Data-driven companies are finding that a sound self-service BI strategy can play a critical role in the digital transformation process and is key to bringing disparate silos of corporate information together.
Historically, different departments have worked from their own silos using legacy reporting methods that don't provide an end-to-end view of the data across their organization. For example, the absence of standard key performance indicators (KPIs), definitions and a centralized data warehouse had led to inaccurate reporting at AirAsia, said Aireen Omar, deputy group CEO for the company's digital, transformation and corporate services.
Harmonizing different silos or eliminating them altogether comes with many challenges. One of the biggest is identifying the source of the data and standardizing KPIs across the board. "Enabling self-service BI requires the right tools, certified data sets and strong data governance policies to be in place," Omar said.
Find common ground
Quick access to information that resides within various internal business applications and databases is paramount. Massive amounts of structured and unstructured data often reside within multiple and disconnected platforms, applications, locations and devices. "This creates compliance and security risks and also leaves employees with only some of the information they need to perform their jobs," said Matthew Meigs, director of marketing communications at content service platform maker Nuxeo.
Patchwork solutions to data silos can be problematic, especially for companies creating a new data integration layer from scratch. Traditionally, consolidating data silos to accommodate a self-service BI strategy required ripping out and replacing existing systems, which can entail high implementation costs. Many companies are creating an intelligent middle ground that enables modern solutions and legacy systems to coexist without embarking on complex migration projects.
Overcome cultural hurdles
Employee resistance to change and manager perceptions of complexity in moving data to a new repository are among the biggest stumbling blocks to creating a self-service BI strategy, according to Dave McCandless, vice president of IT at shipping supply chain app provider Navis.
Rob PerryVice president of product marketing, ASG Technologies
McCandless said he's pushing data science as the new research lab in his company, as well as for its customers. That means selling corporate leaders on the value of creating a data lab staffed by subject matter experts tasked with assembling information on the greatest challenges facing the company.
Many organizations don't provide incentives for sharing information across business units, which "generally have very limited budget allocated toward working with other business units," explained Rob Perry, vice president of product marketing at IT service management provider ASG Technologies. "Sharing information across departments is not a priority, especially given [that] managers are not measured on how well other business units have done."
Instead of eliminating data silos, Perry recommended that managers focus on data sharing that can improve self-service BI. He suggested working from the top down -- issuing mandates that encourage and even gamify information sharing using new kinds of incentives. Another approach, he added, would be to hire a chief data officer with the directive and budget to bring business units together, including the creation of cross-functional teams to manage data domains.
Aggregate smaller analytics cases
A self-service BI strategy can make it easier to explore a wider variety of analytics -- sometimes referred to as sidecar analysis, since it's outside the normal course of analytics. Often too small to automate, this type of analysis can help guide certain business decisions.
"To address these sidecar analytics, we often advise our clients to inventory the types of analytics they are performing and look for projects to automate a collection of the smaller analytics to provide the ROI needed for the automation investment," said Andrew Roman Wells, CEO of Aspirent, a management consulting firm.
Nathan Kollett, former senior manager of web analytics at Wayfair, offers five dos and don'ts for setting a self-service BI strategy.
Keep it intuitive
Desktop or server tools that don't have centralized management are a large contributor to data silos. They're difficult to scale and manage across the organization and often lack IT approval.
"In our experience, we've seen IT successfully counteract these solutions, not necessarily by enforcing stricter policy, but by offering an alternative solution to the tools that are currently being adopted," said Sean Kandel, co-founder and CTO at Trifacta, a data wrangling software provider.
Self-service BI sometimes straddles a delicate line between adoption and governance. Business users want their tools to be intuitive and easy to use, while satisfying IT concerns over security, governance and scalability.
According to Sreeni Iyer, CTO at cognitive sourcing service provider LevaData, data cataloging is an emerging field to help companies catalog all their data sources and metadata. In addition, data wrangling is becoming a key element for a sound self-service BI strategy, enabling staff with no programming skills to visually define transformation, cleanse data and integrate across silos.
Simplify data modeling
Dominic GoDirector of analytics, Olivela
As if silos aren't isolated enough, another type of data silo can arise as a result of the different queries put forth by users. "The issue is that people often view data as objective truth when, in reality, the translation of data into meaningful information can be entirely dependent on the question being asked and the business need behind the question," observed Dominic Go, director of analytics at luxury retailer Olivela. Business users, for example, will typically consider their own business problems, not companywide issues, when interpreting specific data.
Thanks to self-service analytics environments, this "infallible" data is more accessible than ever. "This can lead to deeper divides and mistrust between departments," Go reasoned. "And when we can't trust data anymore, what's next?"
He believes an underlying problem is the data modeling layer, which sits above the data warehouse and just below the reporting UI. "[T]he business user often does not have a say in how those interpretations are being made by the IT team setting up the infrastructure," Go explained. "A poorly defined data modeling layer can institutionalize a poor interpretation with the guise of infallibility, leading to more organizational divides and poor business decisions."
As a result, Go has focused on streamlining and possibly eliminating the modeling layer for self-service BI tools and still allowing easy data access. "It's risky," he acknowledged, "but I believe it's the next step in bridging [or] eliminating communication silos, while still being able to use data effectively to grow our businesses."