How many dashboards does your organization run, and when did one of them last change what somebody did that afternoon?
The question is not rhetorical scepticism about reporting. Dashboards were a genuine advance when they arrived, and the discipline of putting a number in front of an operator on a regular rhythm remains sound. The issue is what happens after the operator reads it.
What question does a dashboard actually answer?
The first of three, and only the first. Dashboards answer What: what happened, what the current state is, how the number moved against last period. That is a real service and it is the easy part, because the answer lives inside a single system that already holds the measure.
The second question is So What, meaning what the movement implies for this part of the business. The third is Now What, meaning the action, its owner, its timing, and what it is expected to achieve.
Most reporting stacks are excellent at the first, thin on the second, and silent on the third.
Why do dashboards answer So What so poorly?
Because meaning is relational, and a chart is not.
A seven-point drop in on-time delivery means something different depending on which customers it touched, whether those customers are in a renewal window, whether the affected lane also carries a product family already under margin pressure, and whether the supplier behind it is showing financial distress. None of that lives in the delivery dashboard. It lives in the relationships between the delivery system, the CRM, the contract repository, and a supplier risk feed.
A chart can show the drop. It cannot cross four systems to tell you which accounts are now at risk, because nothing in the estate states that the customer in one system and the account in another are the same company.
That is why the So What usually arrives as an analyst’s email three days later, if it arrives at all.
Why is Now What missing entirely?
Because a recommendation requires a position on what should happen, an owner willing to take it, and a record of what it was expected to achieve. A reporting layer produces none of those, and it was never designed to.
So the action gets decided in a meeting, executed in an operational system, and disconnected from the number that prompted it within a week. Two quarters later the number has moved and nobody can say whether the action was the reason.
Is the answer more data?
No, and this is the part worth being direct about. The action gap is caused by incoherent data rather than missing data, and adding more incoherent data widens it.
IDC put the global datasphere at 149 zettabytes in 2024, up from 64 zettabytes in 2020. Almost none of that growth has made the average operational question easier to answer, because aggregation moves data without resolving meaning. A lake gathers five records describing one customer into one place. It does not state that they are one customer.
Coherence is a property of the relationships between datasets: the same entity resolved across systems, the same event reconciled across timestamps, and the relationships held as data in their own right rather than rebuilt by hand every time somebody asks.
What changes when the entities are resolved?
The question stops being an integration project and becomes a query.
Which customers grew in volume and declined in margin. Which suppliers appear across three business units under different names. Which contracts renew next quarter below current pricing. Which open recommendations have gone quiet, and who owns them.
Each of those crosses systems that nobody pre-joined, and each is answerable in one traversal once the entities are resolved and the relationships are typed. None of it exists in any public corpus, because it exists only in the relationships between your own systems.
Does this replace the dashboards?
No. The reporting layer keeps doing what it does well, and the systems of record keep running the business.
What gets displaced is the manual work that grew up around the dashboard because nothing held the connected picture: the reconciliation somebody performs each month, the scorecard maintained by hand, the pack assembled from four extracts before every review. That work exists because the connection was missing. Once it is present, the work becomes a query, and the meeting can be about the decision rather than about whose number is right.
Past the dashboard
The PolyPhaze white paper Best decision. Best action. covers the four-step workbench structure, the six value levers, and how a recommendation carries its confidence through to a closed loop. Download the full P&L intelligence ebook for the detail behind this.
Frequently asked questions
What are the limitations of a business dashboard?
A dashboard reports what happened inside the system that holds the measure. It cannot establish what a movement means across other functions, because meaning depends on relationships between systems, and it does not produce a recommendation, an owner, or an expected outcome.
What is the difference between What, So What and Now What?
What is the observed state, which reporting tools handle well. So What is the implication for the business, which requires cross-system context. Now What is the recommended action with its owner, timing, and expected result, which requires a record that survives past the meeting.
Does a knowledge fabric replace business intelligence tools?
No. Reporting tools continue to present measures and systems of record continue to run transactions. What changes is that questions crossing several systems become a single query rather than a reconciliation exercise performed by hand each month.