There has recently been a lot of chatter in the data community about dashboards. Most of which converge on some version of “dashboards suck but what else,” as Benn Stancil points out in Data is for Dashboards:

Ask ten analysts a question, and you’ll get eleven opinions — unless your question is about dashboards. On dashboards, the consensus is universal: They’re bad. Dashboards are outdated, brittle, poorly built, and rarely used. Even our defenses of dashboards are lukewarm.

This resonates with me. I’ve got more than a few stories of times when dashboards let me down. But I also think it’s important to take a step back and think about the jobs that dashboards are enlisted to do in the first place. Only then can we 1) decide if they suck 2) explore how to improve them. So let’s get into it.

Data: what’s the point?

If you were to ask me “What’s the point of collecting data?” then almost immediately, I would shout out something along the lines of “make better decisions.” I suspect I’m not the only one that links these two ideas together. Data can be turned into more information which can be turned into more informed actions. Math checks out.

But let’s get more concrete. How exactly do we use data in order to make better decisions? I see a few main buckets of data interactions:

  • Reporting: Scheduled, recurring interactions with select data. Reporting gets a bad rap, but it’s quite useful. It helps you ensure that things are going according to plan e.g. “Has anything changed more or less than we would expect?” More importantly, reporting also helps build a shared mental model of the business with a group of people e.g. “Our funnel looks like this and these are the areas we care about.”
  • Analysis: Ad-hoc deeper dives into a particular question. Maybe the question is “Why did this metric move last week?” or “What will happen if we roll out this product change?” or something more obscure. Analysis is often performed “just in time” to make a decision or fill a gap in the organization’s mental model of the business. There’s also a big variance in how deep you can go. Maybe it’s a one-hour detour. Maybe it’s a 2-week rabbit hole.
  • Automation: Often the best way to make better decisions is to automate them. This happens once you have confidence around your answer to a question. Things like “Should we reach out to that customer?” shouldn’t be analyzed on a per-case basis. You form a general view and automate the flow based on logic. I won’t talk too much more about this for the rest of the post, but I’m keeping it here for completeness.

Back to dashboards

So where do dashboards fit into this story? For most organizations, dashboards are enlisted to handle the “Reporting” use case described above. This feels right to me. Good dashboards help you 1) monitor select metrics and 2) communicate a simplified model of an area to others. From my former colleague Andrew Bartholomew in Good Dashboard, Bad Dashboard:

Good dashboards are opinionated. A good dashboard communicates, "These things matter; these other things don't." A bad dashboard throws everything on the page and forces the user to decide what is important.

So dashboards aren’t inherently bad. But if you think about it, what are dashboards really? They are a collection of charts and metrics; simply a medium to organize and showcase data in a useful light. If we had a better medium then we would use that instead. But we don’t.

After many long nights, I have no choice but to declare that the medium isn’t the issue. It's the implementation of said medium; and it all starts with tooling. Could our tooling better enable dashboards to do their job? You betcha.

Dashboards are documentation

Today, dashboards don’t effectively interface with practical business knowledge. They suck at embedding and documenting new information from others: conversations about metrics, questions that arise, notable events that occur, and further analysis conducted. These are all pieces of information that pop up and then disappear into the ether. Instead, they should be embedded within the dashboard itself.

All the day-to-day reporting stuff feeds into making the dashboard a more complete representation of reality. Dashboards aren't a static collection of charts. They are living, breathing documentation that should be maintained and built upon over time.

I don’t know the best way to do this, but I do have some ideas. Just for fun, here are a handful of features that might make dashboards a little better at being a source of truth. Note that some of these features exist already but aren’t adopted completely across the board yet:

  • Events and historical context: Circumstantial business knowledge e.g. “This was when we launched our marketing campaign”
  • Comments and threads: Past conversations and questions e.g. “Why did we not see a dip in WAU over New Years?”
  • Links to in-depth analysis: References to deeper analysis on specific questions e.g. “Click here for investigation into this decrease.”
  • Changelogs and previous states: Visibility into how a dashboard evolved over time e.g. “Chart X was last updated on 1/12/22.”
  • Relationships between metrics: Definitions of how some metrics are related e.g. “New Users = New Signups x Account Creation Rate.”

All together, tooling with these additional features would be pretty sweet. You could document and maintain events, conversation, questions, relationships, and references to analysis. Dashboards go from this static collection of charts to a document that grows in value over time as more information is embedded in it. Life is good. But there’s always a catch.

Most of these half-baked ideas above require someone to manually input information. And generally speaking, people suck at documenting things. There’s definitely some more thinking to be done on how to automate as much of this as possible, or at least make it lightweight and close enough to natural usage patterns that it isn’t “yet another thing to do.”

That ain’t an easy problem to solve. But I think it is solvable. And if we can figure it out, then maybe dashboards end up sucking a little less and becoming a little more useful.

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. 🔥