One screen, two conclusions.

What a dashboard actually is

Everything in this track ends up on that screen.

Quantities from the site. Hours from the timesheets. Cost from a ledger that closed on its own date. Drawing status from document control. Deliveries from the store. Approvals from the client. Twenty-five weeks of records, from seven or eight departments, reduced to a page of tiles.

That reduction is the point of it and it is also the problem. A dashboard shows the output of a data model. It doesn't show the model, and it certainly doesn't show where the model is weak.

Why two readers diverge

The divergence has causes and they are all upstream.

If two departments calculate the same indicator differently, the tile shows one of the two versions and doesn't say which. If cost closed on the fifteenth and progress on the twentieth, the two tiles beside each other describe different months. If a quantity was estimated rather than measured, it looks exactly like one that was counted

. If nobody owns a figure, nobody can be asked what it means.

Each reader then fills the gaps from what they know. One has been in procurement all week and reads the screen through the deliveries. Another remembers a conversation about rework. Both are reasoning correctly from the same display and arriving somewhere different, because the display carries the numbers and not what is behind them.

That is the failure this week is about, and it is not a design failure.

Changing the screen changes nothing

The instinct when this happens is to improve the dashboard. Add a chart. Change the thresholds. Show a different indicator. Somebody suggests a different tool.

None of it works, for a reason that is obvious once stated: the disagreement was never about presentation. Two people who don't share a definition of complete won't come to share one because the tile turned a different colour.

Which is why the response to a dashboard argument belongs at the input. Which indicator, calculated how, from which record, closed on which date, owned by whom. Answer those and the screen stops being contentious, usually without anybody touching it.

What it can and cannot do

A dashboard cannot align an organisation. If the organisation is already aligned — shared definitions, agreed cut-offs, owned figures — it makes that alignment visible and useful. If it is not, it makes the disagreement look professional

.

That is worth stating because the opposite is widely believed. Projects buy tools expecting them to produce agreement, and what arrives is a better rendering of a disagreement that was already there.

There is one thing a dashboard does that nothing else does, and it is not the thing it is bought for. It puts every department's output on one page at one moment, which means inconsistencies that were survivable while they lived in separate documents suddenly sit next to each other. A project that has never reconciled anything finds out on the day the dashboard goes live.

The version that works

A dashboard is defensible when three things are true of every tile on it, and they are all upstream of the tool.

The indicator has one definition, written down, used by everybody who produces or consumes it. The data behind it closed on a date the reader can find. And somebody owns it, meaning there is a person who can be asked what it means and who will answer the same way next month.

Tiles that fail any of the three should not be there. A page of six figures that meet all three is worth more than twenty that don't, and it is a shorter argument than any redesign.

Where this leaves the track

Twenty weeks of this have been about single records: what a quantity means, when a ledger closes, which revision is current, who owns a number.

The screen is where all of it becomes visible at once, to people who were not involved in producing any of it. Which makes it the honest test of whether the work was done — not of whether the dashboard is any good.

System design

Nothing here is about the tool. These are the three things that have to be true of a figure before it is worth putting on a screen at all.

Record	Produced by	Required quality	Verified against	Feeds
Indicator definition	Project controls with the owner	One sentence a second department would agree with	The producing department	Every tile using that figure
Cut-off date behind the tile	Project controls	Findable by the reader, not implied	The reporting calendar	Whether two tiles are comparable
Figure owner	Named person	Someone who will answer the same way next month	The data ownership register	Any challenge to the tile
Measured or estimated	Producing department	Marked, because the tile cannot show the difference	Corroborating records	How much weight a reader gives it

A tile failing any of the four is one that two readers can legitimately interpret differently. Removing it is a shorter argument than redesigning the page.

Practical insight

Take your dashboard and pick three tiles. For each one, answer three questions without looking anything up.

What exactly does this measure, in one sentence that a second department would agree with? What date is the underlying data closed to? Who would you go to if it were challenged?

Any tile where you cannot answer all three is a tile two readers can legitimately interpret differently. On most projects that is more than half the page, and it explains the meeting.

Then do it once more, with somebody from another department, on the same three tiles. Where their sentence differs from yours, you have found the reason two people can leave the same review disagreeing.

Key takeaways

• Two people can read the same dashboard correctly and reach different conclusions.
• A dashboard shows the output of a data model. It doesn't show the model, or where it is weak.
• Different definitions, different cut-offs, estimated figures and unowned numbers all look identical on a tile.
• Readers fill the gaps from what they happen to know, which is why they diverge.
• Improving the display cannot resolve a disagreement that was never about the display.
• A dashboard cannot align an organisation. Aligned, it makes that visible; unaligned, it makes the disagreement look professional.
• It does put every department on one page at one moment, which is where a project finds out it has never reconciled anything.
• A tile is defensible when it has one written definition, a findable cut-off date, and an owner who will answer the same way next month.

Records born here. Indicator definition list · the dashboard specification · the tiles removed, and why.

What is coming next

If a tile has to have one definition and one owner, the next question is which tiles are worth the page at all.

Next week: leading and lagging indicators — the ones that describe last month, the ones that describe the next, and the one that always lands just under target.