundational issues before adding sophisticated data analytics tools, Shilling warned.
"If the volume of information is doubling every two to three years, that means the CIO's infrastructure is likely being crushed," he said. "It means it's quickly becoming outdated. It's likely there are processing bottlenecks in that environment. It's likely there are significant cost pressures and that there's a burden to manage, govern and secure in different places within the scope of their responsibility."
A key part of modernization will mean supporting new data types, such as mobile, social, location data, machine-to-machine data, video and images, he added. It all adds up to an unavoidable problem: Infrastructure changes are "almost non-negotiable to run the business," he said.
Embedded Analytics Cannot Do It All
Finally, embedded analytics are not a substitute for enterprise-wide data analytics, according to data analytics veteran Sachin Chawla, who recently joined data transformation startup Trifacta as the vice president of engineering. Chawla, who previously worked for both Oracle and Informatica, said he sees value in embedded analytics, but warns that applications aren't designed to hold the diverse and large amounts of data used for enterprise analytics.
"There are important questions you need to answer that are localized to that application," Chawla said. "So I think that's great, that's of value and that's important, but really the types of questions that people are asking now, they're not going to be constrained to an application."
For analytics that draw from multiple applications and datasets, organizations should think about building a repository with some compute power, so you can make the data available across the enterprise, he added.
For CIOs, these details mean the hard work is still ahead.
"Your predictive abilities are only as good as your ability to comprehend the data, make some sense of it, analyze it. And that requires some very fundamental, basic techniques with respect to how you aggregate data, how you derive semantic operability, how you annotate that data, curate that data and develop algorithms to uncover the right insights, patterns and correlations," Mittal said. "That's the hard work ahead, which companies need to think through and plan in the next six to nine months."
Loraine Lawson is a freelance writer specializing in technology and business issues, including integration, health care IT, cloud and Big Data.