Skip to main content
For general data quality help, see Data Quality FAQ.
Nectar continuously watches your utility data for unusual patterns. When something looks off — a jump in consumption, an unexpected cost increase, a late fee, or a spike in demand charges — it surfaces as an anomaly for your team to review, right alongside your other data quality items.
Anomaly detection is off by default. Turn it on in Settings > Company > Data Quality > Anomaly detection. Once enabled, Nectar checks your data automatically every day, and you can run a check on demand at any time.

​
What Nectar looks for

TypeWhat it flags
Usage anomalyA site’s consumption for a commodity is unusually high or unusually low for the conditions
Cost anomalyA site’s monthly cost for a commodity is unusually high
Unit-price anomalyThe effective price you paid per unit (for example, $/kWh) is unusually high
Interest chargesAn account is paying interest, late, or penalty charges
High demand chargesAn account’s electricity demand charges are unusually high
Each anomaly points you to the specific site or account, commodity, and time period involved, so you can investigate quickly.

​
How detection works

Nectar compares each site or account against its own recent history and flags meaningful deviations. Detection is weather-aware: it accounts for how hot or cold the weather was during each period, so normal seasonal swings (more gas in winter, more electricity in summer) don’t get flagged — only changes that genuinely stand out do.
Weather-aware detection uses historical temperature data from Open-Meteo, licensed under CC BY 4.0. Nectar processes this data into heating and cooling degree days for each site and period.
Interest charges and demand charges work a little differently: any interest you’re paying is worth surfacing, and demand-charge spikes are flagged against the account’s typical pattern. Nectar only flags a site or account once it has enough history to judge what “normal” looks like. New sites, or sites with sparse data, are left alone until there’s enough to compare against — this keeps the inbox focused on real signals.

​
Sensitivity

Each anomaly type has a sensitivity setting:
LevelBehavior
LowFlags only large, obvious deviations
MediumBalanced — the default
HighFlags smaller deviations too
Higher sensitivity catches more, but may include some patterns that turn out to be normal. Lower sensitivity is quieter but may miss smaller changes. Adjust it per type in Settings > Company > Data Quality > Anomaly detection.

​
Where to find anomalies

Anomalies appear in your Data Quality inbox alongside other data quality items — there’s no separate page to check. From the inbox you can:
  1. Filter to anomalies (and by type, site, or commodity) to focus your review.
  2. Open an anomaly to see a chart of the trend, the expected range, and the periods that were flagged.
  3. Resolve it once you’ve addressed it, or dismiss it if it isn’t a real problem.
Resolved and dismissed anomalies are kept for your records and won’t be flagged again on future checks. The Data Quality overview also rolls up your open anomalies by type and shows their estimated total dollar impact, so you can see where the biggest opportunities are at a glance.

​
Getting value out of anomalies

  • Catch overspend early. Cost and unit-price anomalies highlight bills that cost more than expected — often a rate change, a billing error, or a usage problem worth chasing.
  • Stop paying avoidable fees. Interest-charge anomalies surface late fees so you can fix the underlying payment or billing issue.
  • Manage demand. High demand-charge anomalies point to facilities where reducing peak demand could lower bills.
  • Spot operational issues. Usage anomalies (including unexpected drops) can reveal equipment left running, meters that stopped reporting, or changes in how a building is used.

​
Improving accuracy

If an anomaly’s chart shows gaps or unexpectedly low months, the underlying data may be incomplete — and incomplete history can make a normal period look unusual. Records with open data quality issues (such as unmatched accounts) are left out of the analysis. If a site has many unresolved issues, resolve them in your Data Quality inbox — the next check will use the corrected data and produce more accurate results.
See also: Glossary — Anomaly