KPI dashboards: when a screen earns its place

It is a display: tiles, charts and current values, usually pulled automatically and refreshed without anyone touching it. That is genuinely useful. What it is not is a management habit. Nobody has ever been held to a number because it was rendered nicely, and a screen with nobody named against a row is decoration with live data in it.

The distinction matters because the two get bought as if they were the same purchase. A team that wants to manage by numbers buys a KPI dashboard, spends six weeks connecting sources, and ends up with eighteen tiles that nobody opens on a Tuesday. The numbers were never the missing part. The missing part was somebody saying on track or off track out loud with five colleagues listening.

Put the other way round: a company with a weekly reading habit and a shared spreadsheet is managing by numbers. A company with a beautiful screen and no meeting is not. If you only have budget for one of the two, the meeting is free.

A dashboard wins when the question is what is happening right now and who needs to react today: support queues, uptime, orders in the last hour, cash in the account. The weekly table wins when the question is whether the business is on plan, because that needs a small number of rows, a person against each one, and a moment where somebody answers for the red ones.

The clue is the reaction time. If the right response to a number moving is within the hour, it belongs on a screen somebody watches. If the right response is a discussion and a decision, it belongs on a table read once a week, where five to fifteen rows and a trailing view give you the pattern rather than the noise.

Both can be true for the same business. An operations lead can watch a live queue all day and still bring one weekly row about the same function to the leadership meeting. What breaks is putting the live queue in front of the leadership team, who then spend twenty minutes reacting to an hour of data.

Buy when your data already lives in tools that the product connects to and your rows are ordinary: revenue, pipeline, tickets, hours. Build when the number that matters most is specific to how your company works and no connector will ever produce it. Most companies of twenty to a hundred people should do neither for the first quarter and use a spreadsheet.

The reason for the delay is that rows change. A first draft of a weekly board typically loses two or three rows by the end of the quarter, because a measure that sounded important turns out to be unactionable or gameable. Automating a row costs real engineering time, and automating one that gets deleted in six weeks costs it twice.

When you do buy, check the boring things rather than the demo. Whether the tool can express your definition of a number rather than its own. Whether a row can carry an owner's name. Whether last quarter's figures survive a source system changing. Published prices sit on the vendors' own pricing pages and change often enough that any number quoted here would be stale by the time you read it.

Too many tiles, so nothing is the point. No owner, so nobody answers for a number. No target, so a figure is neither good nor bad. Vanity rows that only ever go up. And no cadence, meaning nobody opens it on a fixed day, which turns the whole thing into decoration that took a quarter to build.

The owner problem is the one worth fixing first, and it costs nothing. Put a person's name next to every row, agree that the name means answering for the number rather than producing it, and half the other problems become visible immediately. Rows nobody will accept ownership of are usually rows nobody can influence.

Targets are the second cheap fix. A number without an expectation cannot be read: eleven is fine or a disaster depending on what was supposed to happen. Rough targets set on the second Tuesday and corrected a month later work far better than precise ones that arrive never.

Two or three things per audience, not everything you can connect. A leadership screen carries the same rows the weekly table carries and nothing else. A functional screen carries the live numbers that team reacts to today. A screen for the whole company carries the two or three figures everyone can influence, and it is the one most worth having on a wall.

The company screen is the one that gets skipped and returns the most. Two numbers that everybody in the building affects, updated where people walk past them, do more for shared attention than a suite of departmental views nobody outside the department opens.

Keep the definition of each number written down next to it. Half the arguments in a review are about what a figure counts rather than about the business, and a one-line definition under a tile ends most of them before they start.

Spreadsheet first, for a quarter. Read it out loud every week with a person against each row. Delete the rows that turn out to be noise, keep the ones that survive, and only then pay to automate what is left. Teams that run this order end up with fewer tiles, better rows, and a habit that would survive the tool being switched off.

Manual collection has a hidden benefit for the first few weeks. Somebody typing a number in every Monday finds out quickly which sources disagree with each other, and that argument is worth having before a connector quietly picks one of them for you.

In EOS® the weekly table is called the Scorecard and it is deliberately low technology, a short board read at the start of the leadership meeting. Automation is a comfort rather than the point. The point is that a person says a number out loud and the room decides whether to do anything about it.