Cases

Case studies

20 cases across markets, engineering and commercial decisions. Choose a challenge, work the numbers, then defend your recommendation.

18 of 20
Teaching scenarioADDED OCT 2026

One revised meter record, two different invoices

A data pipeline imports every version of a meter interval. The finance team's invoice increases even though no extra electricity was consumed. Your task is to trace the error and make the correction auditable.

Permanent case link

Your task

Reconcile raw versus valid records, separate estimated from confirmed usage and design a repeatable correction process.

PrepareVersion-aware interval ledger + invoice bridge + validation tests
Your role

Market-data analyst investigating a settlement reconciliation

Practice data & worked solution
PRACTICE DATASET

Do not sum revisions as new energy

Download CSV

Original synthetic meter dataset. Rows contain kWh per 30-minute interval, not kW. Version numbers and quality labels are simplified teaching fields, not a market file specification. Select the highest version for each interval. Price A$200/MWh throughout. Keep estimates visible; no tax, loss factor or settlement calendar is modelled.

Teaching inputs · synthetic data
Interval ending AESTVersionEnergy kWhQuality
00:301100Actual
01:001120Estimate
01:002110Actual
01:30190Estimate

Your analysis

  1. Calculate energy and cost if all rows are summed, then repeat using the latest version per interval.
  2. Report what proportion of the corrected energy is still estimated. Explain why multiplying these kWh rows by 0.5 would be wrong.
  3. Replace the final interval with a new actual version of 80 kWh. Recalculate totals and design tests for duplicate keys, versions, units and repeat imports.
Worked solution

Version-aware reconciliation

Summing raw rows gives 420 kWh = 0.420 MWh, costing A$84. Latest versions give 100 + 110 + 90 = 300 kWh = 0.300 MWh, costing A$60. The duplicate superseded record overstates cost by A$24. Preserve it for audit, but do not count it as additional consumption.

Quality and units

The remaining estimate is 90/300 = 30% of corrected energy. The rows already measure energy, so multiplying by half an hour would wrongly halve it. Only a power reading in kW would need an interval-duration conversion. An estimate is neither necessarily zero nor necessarily an error.

Revised invoice and controls

The new version gives 100 + 110 + 80 = 290 kWh, costing A$58, a further A$2 reduction. All selected rows are now labelled Actual. Make imports idempotent, reject conflicting same-version records, preserve provenance and test timezone/day-boundary handling. A later invoice adjustment still needs the actual settlement rules and timetable.

Review your response
  • Data: latest valid versions are selected by key, not by arbitrary row order.
  • Units: kWh is converted to MWh exactly once.
  • Quality: estimated shares and revision provenance remain visible.
  • Control: reproducible tests prevent duplicate imports and silent overwrite.

This checklist supports self-review. No tutor marking or independent credential is included.

Related topics 5
Sources & evidence

All figures are original teaching inputs, not observations or published forecasts. The official source supplies industry context, not this dataset. Source link checked 17 October 2026. AEMO metrology procedures provide context. Production handling must preserve actual quality flags, version semantics, interval boundaries and participant procedures.

AEMO metrology procedures and unmetered loads

Take the next step: applied projects

Assess a data-centre energy planConnect demand, grid capacity, procurement and backup power.Write an industry briefCheck industry data and turn your evidence into a recommendation.