Embedded analytics is a paradigm for weaving analytics directly into business applications, mobile apps, equipment, consumer electronics and even beehives. The core idea grew out of frustrations with traditional BI development lifecycles that limited the utility and reach of analytics.
Kurt Schlegel, then-research director at Gartner, introduced the term embedded BI in 2007. He observed organizations that presented BI inside of business processes were better positioned to take effective action. Since then, the industry has extended this core idea to support a variety of analytics capabilities.
Closer to the user
"Embedded analytics brings data closer to the user by incorporating charts, graphs and dashboards into the applications decision-makers use every day," said Chris Bergh, CEO of DataOps consultancy and platform DataKitchen.
But care must be taken to surface insight that users find valuable. With traditional BI, an expert can provide guidance and insight as to how predictions might suffer. Careful planning is required to extend this insight into embedded analytics to foresee how bad data might affect users. If your car predicts your future mileage based on highway conditions, you may run out of gas when you get stuck in city traffic. The problem lies in putting the right data into the analytics in the first place.
"When discussing embedded analytics, people will focus on what the user sees," Bergh said.
What is less pronounced, and perhaps just as significant, is the system built around embedded analytics. A better mileage graph could be improved by adding mapping data that reflects where the user is likely to drive.
But, because a medium to large enterprise might integrate thousands of data sources, care is required to ensure the embedded analytics brings accurate and actionable insight to business users.
Pushing analytics to the edge
The growth of IoT capabilities is dramatically increasing the amount of data that could theoretically improve business decisions. This includes things like tracking data from RFID tags in pallets, product data in stores and diagnostic data from the devices used to move products around. Analytics leaders are looking for ways to provide insight closer to the employees on the front lines so that they can take action immediately when required.
"There is simply too much sensor data being generated for analytics to be run centrally and routinely deployed at all source locations in a timely manner," said Robby Powell, AI and analytics product manager at SAS.
Pushing analytics to the edge enables custom-tailored action to be taken at the point where data is being collected. While decisions are being pushed to the edge, centralized evaluation of model performance and decision impact is critical to assuring the accuracy of the decisions.
Thus, the embedded analytics lifecycle needs to consider how to improve the workflow of the person on the front line. It's also important to empower BI analysts and data scientists to evaluate model and decision accuracy through key indicators viewed both in the aggregate and at the detailed level.
It is also essential to understand that many enterprise workflows use Microsoft 365 applications. Integrating insights gained through embedded analytics into Microsoft 365 documents, such as Excel and PowerPoint, reinforce the attention given to those efforts.
How does embedded analytics benefit BI?
Embedded analytics often complements or extends the reach of BI rather than replacing it. The main benefit embedded analytics provides to traditional BI workflows is improved UX that helps create buy-in across the organization.
It can also extend the expertise of BI experts and data scientists -- and their workflows -- to support users beyond the organization.
Building better beehives
SAS worked with Beefutures and Amesto NextBridge to apply advanced analytics to read bees' movements and identify the optimal location for beekeepers to place beehives to help preserve the bees. The beekeepers can access an interactive map on their cellphones and easily track hives, bees and their food sources throughout the day using embedded insights and visual analytics.
By tapping into embedded analytics, the map also identifies where the beekeepers could relocate the hives to optimize conditions or where to plant new food sources based on the time of year. Using maps embedded in a mobile app gives beekeepers who may or may not have any experience with analytics the insights to create a better set of conditions for bees to thrive.
The collaboration plans to capture, monitor and publish various decisions and results to a central system. Model tournaments will be continually run to build better decision models that can be pushed to the edge.
"Their goal of preserving bee populations also inspires social innovation, which is an area that is poised to only become larger in the future," Powell said.
Outsourcing inventory management
"Embedded analytics is a great match for sharing data outside the corporate walls since security and data access can be tightly controlled with custom-fit applications," said David Mariani, CTO and co-founder of AtScale, an analytics tools provider.
AtScale has worked with one retailer on an app that enables suppliers to query and download data in real time to manage inventory across 3,500 stores. This helped the retailer improve product availability during the COVID-19 pandemic, resulting in explosive earnings and sales gains when other retailers struggled. Essentially, the retailer outsourced its inventory management to its suppliers, whose best interest is to make sure they have products available for purchase.
Keeping it real time
Dan Simion, vice president of AI and analytics at Capgemini Americas, said the most compelling use cases for embedded analytics are tracking processes, activities or anything else a company would like to monitor.
The real value is that users do not have to wait for someone else to extract data or use another tool to create reports that showcase results. This starts with accessing the data in real time to understand business needs and matching that to what provides value to the organization.
"Don't just build reports for the sake of building reports, but ensure the data being captured is going to provide valuable outcomes," Simion said.
Watching out for bias
Enterprises are weaving advanced AI and analytics models into a wide variety of algorithms for automating decisions. Despite best intentions and work to mitigate bias in these algorithms, sometimes, it creeps in. For example, a hiring algorithm might ignore factors like gender when making a hiring decision but then end up selecting more male candidates for a job. Embedded analytics could provide a kind of engine warning light for these algorithms before it's apparent to business users.
"Algorithmic decision-making is biased because the humans that design them are biased," said Mark Palmer, senior vice president of analytics, data science and data products at Tibco. "By embedding analytics trained to identify common patterns of factor use that are signposts of bias, it can be more easily identified and reduced."
BI has traditionally focused on descriptive and diagnostic analytics. Embedding predictive and prescriptive analytics can help you take action when it matters. That includes optimization, simulation and predicting what will happen, rather than looking in the mirror to understand what already occurred.