As businesses of all sizes rush to make sense of the growing volumes of data they're collecting, they face a variety of business intelligence challenges that complicate efforts to make BI processes productive, effective and useful.
The challenges are shaped by multiple factors, including diverse data infrastructures, data management issues, new types of BI capabilities and varying levels of data literacy in the workforce. On the one hand, BI teams must ensure that proper data governance and security protections are put in place; on the other, they need to demonstrate how BI can benefit workers, including less data-literate ones.
Another set of BI challenges centers around changes in the ways that business intelligence tools are being used in organizations.
"Traditional BI typically involves curated data and applications driven by IT," said Porter Thorndike, senior director of strategic services at Information Builders, a BI and data management software vendor. The traditional approach commonly provides information to business users via reports and specialized portals, with well-defined workflows, Thorndike said. In contrast, modern BI initiatives are often driven by business units using self-service BI, data preparation and data visualization tools to hunt for insights.
In many cases, the challenges start with getting approval and funding for a business intelligence program and developing a solid BI strategy that meets business requirements for information and can deliver the promised return on investment. In addition to conventional querying and reporting, BI strategies often need to incorporate mobile BI, real-time BI, augmented analytics and other specialized applications, further increasing the deployment and management challenges.
As they address all the issues and problems, BI and data managers need to strike the right balance between agility and good governance. Faster time to insight can provide a competitive advantage. But that needs to be balanced against data security and privacy concerns and the risk that business users may perpetuate inaccurate findings. As Thorndike put it, "Is the speed at which those insights are generated worth acting upon knowing that some of the insights could be faulty?"
Here's a more detailed look at the top business intelligence challenges for enterprises, plus advice from BI practitioners on how to avoid and overcome them.
1. Integrating data from different source systems
The growth in data sources means many organizations need to pull together data for analysis from a variety of databases, big data systems and business applications, both on premises and in the cloud. The most common way is to deploy a data warehouse as a central location for BI data. Other approaches are more agile -- for example, using data virtualization software or BI tools themselves to integrate data without loading it into a data warehouse. But that's a complicated process, too.
"While the BI tools have the capability to merge data from different data sources on the fly, it still requires a combination of technical skills and data understanding to make this happen," said Ramesh Hariharan, CTO at consultancy LatentView Analytics. That limits scalability and increases the time it takes to analyze data and deliver BI insights, Hariharan added. To help speed that up, he recommended creating a data catalog that contains information about data sources and lineage for users.
Other types of data integration challenges require navigating various technical tradeoffs. For example, Sameer Dixit, general manager of data, analytics and AI/machine learning at consulting services firm Persistent Systems, said his team ran into problems refreshing Microsoft Power BI reports on the fly from a Google BigQuery cloud data warehouse at one customer. The team found it could refresh the reports in memory instead, but performance was constrained by the hardware used, Dixit said.
Dixit's team also faced challenges in harmonizing different file formats used by a client's source systems. It explored creating a separate extract, transform and load (ETL) mapping for each file, but that would have taken too much time, he said. A better approach involved writing a Java application to bring all the files into a common format that could be processed efficiently in a single ETL job.
2. Data quality issues
BI applications are only as accurate as the data they're built on. "Data quality is one of the most crucial aspects of BI that is often overlooked," said Soumya Bijjal, director of product management at Aiven, an open source data infrastructure platform provider.
Users need to have access to high-quality data before beginning any BI projects, Bijjal said. But she added that, in the rush to aggregate data for analysis, many organizations neglect data quality or think they can just fix errors once they sort out the data collection issues.
The root cause can be a lack of understanding about the importance of proper data management across an organization. When deploying BI tools, Bijjal recommended that organizations create a data collection process that involves everyone in thinking about how to ensure data is entered accurately, along with a data management strategy that provides a solid foundation to track the entire data lifecycle.
3. Data silos with inconsistent information
Siloed systems are another common business intelligence challenge. Data completeness is a necessity for using BI to speed up and improve business decision-making, but Bijjal said it's difficult for BI tools to access siloed data with varying permission levels and security settings. BI and data management teams must break down silos and harmonize the data in them upfront to get the desired results, she added.
Many organizations struggle with that, though, due to a lack of internal data standards in different departments and business units.
"This is one of the hardest things to overcome because a lot of definitional work needs to be done spanning business functions," said Cameron Cross, technology senior manager at business and IT consultancy West Monroe Partners. For one client project, his team had to assemble senior business leaders in a room and get them to align on fundamental data definitions, such as what constitutes a pair of eyeglasses.
Inconsistent data in silos can lead to multiple versions of the truth, said Garegin Ordyan, head of analytics at data integration vendor Fivetran Inc. Business users then see different results for key performance indicators (KPIs) and other business metrics that are labeled similarly. To avoid that, Ordyan recommended starting with a well-defined data modeling layer and clear definitions for each KPI and metric.
4. Creating a data-driven culture
"One of the ongoing challenges is around creating a data-driven culture, not just at the executive level, but at the front lines, where the business truly interacts with the world around it," said Sudheesh Nair, CEO of BI and analytics software vendor ThoughtSpot. In his view, building that kind of corporate culture requires organizations to succeed on two fronts: equipping workers with the right tools and empowering them to apply the insights those tools generate in business processes.
Nair said BI managers need to enlist business leaders from all parts of an organization to help drive a cultural shift that prioritizes the use of data analysis to inform decision-making. It's also important to include middle managers in facilitating that change, he said.
5. End-user training
Training and change management programs related to BI initiatives also require the involvement of business executives and managers to be successful, according to Nair and others.
For example, Chris Fielding, CIO at disaster recovery vendor Sungard Availability Services, said her team worked closely with the company's HR team to develop a BI dashboard with global data on employee head count, new hires and terminations, compensation and other metrics. Completed in 2019, the new dashboard is updated automatically, replacing a manual reporting process that took hours.
The dashboard was quickly adopted by business leaders in HR, and Fielding's team created a simple training program for other users to facilitate a rollout of the tool across the company. She said that led to other departments and business units recognizing the dashboard's potential, which boosted adoption and spurred numerous requests for additional BI applications throughout the organization.
6. Managing the use of self-service BI tools
Uncontrolled self-service BI deployments in different business units can lead to a chaotic data environment with silos and conflicting analytics results that create confusion in the minds of business executives and other decision-makers.
Information Builders' Thorndike said most modern BI tools have a data and security architecture that provides a protected place for user-generated analytics to be stored and shared. He recommended, though, that BI and data management teams curate data sets in data warehouses or other analytics repositories upfront to help avoid inconsistences.
"We've found that the key to enriching the self-service experience is to expose these tools to curated data and content, which users can leverage to create much better data flows and mashups," Thorndike said.
However, LatentView's Hariharan cautioned that enterprises need to balance creating standardized metrics and dashboards and enabling users to define and publish their own. That requires careful considerations, he added. For example, when the freedom to explore data and publish findings is unfettered by any centralized governance principles, self-service BI users may publish dashboards that have overlapping KPIs or metrics that are defined differently from one dashboard to another. On the other hand, too much control may hamper analytics innovation and agility, Hariharan said.
In addition, BI tools are often modified to include custom extensions that meet specific business needs, Fielding said. Over time, such changes obstruct the ability to deploy standard product upgrades. To prevent that problem, she said BI teams should work with end users to understand their needs and find ways to deliver required data and dashboards using out-of-the-box functionality.
7. Meeting user requirements for BI data and capabilities
A restaurant management software provider that Thorndike worked with was delivering dashboards and reports on metrics like customers, revenue and food consumption to thousands of restaurants via a SaaS application. However, the company discovered that some restaurant managers were loading their data into separate BI and data visualization tools for further analysis. That was a major security risk, and it detracted from the overall customer experience because the managers had to use multiple tools to generate the analytics they wanted to see, Thorndike said.
So, the company added white-labeled versions of embedded BI tools from Information Builders to the SaaS application for advanced users to access. Doing so kept the users within the application for data mining and analysis, Thorndike said. It also provided other benefits, such as making it easier for users to collaborate on BI projects and analyze their data against anonymized competitor benchmarks.
At Sungard, Fielding said she has found that it's important to keep the collective interests of all users front and center when working with business units on their BI needs.
8. Low adoption of BI tools
End users often take the path of least resistance and look to continue using tools they're familiar with, such as Excel or SaaS applications. "Rather than use the BI tools to analyze data to derive insights, they export data and then perform their analyses elsewhere," Hariharan said. "This results in unexpected usage patterns that are not optimal and low rates of adoption of these tools."
Hariharan recommended continuously monitoring user activity and logs of user requests to identify potential adoption problems and issues with BI tools. BI teams should also aim to deliver continuous functionality enhancements with an eye on boosting user adoption, he said.
If you're just starting a deployment, user adoption often hinges on finding a good use case that quickly demonstrates tangible business benefits and encourages people to embrace a new BI tool, Fielding said.
9. Bad data visualization and dashboard design practices
Data visualizations often go wrong, making it hard to decipher the information they're trying to illustrate. Similarly, a BI dashboard or report is only valuable if it's easy for end users to navigate and understand the data that's being presented. But organizations often focus on getting BI data and the analytics process right without thinking about design and UX.
Persistent's Dixit said BI managers should involve a UX designer from the start to develop an intelligible look for dashboards and reports with a visual interface that has advanced features but isn't cluttered. BI teams should also promote good data visualization design practices internally, particularly in self-service BI environments. Those steps are especially important for mobile BI applications on smartphones and tablets with small screen sizes.
10. BI application and dashboard deployment
In many cases, BI applications and dashboards need to be deployed to end users on a continuous basis. But doing so can be a challenge: Manual deployments are hard to scale and prone to errors, according to Fielding.
To combat that and help bring down costs at the same time, her team has developed custom code for DevOps-style automation to standardize the overall BI deployment process. BI code migration and deployment scripts make it possible to quickly deploy dashboards in a repeatable way. "This enables our team to allocate more resources to working on the [project] backlog, rather than the deployment," she said.
11. Data model and dashboard maintenance
In a rush to generate new BI and analytics applications, BI teams and business users sometimes add metrics, data models and dashboards without any consideration for maintaining them and tracking their usage. "Maintenance tends to be the work that no one wants to do, but it's incredibly important," Fivetran's Ordyan said.
BI managers should routinely look at what's being used and what isn't and then deprecate things as needed, Ordyan recommended. Otherwise, neglected data models, data definitions and dashboards will continue to accumulate. "If you don't delete the things that aren't being used, you're going to find yourself in a situation where you have too much to maintain for no good reason," he said.