Grid-Scale Analytics for the Energy Transition.
We build analytics engines on governed grid data: SAIDI/SAIFI/CAIDI reliability marts with defensible exclusion logic, transformer-loading and load-forecast models fed by AMI interval reads, and DER-aware demand analytics for a grid where large flexible loads queue for interconnection. For water and wastewater utilities, the same engine produces serviceability and CML measures and pollution-incident reporting with equal rigour.
Fragmented Silos
Legacy Utility systems and disconnected feeds.
Reproducible Reliability Marts
Grid-Edge Load Analytics
Production Reality
SAIDI/SAIFI/CAIDI Reliability Marts
SAIDI/SAIFI/CAIDI Reliability Marts
Transformer & Feeder Load Analytics
Queue-Aware Load Forecasting
Industry-Specific Friction Points
Reliability Metrics Assembled by Hand
SAIDI, SAIFI, and CAIDI are still computed from OMS extracts in spreadsheets, where major-event-day exclusions and customer-count denominators shift with each analyst. The numbers filed with regulators are hard to reproduce, let alone defend.
Forecasts Blind to New Load
Data centres, electrification, and DER adoption flip feeder load shapes faster than regression-era forecasting can track — and interconnection-queue signals never reach the models that plan capacity.
AMI Data Stops at Billing
Interval reads that could expose overloaded transformers, voltage excursions, and unmetered losses are used once for billing and archived. The analytics layer never sees the richest demand signal the utility owns.
How the Utility delivery flow works
This technical flow diagram reveals how Unolabs treats Utility data to deliver governed, production-ready outputs.
Source Layer
Signal Assembly
Joining VEE-cleared interval reads with outage events and connectivity, so every analysis knows which meter hangs off which feeder.
Industry Logic
Reliability Engine
Computing SAIDI, SAIFI, and CAIDI with codified major-event-day exclusions and versioned customer counts.
Load Intelligence
Forecasting feeder and system load with electrification, DER, and interconnection-queue signals in the feature set.
Activation
Decision Serve
Publishing reliability, loading, and forecast products to planners, regulatory teams, and demand-response programmes.
How the work is engineered for Utility
Reproducible Reliability Marts
We codify SAIDI/SAIFI/CAIDI logic — sustained-interruption thresholds, major-event-day exclusions, customer weighting — as governed transformations, so every filed number recomputes identically on demand.
Grid-Edge Load Analytics
We aggregate interval reads to transformer and feeder level to reveal loading, phase imbalance, and hosting-capacity headroom that the SCADA layer alone cannot see.
Queue-Aware Forecasting
We fuse historical load, weather, and interconnection-request pipelines into forecasting features that treat large flexible loads as schedulable resources rather than passive additions to peak.
Where the Real Work Is
MED Exclusions Are Governance, Not Math
The IEEE 1366 major-event-day method — the 2.5-beta threshold on daily SAIDI — is straightforward to compute. What is not straightforward is governance: many commissions modify or replace the IEEE method, customer-count denominators change mid-year, and a borderline event day can swing a reported metric either way. We implement the calculation and the decision trail together, recording which events were excluded, under which rule set, approved by whom — so the number and its justification travel as one artifact.
Baselines Decide What Demand Response Is Worth
Demand-response settlement is an argument about a counterfactual: what the customer would have consumed without the event. Comparable-day selection, weather adjustment, and same-day corrections each shift the answer, and disputes land on whoever holds the data. We build baseline engines where the methodology is explicit, versioned, and re-runnable per event, producing baseline-versus-actual evidence at the participant level that survives scrutiny from aggregators, customers, and the programme's regulator alike.
Loading Estimates Where No Sensor Exists
Most distribution transformers carry no direct measurement, so loading is inferred by summing AMI interval reads through the connectivity model — which means every identity gap propagates into the estimate. An unmapped meter makes one transformer look cool and hides another's overload. We publish loading estimates with explicit confidence flags tied to connectivity quality, and validate the method against the subset of assets that do have telemetry, so planners know which estimates to trust and which to field-check.
Visible work products, not vague advice
Each deliverable is designed to be used by Utility architects, engineers, data owners, and operations teams after the engagement ends.
How We Measure Grid Analytics
Analytical outputs are judged by whether they can be reproduced and defended. Every measure below is baselined against your current process in the opening weeks, then tracked release by release.
Filed-metric reproducibility
A previously filed SAIDI or SAIFI figure is recomputed end-to-end from retained inputs; the gap between recomputed and filed values — and the time it takes to close it — is the core trust metric.
Forecast error by feeder class
MAPE tracked separately for residential, commercial, and mixed feeders against a jointly agreed baseline model, so improvement claims are class-specific rather than averaged into meaninglessness.
Time to answer a regulator data request
Elapsed time from a commission question about a reported number to delivered supporting detail, comparing the manual before-state with drill-through after the marts land.
Loading-estimate coverage
Share of distribution transformers with a current loading estimate, broken out by confidence tier — growth must come from resolved connectivity, not loosened standards.
Frequently Asked Questions
Do you implement IEEE 1366 exclusions or our commission's rules?
Both, where they differ — and they often do. The IEEE 1366 major-event-day method is the reference implementation, and your jurisdiction's variations are codified alongside it as an explicit rule set. Every reported figure records which rule set produced it, so the same outage history can be reported consistently to different audiences.
Can you reproduce reliability numbers we filed in previous years?
If the underlying OMS records and customer counts still exist, yes — the mart recomputes historical periods under versioned inputs and exclusion decisions, and differences from the filed figures are surfaced rather than smoothed over. Where source data has been purged, we document the gap honestly instead of backfilling estimates into a filing trail.
How reliable are transformer loading estimates without sensors?
As reliable as the connectivity beneath them, which is why every estimate carries a confidence flag tied to meter-to-transformer match quality. We validate the aggregation method against assets that do have measurement, and unresolved connectivity feeds a stewardship queue — so accuracy improves through data correction, not assumption.
How an engagement starts
Pick one filed reliability number you would struggle to reproduce today and bring it to the discovery call — we walk through what data, rules, and decisions its recomputation would need, and scope the first mart from exactly that gap.
Interested in the full industry blueprint?
We have deeper technical documentation for Enterprise Decision Intelligence & Operational Analytics for Utility in the Utility sector.
Bring last quarter's SAIDI numbers. We'll show which outages telemetry saw coming.