Self-serve analytics is a tough game to win at. Lots of companies aspire to “do self-serve” in order to reduce the load on their analytics and data science teams, but few succeed. This doesn’t mean that we’re hopeless. Our tools are getting better at enabling the self-serve experience, as we offer more intuitive and powerful building blocks.

The Building Blocks

Most BI tools today operate on a spectrum of building blocks. Some like SQL are flexible enough to tackle just about any problem you want. Others like dashboards skew more towards usability and lowering the barrier for teammates. The name of the game is finding the right level of abstraction for your given problem. Let’s go through the full range, from small blocks to big ones.

SQL with raw data

Usability: ⭐️  Flexibility: ⭐️⭐️⭐️⭐️⭐️

As I alluded to before, writing SQL is just about the most flexible tool out there for analysts and friends to answer questions about data. If you know it, great. You can effectively serve yourself. If you don’t, no dice. No tooling required here aside from a database.

Visual query builder with raw data

Usability: ⭐️⭐️  Flexibility: ⭐️⭐️⭐️⭐️

We are still working with raw data here but now we aren’t limited by knowing how to write SQL code. We can instead use some sort of visual query builder, which has become table stakes for BI tools. Note that there’s still a large bottleneck here around what raw data to use and how to manage the nuances of it. BI tool required.

Visual query builder with modeled data

Usability: ⭐️⭐️⭐️  Flexibility: ⭐️⭐️⭐️⭐️

So how do you address the aforementioned bottleneck? You model your data. Modeled data is more intuitive and usable. The kinks have been ironed out. With the visual query builder and modeled data, you start to achieve some semblance of self-serve analytics for non-analysts. Progress!

Visual query builder with metrics

Usability: ⭐️⭐️⭐️⭐️  Flexibility: ⭐️⭐️⭐️

With modeled tables, you still run into the problem of needing to piece together logic for answering common business questions. There are an endless number of ways to measure even seemingly simple metrics like “Weekly Visitors.” Instead of putting this task on self-serve analytics users, data teams can instead model out a set of canonical metrics so everyone uses the same definition and doesn’t have to do the legwork themselves. The catch: Not every question fits nicely in a pre-baked metric. The end user is trading off usability for flexibility when they choose this building block.

Dashboards

Usability: ⭐️⭐️⭐️⭐️⭐️  Flexibility: ⭐️⭐️

You know I couldn’t forget about the humble dashboard. In this case, the work is already done by an analyst ready to be consumed by others. There may be filters and date pickers wired up to enable some slight flexibility but in general, what you see is what you get. Dashboards are great for some situations but limiting when it comes to the breadth of self-serve analytics.

What’s Next?

Just to be clear: I’m not saying that any of the above building blocks are inherently better than others. Remember that it’s a spectrum for a reason. Depending on your flexibility and usability needs, you should pick the right block for the problem. However, be aware of the tradeoffs that come with your choice. As is often the case, there’s no free lunch.

But it’s interesting to explore what’s next. Are we stuck with the existing spectrum or is there a path to a slightly-more-free-lunch that we’re used to? I think so, and see two paths forward. Both exist today in some form, but are still relatively nascent as mediums.

Data Apps

Usability: ⭐️⭐️⭐️⭐️⭐️  Flexibility: ⭐️⭐️⭐️

Data apps can be a bit tricky to get your head around, but imagine a more powerful dashboard. A dashboard that is wired up in new ways that enable more breadth in the questions you can answer. Hex is one company that has explored this rabbit hole as of late. I'm very bullish on data apps, but there's still lots of work for us to do here.

NLP

Usability: ⭐️⭐️⭐️⭐️⭐️  Flexibility: ⭐️⭐️⭐️

And then there’s natural language processing applications. The workflow most of us imagine is you type your data question into a search bar and get back an automated answer. I’ve written before about my thoughts here. I’m more skeptical of this as a catch-all solution and more intrigued in the types of questions it can handle as a low-touch building block in the self-serve tooling spectrum.

Wrapping up

That’s the lay of the land as I see it today. But who knows, the data space has seen some really neat innovation and lots of fresh ideas over the past few years so I suspect there will be other directions explored. Regardless, I’m excited for a world with building blocks that better fit what I’m trying to build. I suppose Legos left an impression.

Thanks for reading! If you enjoyed this post, you can subscribe to get future ones like it delivered straight to your inbox. If Twitter is more your speed, you can follow me there as well. 🔥