How to identify and implement embedded analytics opportunities
Business users need to consider data science workflows and software development to identify opportunities for implementing embedded analytics for business value.
Embedded analytics spans both software development and data science workflows. Software development tends to focus on functional requirements. Data science and analytics tend to focus on the best way to develop models. Business users need to step back and consider both disciplines to assess embedded analytics opportunities.
To get the ball rolling, it's helpful to identify a few embedded analytics examples that may map to existing problems. Developing experience with these projects can help teams discover new embedded analytics opportunities.
Start by analyzing the user experience. Teams should also address customization and integration challenges. It's also helpful to develop an iterative process that enables embedded analytics projects to improve over time.
Here are four places to look for opportunities:
- New product offerings. Sal Stangarone, partner at Michaels, Ross & Cole Ltd., a software development consultancy, said many software vendors turn to embedded analytics as part of a new product offering. The technology helps them improve their product without the need to build analytics features into their software. It lets them offer analytic capabilities as another selling point, or even as a new paid feature to improve the customer experience.
- Inferior reporting tools. Stangarone sees many people turning to embedded analytics to improve the lackluster reporting features found in software packages. ERP reporting is a typical example. If a company does not like their ERP's reporting capabilities, they can embed custom reports or a reporting tool into their ERP to deliver high-quality data to the business.
- Process gaps. It is also helpful to work through how analytics might improve common business decisions. For example, before a manager approves a pay raise for an employee, they likely want to check that it is in line with the pay for other people on the team. Interviews with managers can identify gaps in this process, such as needing to leave the system to answer questions before making a decision. This is an excellent opportunity to enhance the user experience by integrating the insight that supports the required action so it is streamlined in the same workflow.
- Decentralized systems. Robby Powell, advisory product manager at SAS, assesses decentralized systems for embedded analytics opportunities. Look for cases where metrics are collected from multiple sites. Decentralized embedded analytics can provide value by helping users identify issues that need to be addressed as soon as possible, and at the source of the problem where it is easier to manage. He recommends focusing on situations with a need for high model accuracy. Embedded analytics can help evaluate model performance at each edge location and in aggregate, and then update models as needed.
Customize the experience
Once teams have identified embedded analytics opportunities, it's time to start weaving them into applications. It's helpful to think about the user experience as part of this process. Sri Raghavan, director of data science and advanced analytics at Teradata, said this first requires understanding each user's unique data and analytics needs, as well as the type of environment in which they consume these insights. For example, some analysts prefer relevant KPIs embedded directly into their apps.
Next, teams need to work through customization issues when embedding analytics. Customization helps analysts focus on their areas of need. Raghavan said this usually requires finding ways to take only the relevant subsets of data and insights that are typically crammed into a large dashboard and displaying them to specific individuals or groups responsible for the task.
Integrate and iterate
Integration is one of the biggest challenges of embedded analytics, Stangarone said. Much of it boils down to architecture: Some vendors build their software on proprietary architecture, while others use open architecture and frameworks.
"It's crucial to understand how your software will integrate with the vendor's platform before you even start," he said.
He worked with one software vendor who spent three years and more than $100,000 trying to embed analytics software within their product, but still had a long way to go. Rather than continue spinning their wheels, the vendor switched to a different platform built on open architecture and got up and running in less than three months.
It is helpful to develop a process of revisiting embedded analytics apps to deliver the intended value and adjust when needed. Embedded analytics is likely to be driven by IT teams, but it is not just an IT initiative.
"It's an iterative process that requires collaboration and alignment from individuals across the organization," said Scott Gnau, head of data platforms at InterSystems.
Different teams should work together to constantly assess the contributions being made by introducing embedded analytics. Use cases and required metrics should be reviewed often to understand whether the efforts are providing value and what changes might be needed to measure progress. This helps to iron out any issues as they occur and ensure that all users are extracting real value from the platform.