Why no BI tool tells you
what a decision was worth

A dashboard says gross is down eleven percent in the Northeast. Somebody calls a meeting, a change gets made, and six months later nobody can say whether it worked. This is not a tooling gap anyone talks about.

The meeting that has no ending

Here is a sequence that happens in every business with dashboards. A number moves. Somebody notices. A meeting is called, opinions are exchanged, and a change is agreed: a price floor in forty stores, a new discount tier, a route change.

The change goes live. Everybody moves on to the next number that moved.

Nine months later, somebody asks whether it worked. And there is no answer, because the metric moved for four reasons at once, nobody wrote down what they expected before the change, and nobody recorded which stores it applied to. The honest response is a shrug with a chart attached.

Why the tools stop where they stop

This is not an oversight. Reporting tools were built to answer questions about the past, and they do it well. Measuring a decision needs three things a reporting tool does not have.

A metric the whole business shares. If finance and operations compute gross margin differently, there is nothing stable to measure against. Any verdict is arguable.

A prediction, recorded before the fact. Without it, the result is judged against whatever people remember expecting, which is reliably generous after the event.

A comparison group. The number would have moved anyway. Seasonality, the market, four other changes. Without knowing what happened to the units you did not touch, you cannot separate your effect from the weather.

What recording a decision properly requires

A proposal has to name the metric it expects to move, by how much, by when, and over which units. If it cannot name both a metric and a scope, it is an opinion, and opinions cannot be checked later.

That constraint sounds bureaucratic and is the opposite. It is the thing that stops a decision register filling up with entries nobody can ever evaluate, which is what happens to every attempt at this in a spreadsheet.

Then the expectation gets locked at approval. Not because people are dishonest, but because memory is not a reliable record and everybody adjusts their recollection to fit the outcome.

The part that builds the most trust

Sometimes the answer is that you cannot know.

If a second change touched the same stores in the same window, the two effects cannot be separated. Any number you produce is a guess wearing a decimal point. The correct output is to mark it unmeasurable and leave it out of the totals.

A portfolio that quietly includes results it cannot defend is worth less than one that admits the gap, because the second one can be believed.

What it looks like when it works

A price change went live in forty stores on the third. Nine weeks later: predicted plus three percent, actual plus two point one, the ninety-six untouched stores moved plus nought point four, so plus one point seven is attributable. Volume fell nought point nine percent over the same window, which nobody predicted and which is reported next to the win rather than buried.

That is not a chart to interpret. It is a verdict, and the next proposal is better because of it.

Why this is hard to add later

Because it depends on the semantic layer. A decision can only be measured against a metric the platform computes the same way every time. If three teams compute gross margin three ways, there is nothing to measure against, and no amount of decision-tracking software fixes it.

Which is why this arrives as part of a platform rather than as a product you can buy on its own. We build both halves because the second half does not work without the first.