LIGHTHOUSE FOR ENERGY / REGIONAL RETAIL

Data observatory

Operating history and future scenarios, July 2021–June 2027.

Ask Lighthouse
Retail customer portfolio / 763,000 customers / 48 cohorts
July 2026 – Mar 2027
Explore the modelled regional retail book. It is separate from the modelled 6,052-account decision book. Open the decision book → · Sign in for Ask Lighthouse and saved decisions →
I run a book of:

Book: Residential + SME + C&I · 6,052 accounts in the decision book.

Regional scenarios cover residential and SME customers; C&I sites are in the decision book only.

Six years of customer energy

Monthly customer imports and PV exports (MWh), with published market and weather context

View data
Customer importsPV exports

Market context

Monthly mean QLD1 spot price · $/MWh · published to the observation date, modelled after

View data

Customer contact history

Linked operating events

View data

Pack integrity

72 months of retained operating inputs, with separate historical periods and future scenarios.

Disjoint cohort population control48 unique cohorts sum to 763,000; modelled control 763,000.

Passed

Linked sample meter register5000 unique modelled NMIs; minimum sample 104.

Passed

Joint meter-day weights and separated channels2304 joint daily samples, each 48 intervals; each cohort/season/day-type weight sums to one.

Passed

Export only for solar cohortsB1/export channel is never netted against E1/import.

Passed

Full six-year fixed-AEST coverage2191 complete days July 2021 – June 2027, 48 shape intervals per day. Settlement archive keeps original source resolution.

Passed

72 linked months per cohort3456 monthly cohort records; static modelled population, future months labelled scenario.

Passed

Effective-dated tariff coverage without overlapsEvery cohort/month has exactly one retail tariff; QCA published versions replace modelled design versions for the same period.

Passed

104 modelled CTS activitiesEvery modelled activity has non-empty cohort allocation weights. Cost arithmetic is regression-tested.

Passed

Traceable workload event lags15 modelled event→activity elasticities with unit-sum lag distributions.

Passed

Credit fields are integer centsGenerated credit ledgers use integer cents; runtime tariff multiplication uses the shared decimal engine.

Passed

No modelled observations claim measured authorityOnly independently retained public schedules/sources may be published. Modelled book, register and elasticities remain modelled.

Passed

No non-finite generation outputsFinite non-negative usage and finite prices; negative wholesale prices remain valid. Missing ingestion data is separately tested as unknown.

Passed

Full six-year NEM12 archive generated5000 gzip files; 17,343,956 channel-days; 2054 explicitly blank unknown reads; 1704 estimated-quality reads.

Passed

Residential annual usage vs QCA typical customerModelled residential mean 4958.23 kWh in 2025–26. QCA typical 4748 kWh; difference 4.43%. Mean-vs-tariff-median scopes differ and are stated explicitly.

Passed

Published typical T11 bill reproductionAt 4748 kWh: calculated GST-inclusive $2,030.78, QCA $2,031; difference 0.011%. No feed-in tariff or concession applied to this reference bill.

Passed

Gross import shape vs QLD1 ≥0.8Pearson correlation of 48-point average day over 1888 common source days: gross import 0.359. This original gross-load gate is not met; no reclassification to a different metric.

Needs review

Net grid shape diagnostic vs QLD1Net import minus export correlation 0.844. QLD1 operational demand is a net regional measure. This diagnostic does not silently replace the gross-import acceptance gate.

Passed

Joint shapes vs full sample source energy ±2%All 5000 meter index totals compared with the joint model's sample-weighted six-year import+export energy; worst cohort difference 0.00175%. Blank gaps are excluded from observed totals, never converted to zero. Full parser validation independently verifies indexed totals.

Passed

AER weighted hardship/concession controlsModelled residential shares hardship 2.05%, concession 33.59%. AER Q1 2026 rates applied to modelled 663,808 residential accounts; not a claim that modelled counts equal the 741,548-account observed retail population.

Passed

All market inputs are public observationsNot met: future prices/weather, pre-2025 modelled retail schedules, earlier network schedules, losses, green/market fees and hedge book assumptions remain modelled. Published records are specifically identified; a modelled history is not relabelled real.

Needs review

Files and provenance

Large meter archives are generated locally; published inputs retain their identity

FileRowsSizeAuthorityFingerprint
Loading file index…
Page 1 of 1
Published inputs stop at their available observation date. Customer records and future periods are modelled; they are never presented as actual customer history.
Lighthouse for EnergyCalculation engine v0.4.2 · 72-month data, v1Scenario analysis from stated assumptions