What (actually) is self service analytics?

Why data-informed decisions are more important than broad access to data for self service analytics

What (actually) is self service analytics?
Photo by Noah Silliman / Unsplash

And why you need to think in terms of data-informed decisions.

Self-service analytics is a really broad phrase. It can cover both an aspiration that we've never quite reached, and something that we already do every single day.

The simplest definition I've seen is that business stakeholders aren't stuck waiting on the data team. If a marketing manager wants to understand sales this quarter they can look it up themselves without putting in a request to the data team to provide them with the info.


As an aspiration this is great — imagine a world where everyone could quickly find the data suitable for their needs. Even if we're not there yet, our (really data-driven) stakeholders will request data and wait patiently for us to put it together for them. And all we have to do is just keep making dashboards to cover each of their needs and eventually that queue of data requests will go away.

In reality that's generally not the case.

We end up in a continual state of partial self-service. To illustrate, let's look at our hypothetical marketing manager. The next time she's making a decision she could do one of these things

  • look at an existing dashboard to see the metrics she's interested in
  • read a data analysis provided by the data team and follow its recommendation
  • export data into Excel and filter / pivot to get what she needs
  • know SQL and query the data herself (yay)

But there's a risk she could do one of these instead

  • look at Salesforce reports since there's no dashboard available
  • make decisions by gut feel instead of looking at data at all
  • not be clear on which version of the various sales-related dashboards is most relevant and doesn't want to take the time to ask
  • know SQL and query the data herself (oof)

Self-service clearly isn't about stakeholders getting data themselves. If we build out a dashboard covering every possible scenario, it's not clear that it will even get used. A dashboard might not even be the right approach in the first place. We need a better way to think about this.

As defined above, self-service covers every scenario except a user directly submitting a request to the data team. The goal shouldn't be to leave the data team alone. We also need to think about data-informed decision making.

Data-informed decisions

A data-informed decision is also a super-simple definition — it's just a decision made using data.


Everyone at a company makes decisions. Those decisions are data-informed if (relevant) data influenced the decision in some way. The data could be in the form of dashboards, analyses, presentations, data products, raw data dumps, data support tickets, or anything else. The format doesn't matter so long as the data was made available in some way that could be properly understood.

Providing data to support decision making is the primary goal of the data team.

The concept of 'data-informed decisions' isn't exactly a real metric that can be tracked, but it's still a neat way of thinking about how a data team can have impact on the business.

As we saw in the section above, providing unstructured or unfocused access to data for everyone isn't data-informed and is therefore probably not very impactful. It's not a great definition for self-service.

A better definition for self-service

Self-service is the intersection of data-informed decisions and minimal data team involvement

self-service is when people in a company are making data-informed decisions themselves

Why is this useful? There are two truths buried in this diagram

  1. if your company isn't making data-informed decisions then you've got a problem
  2. the data team can't be involved in every decision

Let's say our marketing manager needs to decide whether to put more money in the current ad campaign she's running. She would strongly prefer to use data when making her choice. Because she has a great data team she can submit a ticket to ask for the customer acquisition cost of her campaign.

In this case we need to keep in mind that

  1. the data team is working on far more business critical things right now
  2. she needs an answer within 24 hrs
  3. this decision is one of dozens she'll make this week

We need to provide her with relevant, understandable data without a lot of friction or waiting. If she can get the data she needs in a few minutes, that's true self-service.

Self service in practice

In practice, providing this kind of self service analytics is tough. We'll talk about how to actually do it successfully in a future post.

In the meantime we'd suggest stepping back and thinking about what you're trying to accomplish with self-service. Focusing exclusively on data democratization — on ensuring stakeholders have access to data — will only cause problems.

Instead, try to increase data-informed decisions. Ensure that stakeholders have access to relevant data at the right time, even if it's a manual process that requires the data team to be involved. Invest in tools to make the data team faster and coach people on the right question to ask.

Check us out on the Data Engineering Podcast

Find it on the podcast page or stream it below