Residential
What this tier can tell you
Half-hourly, or ~10 s with a Consumer Access Device · via n3rgy consumer API (consented DCC)
- Separate standing load from occupied-hours load
- Flag overnight consumption that should not be there
- Detect a rising standing load — a failing appliance
- Run entirely on the same detectors, configured differently
What it cannot
Stated, because a capability nobody contradicts becomes an assumed one.
- Do appliance-level disaggregation from half-hourly data
- Weather-normalise without a local weather series configured
- Produce numbers as large as a commercial site — expect tens of pounds, not thousands
Connect a meter
- Register the meter at data.n3rgy.com. You need the MPAN and the MAC address from the In-Home Display.
- Export the credentials — keep them outside the repository:
export N3RGY_MAC=<IHD MAC, no hyphens> export N3RGY_MPXN=<your MPAN> # optional
- Pull the history and analyse it:
make home DAYS=365
The adapter handles what real data does and seed data does not: the 90-day request cap, roughly 13 months of retained history, comms gaps marked explicitly as MISSING so a hole cannot read as low consumption, and gas units read from the payload rather than assumed.
What to expect on the first run
Two things will look like bugs and are not, so they are worth stating before you see them.
No weather series is configured for a home, so degree-day normalisation has nothing to work with and the weather-sensitive detectors stay quiet. A silent detector is not evidence of a clean building — the CLI says so on every run.
The detectors are tuned for commercial sites with defined opening hours and cohorts of comparable buildings. Zero findings on a single dwelling is a plausible outcome, not necessarily a fault. That result is the point: it tells you whether the thresholds transfer, which no amount of synthetic data can.
Why the same engine works unchanged
A home has occupancy hours rather than opening hours, so “energy used while the site is closed” becomes “energy used while the house should be asleep”. That needed no new code — the operating calendar is configuration, and the analytics core knows nothing about whether a site is a restaurant or a house. It is the same property that lets a GB pub and a Riyadh restaurant share one engine.