Building Team-Driven Analytics and Trusting Data
Most enterprises can't fully leverage their data because they haven't established policies that build trust in the data and in collaboration.
- By Rami Chahine
- March 18, 2019
With the recent emergence of self-service BI tools and platforms, data analysts and business users are now empowered to unearth timely data insights on their own and make impactful decisions without having to wait for assistance from IT. It's the perfect situation for more agile, insightful business intelligence and therefore greater business advantage, right?
The reality is that even with these new BI tools at their fingertips, most enterprises still fall short of leveraging the real power of their data. If users don't fully trust the information (even if they're able to find and comprehend it), they won't use it when making decisions. Until organizations approach their data analytics strategy differently -- by combining all aspects of how the data is managed, governed, prepared, analyzed, and shared across the enterprise -- a lack of trust will prevent a business' data from being useful, ultimately turning it into a liability rather than an asset.
Finding the Balance Between Agility and Trust
Although the self-service features of modern BI platforms offer more freedom and greater analytics power to business users and data analysts, they still require enterprises to manage and maintain data quality over time. Various roadblocks impede enterprise users and data analysts from gaining access to the trusted data they need. Tactics for overcoming common and critical challenges include:
Building agility through proper data preparation. Many times, data prep -- the process of gathering, combining, cleaning, structuring, and organizing data -- is missing from the analytics equation, especially when business users or data analysts are eager to get results quickly. However, having the data clearly structured with a common vocabulary of business terms (typically held in a business glossary of a data catalog) and data definitions ensures that people can understand the meaning of available data, instilling trust.
Because data is pulled from both internal systems and external sources for reporting, profiling and cleansing data is essential to assure trust in data as well as to improve the accuracy and reliability of results. Any changes made to the data should be tracked and displayed, providing users with the full history of the data should they have questions.
Freeing (and maximizing) the siloed data. Data is often siloed within different business units, enterprise applications, spreadsheets, and data lakes, making it difficult to scale and collaborate with others. The rise of self-service BI has exacerbated this problem as more business users and teams have generated department-specific reports. People working in one silo are likely unaware of what data has already been prepared and shared in other silos, so time is wasted by reinventing data prep efforts and analytics rather than reusing and sharing them.
Integrating data prep with self-service analytics unifies teams across the enterprise -- including shrinking gaps between data analysts and stewards who have more context about the data -- and empowers citizen data scientists with trusted, curated data so they can focus less on hindsight and more on foresight.
Establishing a true north through data governance. Strong data governance practices provide an organization with structure and security for its enterprise data. This is especially critical when data is distributed through many systems, data lakes, and data marts. Governance is the umbrella term for the processes and rules for data, including assigned owners (stewardship) and data lineage -- so users can clearly understand the data's past use, who has accessed it, and what changes were made (if any).
For an organization to fully realize the value of its data, it needs a shared, user-friendly approach where all users across the enterprise have easy access to data they can trust to do their jobs, but in a way that is controlled and compliant, protecting data integrity. Organizations can balance the demands for convenience and collaboration with those of control by establishing and maintaining a three-tiered approach.
Tier 1: The Data Marketplace
Enterprisewide data use begins with the data marketplace, where business users can easily find (or shop for) the trusted business data they need to gain analytics insights for critical decisions. The data marketplace is where the rules of governance, shared common data prep, and shared data silos all come together.
This data marketplace concept is not a single tool, platform, or device. No single self-service data analytics tool can deliver the results organizations are looking for. Instead, the data marketplace is an overarching strategy that addresses data management and discovery with prep and governance to collect trusted data. The marketplace helps organizations address the challenges of finding, sharing, transmitting, analyzing, and curating data to streamline analytics, encourage collaboration and socialization, and deliver results. Creating a standard, collaborative approach to producing trusted, reusable, and business-ready data assets helps organizations establish a common portal of readily consumable data for efficient business analysis.
Tier 2: Team-Driven Analytics
Just as important as having quick and easy access to reliable data is the ability to share it with others in a seamless, consumer-friendly way -- similar to how sophisticated online music, movie, and shopping platforms do. Through the data marketplace mentioned above, users can visually see the origin and lineage of data sets just as a consumer can see background information about the musical artist of a song just streamed on Spotify. Through this visualization, users see consistency and relevancy in models across groups and teams -- and even ratings on data utilization just as we use Yelp for reviews.
Team commentary and patterns of data use dictate which models are most useful. Similar to sharing and recommending music to a friend, business users can collaborate and share data sets with others based on previous insights they've uncovered. This team-driven and "consumerized" approach to data discovery and analytics produces quick and reliable business results.
Tier 3: Augmented Analytics
A newer, more advanced feature of self-service analytics starting to emerge is augmented data insights: results based on machine learning and artificial intelligence algorithms. Using the Spotify analogy again, when augmented analytics is applied to the marketplace, data recommendations are made based on data sets the user has accessed, just as playlists are recommended to consumers based on songs they've listened to.
By automatically generating data results based on previously learned patterns and insights, augmented analytics relieves a company's dependence on data scientists. This can be a huge cost savings for organizations because data scientists and analysts are expensive to employ and often difficult to find.
By creating this fully integrated approach to how enterprises view and use their data, a natural shift will start to occur for the organization, moving from self-service analytics to shared business intelligence and "socialization" -- where all users across the enterprise are encouraged to contribute to and collaborate on business data for greater value and business advantage.
A Common Marketplace
Organizations that have started to make this shift are already starting to see business benefits. Similar to the consumer platforms of Spotify and Amazon, in an interactive community of trust, enterprise users thrive and are inspired to share and collaborate with others. It's through this collaboration that users gain instant gratification for more insightful decision-making. Through social features and machine learning, they learn about data sets they otherwise never would have known existed. Because analysts can see business context around technical data assets and build upon others' data set recipes and/or reuse models, they can achieve better, faster decision-making and work more efficiently.
As data complexity increases, the key to realizing the value of business data is pulling all of the different data management and analytics elements together through a common marketplace with a constant supply chain of business-ready data that is easy to find, understand, trust, and share. Only then does business data become truly intelligent.