Is ‘self-serve’ the end goal for business analytics?

In the analytics industry, we talk a lot about “Big Data,” but the real value is in insights derived from that data. The data itself is hardly ever what actually matters. Business leaders realized long ago that whoever is the first to tap into accurate, up-to-date analytics would be the one to capture market share and cement customer loyalty. This need for speed pushed them to roll out access to business intelligence (BI) across departments and stakeholders.

First, companies expanded data science teams with new tools and more experts. Then they noticed that data scientists had turned into gatekeepers. This creates a bottleneck that deters stakeholders from seeking insights and limits the use of data for decision-making. In response, they started moving towards self-service as a BI goal.

Self-proclaimed self-service BI (SSBI) solutions have been flooding in since the early 2010s. Now, over a decade later, many people hail AI chatbot integrations into BI platforms as finally making data insights available to all and crossing the line into true SSBI. It’s a sentiment that I identify with, but with nuance.

Despite many enthusiastic predictions of instant adoption and BI utopia, adoption is slow. BARC has reported that even though, over several years, 15% of participants in their surveys have responded that they will implement self-service in the next 12 months, the percentage of participants actually using such tools remains unchanged at around 55%.

This gap between expectation and reality raises the question: is self-service really the be-all and end-all for business analytics? Based on my 17-plus years in the BI world, I have my doubts.

Why all the fuss around self service?
Putting cynicism aside for a moment, there are certainly legitimate reasons to embrace self-service. It democratizes access to insights and removes the bottleneck to data analysis. With self-service, all stakeholders can run analytics, answer queries, and understand their challenges, opportunities, and how to resolve or maximize them.

Once line-of-business users have better access to data insights, they can improve decision-making and strategic planning. They’ll come to data-driven conclusions and produce more creative and relevant solutions, making the organization as a whole more competitive.

Access to self-serve insights also eases pressure on data analysis specialists. They won’t have to keep jumping to answer queries, freeing them to do their own work and releasing them to drive the innovation that sharpens your competitive edge.

We’re even seeing significant steps towards making self-serve a reality. The marriage of GenAI and BI has produced BI self-serve chatbots, where business users can use natural language to ask queries without needing to learn to code, use MySQl, or know the best visualizations to request.

A new generation of tools can deliver the answers that users need, regardless of their level of understanding or proficiency in data manipulation. These solutions can explain the complex charts, graphs, and visualizations that business users receive but don’t know what to do with.

Why aren’t we in an analytics utopia?
As I mentioned above, this achievement hasn’t ushered in a BI rapture. The disappointment is mainly because decision-makers have been too distracted by the shiny new tool of self-service.

From where I sit, it seems that many haven’t given enough thought to the surrounding infrastructure and workflows needed to make that self-service usable and practical. Data literacy, data quality and governance, user adoption and trust, and security access are the specific obstacles that still bar the way to analytics utopia.

Line-of-business users lack data literacy
A deficit of analytical ability is a serious handicap in analysis, even with today’s self-service tools. I’m talking about knowing how to ask impactful questions, break them down into sub-queries that bring useful answers, and then apply them to the data at hand.

Either regular business stakeholders need to be fast-tracked to gain this data literacy, or we need to develop solutions that can bridge the data literacy gap for them. That means tools that can understand the requests, formulate them into appropriate queries, execute them, and then interpret the output in a way that is easy for the user to comprehend.

Data quality and governance falls by the wayside
Every business intelligence tool, whether or not it’s hailed as being optimized for “self-service,” requires rigorous data management and governance to maintain data trustworthiness. GIGO (garbage in, garbage out) still holds true for self-service.

However, a typical business user doesn’t know how to ensure data quality, and too many cooks can spoil the data broth.

Everyone is pouring in data and applying their rules and queries, but not enough know the basics of good data management and governance. The resultant poor data integrity and muddied data quality produce insights that no one can trust, not even data scientists.

Users need to trust and adopt the tech
In my opinion, this is an important step that’s been inexcusably overlooked: stakeholders have to actually use self-serve tools. That requires them to trust the system and the results, which in turn requires the results to be reliable and trustworthy.

Unfortunately, that’s not always the case. Hallucination is a known, serious, and common problem for AI tools. You’ll often need humans in the loop to verify the analysis and interpretation, which removes much of the benefits of self-service.

What’s more, it’s harder to generate reliable results when the users have poor data literacy. They’re more likely to formulate requests in ambiguous terms that are hard for the system to understand, setting up a vicious circle of slow adoption.

Security can be a double-edged sword
GenAI tools are already infamous for leaking data, especially third-party LLMs like OpenAI, and it’s crucial to protect sensitive customer information and proprietary data. Obviously, organizations need to set up security fences and limit access to sensitive data.

But at the same time, self-service users have to be able to access relevant data in order to get meaningful and useful results. If you cut them off too much, they won’t have the data necessary to generate the insights they need, and your self-service tools aren’t much use.

Where does true nirvana lie for business intelligence?
The truth is, the end goal was never self service. It was reliable, accurate data insights that business users can access independently and put into use immediately. The mistake was in thinking that this is the same as self-service. Data managers and data scientists will always be key pieces in the puzzle, laying the groundwork and empowering citizen analysts.

True nirvana is access to actionable insights, and that comes when organizations balance self-service with data governance, data literacy, and data security. Self-service can’t be allowed to override the importance of good data management and a central source of truth.

As CTO and co-founder, Avi Perez is responsible for creating, developing, and managing the Pyramid Analytics decision intelligence platform.