Most dashboard projects begin with layout. Which KPIs go at the top, which chart best suits which data, how wide are the columns, which colour flags a deviation. Those choices are not unimportant, but they only become relevant once the problem underneath has been resolved. And that problem is rarely the chart.
Why a well-designed dashboard can still fail
A dashboard that presents an incorrect number elegantly is no improvement over a spreadsheet doing the same. The credibility of a dashboard depends not on the visualisation tool but on the quality of the data flowing into it: the tag structure that determines what name a measurement carries, the aggregation logic that determines how raw values are summarised over a period, the status logic that determines when a machine is considered running, stopped or in changeover, and the normalisation that ensures a measurement on line 3 is comparable to the same measurement on line 7.
When an operator looks at an OEE figure of 78 percent and his experience tells him it is actually closer to 85, the instinctive response is to distrust the dashboard. That distrust is almost always justified, but the cause is not in the chart. It is in a definition that is somewhere wrong: the target speed used as a performance reference is outdated, or the downtime registration includes a planned stop that is operationally considered normal, or the data quality labels are marking a period as uncertain while the machine was simply running.
What data quality tagging actually resolves
Data quality labelling, marking each measurement as good, bad or uncertain, is not a formality for advanced users. It is the mechanism by which a dashboard can be honest about what it knows and does not know. A measurement labelled uncertain because the sensor connection was temporarily interrupted should be treated differently in a KPI calculation than one that arrived reliably. A system that does not make that distinction averages reliable and unreliable data together into a number that looks like a fact but is partly an interpolation.
Normalisation is the complementary step: ensuring that values measured in different places or different units are reduced to a common basis before reaching a dashboard. Energy consumption in kWh per tonne of output is a more meaningful number than kWh in absolute terms, but that calculation presupposes that output data and energy data share the same time base and the same granularity.
Where the dashboard project really begins
A good dashboard project begins with an audit of the tag structure: are the names consistent, are the units correct, are the sources reliable, are the definitions shared by everyone who will read the output? Then come sampling and aggregation: at what frequency is data stored, how are values averaged over a minute, hour or shift, and which aggregation method is correct for which type of parameter? Only once those questions are answered does it make sense to discuss layout and colour.
The role of Capture
Capture starts dashboard projects where they actually become credible: in the data foundation. Before layout, colours or chart types matter, the platform helps structure the tag model, validate data quality, define aggregation logic and normalise values so that measurements from different lines, machines or sites can be compared meaningfully. That is where dashboard trust is won or lost.
This is especially important in industrial environments, where a clean visual can easily hide weak assumptions underneath. A KPI that averages reliable and uncertain data together, uses the wrong target speed, or combines energy and production values at mismatched granularities may look polished but still mislead the people using it. Capture makes those assumptions explicit and manageable before the data reaches the visualisation layer. Data quality labels, consistent units, timestamp handling, aggregation rules and shared definitions become part of the system, not afterthoughts. The result is a dashboard that does not merely look clear, but deserves to be trusted.