BI Is Decision Infrastructure, Not Decoration

There's a moment in many BI programs where the dashboard count quietly passes the decision count, and nobody notices because both numbers feel like progress. More dashboards looks like more insight. It usually isn't. It's more surface area to maintain, more places for a number to be subtly wrong, and more reasons for someone to say "which one do I actually look at?"

The reframe that fixes this is in the word infrastructure. We don't celebrate having a lot of bridges. We care whether the bridges that exist hold weight, go somewhere people need to be, and don't fall down. BI deserves the same standard.

Decoration vs. infrastructure#

Decorative BI is built to be shown. It looks comprehensive in a review, it has a lot of charts, and its primary audience is whoever approved building it. Infrastructural BI is built to be used. Someone makes a real decision because of it, on a real cadence, and would notice — loudly — if it broke.

The tell is what happens when a dashboard goes down. If a chart breaks and the only person who notices is the engineer who maintains it, it was decoration. If it breaks and someone who makes decisions is blocked within a day, it was infrastructure. Most organizations have a lot of the former and a little of the latter, and they spend their maintenance budget evenly across both.

The usefulness chain#

I run any dashboard — proposed or existing — through a chain of six links. It's only as useful as its weakest link.

1. Decision. Is there a specific decision this supports? "Visibility" is not a decision.
2. User. Is there a named person who makes that decision and will look here?
3. Action. Can they do something different based on what they see? If every value leads to the same behavior, the chart is theater.
4. Explain. When the number moves, can they explain why without filing a ticket?
5. Drill. Can they get from the headline to the records underneath it?
6. Workflow. Does looking here fit the rhythm of their week, or is it a place they have to remember exists?

Decorative dashboards almost always pass link one — a decision exists in theory — and fail somewhere in two through six. They have no named user, or no possible action, or no path from the number to an explanation.

A metric with no decision attached is not a metric. It is a number wearing a costume.

Retiring is a feature#

The reason decorative BI accumulates is that building a chart is easy and removing one feels rude. Someone asked for it. Someone might be using it. So it stays, and the next one stays, and the surface area grows until no one trusts any of it because there's too much to keep correct.

Treating BI as infrastructure means treating retirement as legitimate engineering work. A dashboard that fails the user or action link is a candidate for deletion, and deleting it makes the rest of the system more trustworthy by concentrating attention and maintenance where decisions actually happen. I've never regretted removing a dashboard nobody could attach a decision to. I've often regretted keeping one.

What infrastructure-grade BI feels like#

When BI is genuinely infrastructural, a few things are true. There are fewer dashboards than you'd expect, and each one has an owner who'd be paged if it lied. Every headline number can be drilled into until it bottoms out in real records. Freshness is visible, so nobody acts on stale data without knowing it's stale. And the metrics that matter are defined once, somewhere canonical, so two people pulling "active accounts" get the same answer — because nothing erodes decision infrastructure faster than two numbers with the same name and different values.

None of this is glamorous. Infrastructure rarely is. But the goal of business intelligence was never to be impressive in a slide. It was to be the thing a decision can safely rest its weight on — and that's a standard you can actually hold a dashboard to.