From Spreadsheets to Dashboards: Getting Numbers You Trust

Decide which numbers deserve a screen

Choosing the software is the easy part of a dashboard project. The harder question is which numbers earn a place on the screen. Every extra tile is one more thing to maintain, and one more place a wrong figure can hide.

One filter does most of the work. A number belongs on the dashboard only if somebody would do something different this week because of it. Most owners can name four or five that pass. Cash in and out, orders in progress, how fast new enquiries get answered, and the current bottleneck. Adding a tile later is easy. Taking one away once people are attached to it is not.

• Who looks at it, and how often.
• What decision changes when it moves.
• Which system holds the raw figure, and whether software can read it unaided.
• The one sentence definition everyone agrees on.
• Whose name is on it the day it looks wrong.

Agree what each number means before anything gets wired

Sales counts a deal the day the buyer says yes. Finance counts it the day the money lands. Operations counts it when the job goes out the door. All three are reasonable, and all three give a different revenue figure for the same month. A number challenged once tends to get ignored from then on.

So write the definition beside the number, in the words your team already uses. Wobble's view is that the first week of a dashboard build should go on wording rather than wiring. A wrong definition wired perfectly is still wrong.

How do you know the dashboard numbers can be trusted?

Trust gets earned by comparison. Run the new dashboard alongside the manual report you already produce, for one full cycle, then compare them line by line. Where they disagree, the disagreement is the useful part. Usually it is a filter nobody wrote down, or a row somebody has been excluding out of habit. Keep the old process alive until two cycles in a row come out clean.

Then put a last updated time on every panel. A stale spreadsheet looks stale, and people handle it carefully. A stale dashboard looks live, and people act on it. If a feed breaks at two in the morning, the screen should say so.

When is a dashboard the wrong thing to build?

A dashboard installed over a broken process only makes the problem arrive faster. There are cases where waiting is the honest advice.

In each case the fix sits underneath the reporting. A dashboard shows you a process. It cannot repair one.

• The raw data is entered inconsistently at the source. Fix how it gets captured first, or you will simply see bad numbers sooner.
• Nobody has agreed what happens when a figure goes red. Without an owner and an action, a dashboard is wall art.
• The question is a one off, like whether to open a second branch. That is an analysis you do once, not a screen you maintain.
• The owner still knows every order by name. Build when the volume outgrows memory.

What a dashboard costs, who agrees the definitions, and what you keep

No price is published on this site, and on a dashboard the drivers are unusually easy to name. How many sources have to be read. Whether each will permit a read at all. How many numbers genuinely earn a screen. Who maintains the definitions when the business changes shape. Whether it is worth doing is settled by the section above about which numbers deserve a tile: five numbers a business trusts pays back, and fifteen nobody trusts is a maintenance job with a screen attached to it.

Do the definitions in-house first, whatever you buy afterwards. Writing down what a sale means, on one page, agreed by the three departments currently counting it three ways, is the whole project, and your own team is the only party who can do it. A supplier can wire a source to a tile. No supplier can decide whether finance or operations is right about the day a deal counts.

A dashboard reports and does not decide. Where a number moves far enough to matter, the useful behaviour is to hand it to a person with the underlying rows attached rather than to trigger anything, and any action taken off the back of a figure waits for human approval. The dashboard, the definitions and the connections sit in accounts under your own logins, so you own the system, and the parallel run described above can be repeated whenever somebody doubts a tile.

The published work behind this is the reporting layer inside larger builds rather than a dashboard sold on its own: Quillon's audited automation of 34 AI nodes, and the full operating systems at Emraan Rajput and THE MAGBOOK. Moiz Khan owns automation architecture at Wobble, which works from Karachi, bills month to month and carries 25 engagements across six countries. There is no dashboard-only engagement on the work page, which is worth saying plainly on a page about dashboards.