What Makes a Public Dashboard Trustworthy?
From the earlier reference collection. Read the current software blog.
Trust comes from making a metric’s origin, timing, coverage, units, and limitations visible where people use it.
Put context beside the number
A large number can look authoritative even when its meaning is unclear. A dashboard becomes easier to trust when every important metric carries the context needed to interpret it: where the data came from, when it was last updated, what area or population it covers, and which unit it uses. Put those details next to the value or in a clearly labeled, keyboard-accessible detail panel.
For example, “42 outages” is ambiguous. Is that 42 currently active incidents, reports received today, or customers affected? A more useful label might say “42 active reported outages, countywide; reports from the utility feed, updated 10:15 a.m.” A person can then judge whether that count answers their question.
Make coverage and gaps explicit
Coverage is part of the metric, not a footnote. If a sensor network covers only selected stations, say so. If a dashboard displays the latest reading from each site, users should not mistake that for a complete regional average. List exclusions in plain terms, such as “Four of 12 stations are not reporting.”
Known gaps also help readers avoid overinterpreting apparent change. A chart might fall because fewer sites reported, rather than because conditions improved. Showing the reporting count alongside the total makes that distinction visible. When coverage changes over time, annotate the period or provide a small coverage series.
Choose precision the source can support
Units and definitions should be consistent across charts, tables, downloads, and mobile views. A temperature without °C or °F invites a wrong comparison. A rate should identify its denominator and interval—for instance, reports per 1,000 residents per month—rather than relying on a nearby paragraph to explain it.
Round values to reflect the source’s resolution. If a sensor reports tenths, showing six decimal places suggests accuracy the instrument does not provide. Document transformations such as daily averaging or deduplication, and distinguish a calculated estimate from a directly observed reading.
Treat trust as an ongoing practice
A practical review asks whether a new reader can answer five questions without hunting: what is measured, where it came from, when it was updated, what it covers, and what could be missing. Check those answers on the page itself, in exported data, and after a feed interruption.
Trust does not require pretending data is perfect. It grows when the dashboard states what it knows and what it does not, then gives readers enough detail to decide how much weight to place on a number.