What is data visualization?

A chart is an encoding: values become position, length, angle or colour, and some encodings are read far more accurately than others. Position and length are read well, which is why a bar chart beats a pie chart for comparison. The aggregation should happen in the database, not the browser — sending a hundred thousand rows to the client to be summed is a common cause of a dashboard that takes seconds to appear.

The more expensive mistake is upstream of the design. A dashboard full of metrics nobody acts on gets checked twice and then ignored, so the question worth asking before building any of it is what decision each chart is meant to inform. Two charts someone looks at every morning are worth more than twelve nobody opens, and considerably cheaper to keep accurate.