A generation report has one job: to be believed. The moment an operator sees daily energy that does not match the revenue meter, or a capacity factor that contradicts what they watched on the SCADA HMI all afternoon, the whole report becomes background noise. The model and the visuals rarely cause this. The data layer underneath does, and it usually fails quietly.
A figure with no lineage is a rumour with a chart around it.
Why reports get ignored
Trust breaks in a handful of repeatable ways. Reported daily energy disagrees with the fiscal meter because the dashboard integrated inverter power at the wrong resolution, or summed instantaneous kW as if each reading were a kWh. A plant shows full availability while one string sat dark for six hours, because a frozen sensor kept repeating its last value and nothing flagged the reading as stale. Two screens cite two figures for the same hour and neither one wins.
Underneath all of it sits the same gap: no traceability. When a figure cannot say which device, which timestamp range and which transformation produced it, an engineer cannot confirm or refute it. Faced with a number they cannot check against a meter they trust, operators do the rational thing and ignore the report.
Earning the trust back
Trust is engineered into the data layer, not styled onto the front end. It starts with one source of truth: a single canonical time series that every report, alarm and forecast reads from, so no second spreadsheet is quietly disagreeing. Each value carries lineage — source device, raw register, units, the timestamp window and the aggregation applied — so any figure traces back to the Modbus or OPC UA tag it came from.
Then make uncertainty visible instead of hiding it. A confidence flag marks a value as interpolated, estimated across a gap, or built on a stale sensor. And reconcile continuously: settle the metered energy total against the SCADA-derived total each settlement interval, show the residual, and treat a persistent divergence as a defect to investigate, not a rounding quirk to paper over.
Operators do not need prettier charts. They need to open a report, see a figure, and know exactly where it came from and how far to trust it. That comes from lineage, honest confidence flags and reconciliation against the meter, built into the data platform from the start. It is the kind of plumbing we build into our data work, because a report nobody believes is worse than no report at all.