To fix BI, build it into your applications

Business intelligence (BI) was once heralded as a technology that would democratize data, enabling everyone to become more productive and make better decisions. Today, though, analysts in the BI space like to share the same (and possibly apocryphal) statistic: The global business intelligence adoption rate is only 26 percent.

If only 26 percent of potential users ever access BI, something is broken. Why is access so poor? What can developers and engineers do to make BI achieve its full potential?

People tend to think of BI as dashboards that live in a siloed app, away from the productivity and business apps we use to work. I believe app developers have overlooked opportunities to make BI a core feature and source of value in their apps. While enterprises can retroactively introduce BI, the real shift needs to happen with the many enterprises and software vendors that build apps for their customers. We need to stop thinking of BI as a satellite and exclusive-access source of information and start embracing easy-to-use embedded analytics as a core strategy. 

Out of Context

Before joining Sisense, I spent 21 years at Microsoft, most recently leading one of their CRM services (Microsoft Dynamics 365). I regularly spoke to sales managers who wanted employees to enter quotes, forecasts, and so forth into their CRM. However, their employees rarely benefited from all that data entry, so they put minimal effort into it. They were among the approximately 75 percent of workers with no access to -- or no use for -- BI as a way to drive insights and decisions from their CRM data. 

Well, I’d ask customers, what if we could embed analytics inside the CRM? 

While that would be an improvement, they assured me, their users didn’t spend much time in the CRM anyway. They wondered instead if we could embed analytics in Microsoft Teams and Outlook, where the real discussions and decision-making happened. 

It struck me that these customers wanted convenient access to analytics, integrated seamlessly into their everyday apps, and more importantly, meshed with their other data to get to more powerful insights. 

Death to Dashboards. Long Live Embedded Analytics 

In almost everyone’s mind, BI was a chart or graph. Analysts would connect to a data source, run some queries, and produce some sleek visuals. That process became commoditized by vendors years ago. Today, the real innovation in this space is about shifting from BI to analytics, shaping decisions and actions in real-time, and embedding these analytics into everyday apps. 

As an example, the navigation app Waze has learned my morning commute to work, and if traffic is unusually heavy, it will send me a notification recommending that I leave home early. That way, I’m not late to the office. I don’t have to look at a map, graph, or chart of traffic to know when I should leave. Waze’s analytics does the analysis for me, which is ideal. Most users don’t want metrics -- they want to know what to do based on those metrics.

I also see manufacturers integrating analytics into the device itself. For example, one maker of CT devices uses analytics to measure the quality of the imagery. If the patient’s posture is improper for the image, or if the result is too grainy, it guides the technician to correct the issue immediately. This improves patient care and saves time, costs, and the hassle of retaking images at a future appointment. Analytics are embedded in the device and thus shorten the loop between insight, diagnosis, and action.

These examples are about business analytics infused into apps (and devices) by app developers. None of these examples require dashboards. True business analytics tells us what we need to know at the moment in context rather than dumping a pile of metrics on a screen and expecting us to figure it out. 

The Road to Data Democratization

So far, I’ve argued that around 75 percent of people in organizations don’t get the benefit of analytics because it doesn’t exist where they do their work. And even when it does show up in those places, it’s more of a disconnected visual artifact than a real-time guide to discrete actions. 

Many companies that develop apps for enterprises and consumers missed an opportunity to integrate business analytics into their products. Those who do so today will make their products more valuable to their users and develop new ways to monetize their product data. So, how can developers accomplish that?

1. Compose Analytics

Suppose you wanted to build e-commerce capabilities into your app. You would probably avoid building systems for checkout, payments, and inventory management from scratch. Why devote development resources to something that other companies do better? You’d find an SDK that enables you to compose your ideal e-commerce experience within your product. 

The same goes for business analytics. Vendors have already built tools for connecting to different data sources, cleansing data, caching data, writing analytical queries, and visualizing data. These days, standard practice is to compose analytics into your app via an API from a trusted analytics vendor. 

2. Semantic Modeling

The semantic model lives between the data warehouse and the analytical layer, where it translates between the physical world of data and the business world where that data has meaning captured by words. It allows the data experts to reframe their data in a way that makes it accessible to users who aren't as data literate, including the approximately 75 percent who don't interact directly with business analytics.

Recall that programming, like data literacy today, used to be a rare skill. We couldn’t train developers fast enough to meet the demand for their talent -- so we made programming more accessible to a wider audience by means of abstraction. From Object Oriented Programming (OOP) to declarative code, today’s high-level programming languages and tools are far easier to learn and use well because their terminology and structure have moved away from binary bits to human-readable language.

Likewise, if we are to overcome business analytics’ data-literacy hurdle, we must abstract our enormous data sets into something more human-friendly. That is the role of the semantic layer -- humanizing and abstracting machine-oriented data structures into something a wider audience can interact with and gain value from.

3. One CI/CD process

CI/CD (Continuous Integration and Continuous Delivery/Deployment) is a set of practices and tools that enable app development teams to deliver new code changes to production quickly and with high quality. analytics development should adhere to the same CI/CD process used to build the core app.

For analytics to become an integral part of applications, and one of the many tools the modern developer wields, it must act more like application code. And why shouldn’t it? An error in a chart displayed to millions of users is as impactful, if not more, than a typo in a paragraph or a visual glitch in the UI, let alone with more silent but much more dangerous failure modes such as displaying inaccurate data. 

If our analytics assets can impact our product’s success as much as any other code asset, then we should be using the same tools and processes to reduce and mitigate that risk -- whether that’s by making sure our analytics assets are source-controlled, just like our code, or by ensuring that they are also assembled, tested, and deployed the same way as our code -- via CI/CD.

4. Combining Analytics Insight with Actions 

In the past, the dashboard or report was the end goal. We put hours of work into generating a static report, hoping readers would take action based on the insights presented. But for analytics to live up to their potential impact throughout the organization, it cannot be the end of the line in a user’s data journey. Instead, data and analytics must be embedded and employed in day-to-day work. 

Integrated into an application, modern analytics platforms allow users to adapt charts to their needs by clicking menu options and filters. They also enable users to set triggers that automate actions in the app based on changes in data -- without writing new code. 

Waze, described above, is a perfect use case for data-driven action. The data indicated traffic is getting worse (Insight), pushing a notification (Action) to my phone that encouraged me to leave 10 minutes earlier than usual. 

We can bring analytics to users, but if we want them to engage with it, it has to be worth their while! That means ensuring that users can act on the insights they are presented with, as quickly and effortlessly as possible. Only when users can complete the cycle of analysis-decision-action will modern analytics truly close the feedback loop and allow users to gain the full benefit of their data: better decisions and better outcomes.

Living up to BI’s Promise

Often, we think of BI as a retroactive solution to a stack of applications with siloed data. To be clear, if organizations want standalone dashboards that around 75 percent of people don’t use, there are off-the-shelf solutions for that. However, I believe analytics need to be infused into applications and surfaced to developers via APIs to achieve its full potential. 

If developers build business analytics into their apps, they will democratize access to the benefits of using data. Embedded analytics is how BI will meet its full potential.

Photo credit: Maksim Kabakou / Shutterstock

Ariel Katz is CEO at Sisense.