We have to fundamentally rethink the idea of "self-service" in the world of data. Dashboards aren't self-service; they're just basic service.
Terms like "democratize data" or "self-service analytics" are widely used today. The idealized image they conjure is that anyone in the organization can easily access all the data or analytical information they need with just a few clicks, without needing to ask engineers or data teams. But the reality is quite different.
First, data consumers are usually “served” dashboards - a veritable swamp of precomputed datasets presented as tables or visualized in colorful charts. These dashboards do capture business metrics, but the consumer is actually interested in the "why" and "what's next." They want to understand the drivers behind changes in metrics week over week, broken down into specific segments and their contributions, or to quantify the system-wide impact of a new feature launch, operational process change, or small growth experiment.
Even for these simple use cases, the consumer has to sift through several dashboards and hope that all the data is there at the right time grain, time windows, level of aggregation, with all the correct segments mapped to the correct metrics in order to successfully complete their task. A simple question can turn into a full day or multi-day odyssey of navigating the dashboard swamp. It's no wonder that the consumer finds it easier to send a Slack plea for help to their data team.
In short, dashboards aren't self-service; they're just basic service. And because they are just datasets created at different points in time for specific needs, they are not the right foundation for self-service. True self-service is about empowering the user to explore the entire system if needed, but also to get precise solutions to common use cases quickly and efficiently.
True self-service is also not about a no-code solution, which is an approach touted by vendors. Learning SQL, the lingua franca of data, isn't the main obstacle to self-service analytics. The real challenge is in taking the consumer's mental model of their business processes and associated questions of interest, and translating them into the data world of two-dimensional tables, joins, filters, and aggregations.
So, what if we started with the business processes, use cases, and common questions, and then worked backwards to identify abstractions, the building blocks that users can flexibly assemble and reassemble? This would represent a fundamental shift not only in analytics tooling but also in how work flows in organizations. Instead of BI or data teams reactively building more and more reports and dashboards, their work would shift to defining the building blocks, leaving software modules to take on the heavy lift of helping consumers effortlessly assemble and reassemble these building blocks.
This paradigm shift is essential if we want to truly empower data consumers. Simply serving data within current BI tools isn't going to cut it, and it certainly isn’t “self-service analytics”.