Dashborad Wars
Dashboard Wars
AI has made it easy to create dashboards — anyone can spin one up with the right access and a set of prompts.
But are all dashboards created equal?
Numbers mean very little until there is meaning and reasonable degree of trust behind them. If your organization has spent considerable time investigating discrepancies between metrics reported through different channels, or has faced a situation where no one knows the right answer but most people can kind of tell you what the issue is, this might sound familiar.
What happens when two dashboards reporting the same metric cannot agree on the numeric value? If the numbers are different, how much difference is acceptable? What happens when someone asks why the numbers are different?
There is usually a good explanation for these differences. The data source is different, or the definition is different. One system has latency, so it underreports. The metric in Dashboard A counts three edge cases, while the one in Dashboard B counts five. Who is right?
The problem is not the code. Code is the easy part in all of this. The problem is much more nuanced and messier — it is about alignment. Not spiritually, although that never hurts. I am talking about organizational alignment.
Alignment on the definition of a metric, which edge cases to consider, how much uncertainty is acceptable, what limitations exist in the systems measuring it, what the measurement actually represents, and how it will be used.
The problem is also data. Data is a highly malleable asset. Anyone can take the numbers, apply whatever bespoke logic makes sense to them, and come up with a metric. Getting twenty-five people across six teams to align on that metric definition, however, is a whole other story.
Here is how well-meaning dashboards sometimes mislead.
An analyst creates a time series of “active” customers for monitoring. The moment there is a large movement in that daily volume, the dashboard that was created to make the analyst’s life easier suddenly has dozens of eyes on it. People start arguing about the filters, joins, and definitions. Usually, these situations end with a completely different understanding of what “active” means than what we started with.
That is a problem.
Data is also political. People want ownership and a claim to expertise over the data they produce, ingest, enrich, or modify for a business use case. Anyone else’s attempt to standardize and streamline metrics and data can feel like an infringement on someone else’s territory — every team thinks they know their data best.
Get Javidsidra’s stories in your inbox
Join Medium for free to get updates from this writer.
The consumers and producers of data also care about different things. They optimize for different outcomes. So, a metric that means one thing in one function can mean something else entirely in another context.
An active customer for engagement is different from an active customer for revenue. Neither definition is necessarily wrong, but over- or under-reporting those numbers without agreement on what they represent can create very real friction.
Have you started to see the problem yet?
So, whose dashboard is better? The domain expert’s or the dashboard expert’s? The technical genius’s or the person reporting to the executive who is on the hook for the numbers? Whose number is better? And what does “better” even mean?
If two dashboards report the same metric differently, which one should you trust?
I wish the answer to this were a few neatly tied sentences that made perfect logical sense. Sadly, that’s not realistic. Data problems are expensive and messy, and consistency across an organization requires carefully designed processes, implemented over time, with people actually willing to use them.
I think of trust as a spectrum across three broad areas:
Technical trust: Can the data be relied on operationally? This includes monitoring, anomaly detection, access controls, resilience, and clear escalation paths when something breaks.
Semantic trust: Do we agree on what the number actually means? This includes metric definitions, calculation logic, lineage, documentation, known limitations, and clarity on which source should be used when the same metric can be calculated in multiple ways.
Organizational trust: Do we agree on who owns the metric and who has the authority to make decisions about it? This includes ownership of data assets, alignment across teams, accountability when definitions conflict, and clarity on who people should go to when something goes wrong.
These are not separate categories. They build on and reinforce one another. A metric can be technically sound but poorly defined, or perfectly documented but disputed across teams. Trust comes from having enough strength across all three.
That clarity is what allows you to say what “sales” means and, more importantly, how confident you should be when reporting that sales are up 20% YoY.
So, which dashboard is better?
It is the one built on data that has been validated, discussed with stakeholders ad nauseam, and where every metric can be traced to a clear owner who can explain it when, one day, it is suddenly down 45%.
