Empowering Self-Service Analytics: What They Forgot To Tell You

For today’s company drowning in data, the benefits of self-service analytics are clear. 

By turning their employees into citizen data analysts, these tools can offer much-needed visibility at the frontlines. It allows business teams to quickly identify opportunities and sidestep onrushing challenges, while their input and analysis will provide valuable intelligence for decision-makers. 

The pandemic has also made self-service analytics a must-have. For example, supply chain executives looking for transparency into upcoming shocks and current situations use self-service analytics to navigate. Likewise, warehouse managers looking to ensure the right stocks are always available online and at retail stores want better insights. 

“Analytics users are looking for rapid answers to sophisticated analytics questions. Analytics users prefer to solve their problems, and the larger organization would also prefer that users solve problems themselves,” says Joseph Antelmi, Gartner’s research director for analytics and BI. 

Jon Teo, Informatica’s healthcare & data governance specialist, sees self-service analytics and data democratization in the same light. It also points to the increasing appreciation of change management. 

“Self-service analytics (or Data Democratization) is an outcome, a result of change management (education, awareness) as much as about having the right tools and skill-sets. Readiness can be seen as a progression of different departments, roles or individuals in both their ability and demand to become more involved in using organizational data,” he says. 

However, some companies fret over putting valuable data before junior executives. Many fear that they will interpret wrongly or make poor judgments. Some are afraid of abuse and disgruntled employees siphoning off data to their competitors.

Informatica’s Teo believes these fears are understandable but points to other ailments. They include leaders not having the confidence in the current quality of their data assets, a belief that junior executives do not yet have the acumen to ask the right questions, nor interpret the data in a relevant way, and a general lack of trust in terms of data protection, risk of data misuse. 

“To be fair, an unguided, ‘open buffet’ to the organizations’ data assets is not likely to end well. However, experience from more agile organizations shows that providing the data autonomy and tools to make tactical decisions or even just generate insights bottom-up results in a tangible competitive advantage,” says Teo. 

This is why many see data governance becoming more critical for self-service analytics to work. 

Data governance becomes critical

Just ask AIA Singapore. The insurance giant wanted to personalize its service to its customers and sees this as an essential differentiator in an increasingly crowded and competitive market. 

But to personalize, it needed trustworthy data. AIA Singapore wanted to understand its data in context based on lineage and intelligent metadata. The problem: it had no data models available for metadata definitions. It also had no visibility into the end-to-end lineage, leading to consumers questioning where the information came from. 

The insurance giant also wanted to improve data quality to identify customer insurance gaps and make more effective proposals. But it did not have a standardized, reliable way to measure data quality, making it challenging to tackle reports about data inaccuracy.

AIA Singapore’s quest for better data governance led to establishing a Data Governance Council. It also deployed an end-to-end Informatica solution for data governance and discovery, starting with Informatica Axon Data Governance and later implementing Enterprise Data Catalog and Informatica Data Quality.

“Informatica Axon integrates with Enterprise Data Catalog and Data Quality to give us a complete solution for next-generation data governance,” said Ronald Chung, AIA Singapore’s section head for enterprise data, in a published case study. “For example, if we query for date of birth, we can see that definition from our business glossary, and we can also identify which system, table, and column that date of birth maps to. We can see the end-to-end lineage, including where people are actually viewing that data and where the points of entry are for critical data elements.”

Self-service analytics can’t be for loners

Self-service analytics works if the data sets are good, unbiased, and relevant to the problem statement. But the reality is much more sobering.

“You can’t guarantee that bias or poor quality data will be identified; however you can support your self-service analysts by building pipelines to clean and store data as best you can. The more ‘fit for purpose’ the data is, the less work your end users will have to do to prepare it for analysis and the less margin for error,” says Georgia O’Callaghan, Gartner’s senior principal analyst.

Informatica’s Two noted deploying self-service analytics should not be seen as the end game, nor does it make the data science team irrelevant. Instead, successful self-service analytics will need good collaboration between the data science team and the business users. 

“To use a supermarket analogy, a good start would be to ensure that any data ‘product’ put up for offer to data consumers are clearly labeled, meet a minimum standard of quality, and have clear recourse back to the ‘manufacturer’ or data producer. An intuitive organization of the data by domains as well as a powerful ‘google-like’ search never hurts either,” he says, offering analogies. 

Over time, data science teams can add new data science tools or even AI-based tools like recommendation engines or similarity algorithms.

“Building a collaborative data stakeholder community may be just as good or better because sometimes you just need to ask the right person or read the right comments on the data sets to solve this,” Teo adds.

But these efforts do add pressure to the data science team at a time when fatigue and impossible expectations are beginning to take their toll. 

“Successful adoption of self-service analytics naturally increases the demand for more data. This indirectly puts pressure on data engineering and DataOps to onboard more data sources as well as ensure that the data is curated to the expected levels of quality before being ‘published’ to the requestor or data consumers,” says Teo. 

 AI and machine learning can ease the pressure. “Greater efficiency in these processes will be required, and AI/ML techniques to automate data discovery, classification, as well as quality management processes can be a powerful part of the answer,” Teo explains. 

It begins and ends with trust

For self-service analytics to work, you need implicit trust. But there are three ways trust matters for the business user.

“It might be useful to differentiate between trust in the data being used, trust in the analysis or analytical models, as well as trust in the final insights (outputs),” explains Teo. 

The outputs are the easier part to address. There is already a range of techniques to validate the outputs of analytical or decision-making models. “Human sanity checking will always be valid in this context as well,” adds Teo. 

Teo believes it will take time when it comes to data inputs and analytical models. And it helps when there is greater transparency. 

“Trust is built up over time when there is greater transparency on the source, context, and measured quality metrics that come with the data set, as well as ‘social’ trust that can be increased through the feedback of other data users who have previously worked with the same data,” says Teo.

“The important thing is to find efficient ways to provide all these important trust cues without a lot of manual work, phone calls, and back-and-forth emails. In this regard, establishing a business-friendly data catalog and consumption portal goes a long way to bring these trust elements into a centrally accessible location,” he concludes.

Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].

Image credit: iStockphoto/Khosrork