IGNIS · meter-to-money cockpit

280,547 meter-points across 9 English grid regions · live month May 2026 · every figure computed from the data spine at request time · V1 Margin-Truth page

SYNTHETIC DATA · OPEN-DATA BACKBONE

True gross margin

£—

—

Arrears book

£—

—

Value at risk

£—

—

Cash collected

£—

—

Switching promise (5 working days)

—%

—

One live month on the books (May 2026) — the mini-charts show live composition of each number, not fabricated history. Trends arrive with June.

The AIgent crew — what the machines did, and where a human takes over (live, all runs · synthetic)

no AIgent runs yet

no decisions waiting

How revenue becomes true margin — May 2026, ex-VAT

Billed revenue, minus every cost the business actually carries — wholesale energy, networks, balancing, policy — down to the true gross margin. Reconciled per meter-point, to the penny.

—

✍️ GM pack sign-off — the human moment

Checking for a pending pack…

Where the money leaks — exceptions ranked by £ (pareto)

The few meters that carry most of the value at risk. Bars are pounds at risk per meter; the line is the running share of the top-12 total.

Why it leaks — by cause

—

The worst offenders — top meters by value at risk

#	Meter (MPAN)	What's wrong	Value at risk	Detail

Billing integrity — every bill reproducible from reads × tariff

—

Check	Failures	Result

—

Settlement exposure by region (first settlement run)

Region	Volume billed (kWh)	Imbalance (kWh)	Exposure

From quote to supply — the funnel

—

Switching health

—

Moves & empty properties

—

A change of tenancy with a read at the move keeps the bill clean for both households.

Switches in flight — registration queue

Switch	Meter (MPAN)	Region	Losing supplier	Requested	Due to complete

Every switch event this month

The arrears book, by age of debt

—

🛡️ Protected customers — human review only

—

Never auto-actioned. Customers on the Priority Services Register who fall into arrears are excluded from every automated collections step by a coded rule, verified live in SQL on every page load. The forbearance decision belongs to a named human — the system can only ask.

Payment plans

—

Direct Debit health — are we collecting the right amount?

—

What the Credit-risk agent did (this run)

The actionable queue — protected customers already removed · ranked by amount owed

—

Account	Owed	Days overdue	Age of debt	Next step taken	Payment plan

Cash in the door (settled, May 2026)

£—

—

Collections that bounced

—

Every bounced collection lands in the arrears book automatically; protected customers route straight to a human.

Refunds paid back

£—

—

Does the money tie? — sales ledger ⇄ double-entry ledger

Eight independent reconciliations, recomputed live. Each one must hold for the books to be trusted.

Reconciliation	Holds?

—

The books — chart of accounts

GL	Account	Type	Debit	Credit

Settlement maturity — exposure decays to zero at the final run

Each industry settlement run refines the position; by the final reconciliation the exposure must reach £0.

Run	Run date	Status	Imbalance (kWh)	Exposure

Industry data flows feeding the spine

Flow	Carried by	Rows on the spine

Honest seams: the transports marked “simulated” are stubs — the rows behind them are real spine data they would carry in production.

One customer, the whole story — every field a direct join on the spine (/360)

Portfolio KPI board — every figure deterministic from the spine

—

Live data heartbeat — two planes (consented 10-second telemetry vs half-hourly settlement batches · synthetic book paced by the real clock · — UTC)

The 10s stream is the consumer-consented CAD/HAN route (it never touches DCC); half-hourly blocks land in DCC overnight scheduled-read windows plus daytime stragglers, and the settled margin converges on the spine daily target to the penny by day end. Two planes that reconcile but never merge.

GSP groups (click to drill)

GSP	Meter-points	Coverage	Live power (consented)	Settled / target

Arrival feed (settlement plane)

Explain any number — ask the IGNIS Captain (grounded in the live spine)

Operational data is synthetic (provenance-labelled end to end); the property backbone is open data (England EPC, DESNZ fuel-poverty, deprivation indices). Every figure on this page is computed live from the V2 data spine plus the half-hourly meter fact (150.5M rows) by real SQL at request time — each dial equals its API endpoint equals direct spine SQL to the penny. Coverage is honestly 9 of 14 grid regions (England only). Estimated-read exposure is mostly billing latency, not leakage — metered energy not yet on a finalised bill. Customers on the Priority Services Register are never auto-actioned: a human owns every forbearance decision. Not affiliated with ENSEK. IGNIS V2 · anaigent / itsorted.