VEE exception management that releases every held bill inside its cycle, each estimate with its method and source
Every held bill released inside its cycle, each estimate with its method and source.
See one case, screen by screen ↓Rules-based VEE in the meter data management system raises thousands of exceptions a night: gaps, spikes, zero consumption, register-interval mismatches, multiplier errors, reverse energy, usage after move-out, water night flow. The rule says which test failed. It does not say why, which estimate belongs in the gap, or whether to send a truck. So the MDM analyst opens raw interval data, the meter event log, the outage log and the premise record one exception at a time, while the bill cycle clock runs and bills sit on hold.
Rules that are too strict or badly calibrated bury the real faults. A collector firmware update drops 212 meters for six hours and raises 212 exceptions. Vacant premises raise zero-consumption alarms every night. The default straight-line fill flattens a clinic's lunchtime dip and afternoon peak and puts kWh in the wrong time-of-use band. Meanwhile a CT multiplier keyed wrong at a meter exchange sits in the same queue and under-bills for months, and service orders go out for meters that are working.
Estimated split of a typical analyst's day on the exception queue, by hand and with the solution. Our estimate, not a measured result.
Six specialist agents explain, estimate and route every exception; the MDM analyst approves what reaches a bill.
Lumina Utilities reads 1.42 million electricity meters every 15 minutes and 890,000 water meters every hour. Overnight, the rule book raised 6,842 exceptions from Tuesday 13 October's reads, and cycle 20 closes on Thursday at 18:00. Here is what happens next, screen by screen, in the working solution.
The run finishes at 02:46. Standing rules approved by the supervisor close 4,906 known false alarms, 1,584 are ready to accept in six groups, and 352 need a person, ranked by the bill they hold and the hours to cycle close. At the top: Corran Bay Dental Clinic, meter E-4471802, a 6-hour gap from 10:00 to 16:00 on a Small Business Time-of-Use account, its $648.20 bill held in cycle 20.
“Rules raise the exception · agents find the cause · you decide what reaches a bill”
The evidence agent reads 72 intervals, 4 register reads, 3 meter events, the outage log (none) and the weather. Comms were lost at 10:02 and restored at 16:04 with no power-fail event, so the clinic had supply. Collector CR-118 took a firmware update at 09:58 and 212 meters dropped in the same window. The register moved from 48,171.63 to 48,213.47 kWh: 41.84 kWh went through the meter in the gap.
Evidence the agents used: “Comms lost 10:02, restored 16:04, no power-fail”
The agent hides 10:00–16:00 on the meter's last 8 Tuesdays and measures each method: similar day 2.8%, average of 4 Tuesdays 4.9%, neighbouring meter 7.4%, straight line 11.6%. Tuesday 6 October matches on weekday, clinic hours (08:00–18:00) and weather, 29°C against 30°C. Swapping to the neighbouring meter, Corran Bay Pharmacy on the same transformer TX-CB-207, redraws the pale line and moves the peak band from 15.44 to 14.89 kWh.
Straight line: “09:45 and 16:15 reads · over 2 h not allowed”
Nadia keeps the similar day. Tuesday 6 October's six hours held 40.31 kWh, scaled ×1.038 to the 41.84 kWh the register counted: 26.40 kWh in the shoulder band and 15.44 kWh in the peak band. She approves it with reason code E-SD, and the $648.20 bill is released to cycle 20.
“The original (missing) reads stay on record. The estimate is written as a new version.”
The other meters on CR-118 are flat-rate homes with a register read either side of the gap: same cause, same method. Each meter keeps its own backtest; the median is 3.1%, the worst 6.8%. Nadia accepts all 211 as one batch, and 211 bills worth $19,855.10 go to cycle 20 with no field orders.
“CR-118 batch written: 211 estimates · 211 bills released to cycle 20 · 0 field orders”
Westhaven Cold Store's daily kWh halved on 21 August, the day its meter was exchanged, while refrigeration hours did not change. The installation record says CTs 400:5, multiplier 80; the meter was set to 40. The proposal: correct the multiplier to 80, rebill 21 August to 13 October, and check the CTs on site. $18,640.32 was under-billed.
“A multiplier change needs a second approver. Hugo Lindqvist will be asked before the rebill of $18,640.32 goes out, and a CT check goes to Darius Okoro.”
At 7 Ostler Way, hourly night flow has not fallen below 18 L/h since Wednesday 7 October, while daytime use follows the household's usual pattern and no network leak is reported on the street. The cause found is a probable leak on the customer side. No edit and no field order: the customer gets a leak alert.
“No edit. Send a leak alert to the customer; no field order”
The rules alone would have raised 418 investigations this week. The field-order agent recommends 61, each with the meter, the fault, the evidence and the work type: 6 CT or multiplier mismatches, 14 stuck registers, 4 wiring or reverse-rotation faults, 29 meters silent 7 days or more and 8 tamper events. The other 357 meters are fine: communication loss with a consistent register, vacant premises, confirmed outages, unrecorded solar and real usage.
“357 × $250–$500 per visit = $89,250–$178,500 not spent this week”
Cycle 20 closes Thursday 15 October at 18:00. With the clinic and the CR-118 batch cleared, the bills it holds fall from 1,214 ($486,300) to 1,002 ($465,797). Mei-Ling sees what still holds them, gaps, register mismatches and spikes, and the largest bills held, and releases cleared bills to billing.
“Total 1,002 bills · cleared bills are released by Mei-Ling Tan”
Each week the rule-tuning agent groups exceptions by rule and the cause found, and tests every change on 8 weeks of reads. V-03 zero consumption: skip premises with a vacant flag or a recorded move-out, 1,041 fewer exceptions a night and 0 true faults missed. V-01 missing intervals: one exception per collector during announced firmware windows, 812 fewer a night. Each proposal is sent to Hugo Lindqvist, who approves it or not.
V-02 spike: “−604 exceptions a night · 1 true fault found later by V-04 (8 weeks) · residential unchanged”
The same night's exceptions, seen by the five people at Lumina who carry them, from the analyst to revenue assurance: what their day looked like, and what it looks like now.
Six specialist agents explain, estimate and route every exception; the MDM analyst approves what reaches a bill.
Picks up the exception queue when the overnight run finishes, joins exceptions on the same meter, applies the standing rules and ranks what is left by the bill it holds and the cycle close.
Gathers everything about the meter and the window: interval and register reads, meter events, outages, premise and account, neighbouring meters on the same transformer and the day's weather, each cited to its source system and time.
Names the cause of each exception: communication loss, power loss, meter fault, configuration error, real usage, premise change or leak, with the evidence for it and a confidence.
Proposes the estimate or edit. Tests each method the rule book allows on the meter's own last 8 matching days, picks the lowest error, scales it to the register and explains the day or meter it borrowed from and the split by tariff band.
Decides service order or no action. Recommends a visit only when the evidence shows the meter or installation is wrong or silent for 7 days or more, and drafts the order with its evidence.
Every Monday, finds the rules that raise the most exceptions with a known cause, drafts a change and tests it on 8 weeks of reads: exceptions removed and true faults missed.
Approves estimates; supervisor approves rules. The agents propose; a named person decides.
Ask about any exception, estimate, rule or bill cycle, or say what to change. Every answer points to its reads and records.
Why a similar Tuesday for Corran Bay Dental Clinic?
The gap is 6 hours, so the rule book rules out a straight line. On this meter's last 8 complete Tuesdays, with 10:00–16:00 hidden: similar day 2.8%, average of 4 Tuesdays 4.9%, neighbouring meter 7.4%. Tuesday 6 October matches on weekday, clinic hours and weather (29°C against 30°C). Its six hours held 40.31 kWh; the register says 41.84 went through the meter, so it is scaled ×1.038.
Show every exception from the CR-118 firmware update
212 meters lost communication when collector CR-118 updated its firmware at 09:58 and came back at 16:04. None had a power-fail event, and all 212 have register reads either side. 3 are small businesses on time-of-use tariffs and need you; the other 211 are flat-rate homes, ready as one batch once you settle the method on Corran Bay Dental Clinic.
Which rule raises the most false alarms?
By share, V-01 missing intervals (93% had a known cause last night) and V-03 zero consumption (91%). V-03 is the cheapest to fix: 58,296 of its 72,128 exceptions in 8 weeks were on premises already flagged vacant or with a recorded move-out. Skipping those would remove about 1,041 exceptions a night and miss no true fault.
Hold any bill over $5,000 for a second check
Done. Bills over $5,000 now need a second check by Hugo Lindqvist before release. One bill in cycle 20 is affected: Westhaven Cold Store, $9,412.60, which also carries a multiplier correction. The change is under Settings, When a person must approve, and on the audit trail.
13 screens from the working solution, on its sample data. Pick one to see it large.
What the agents may propose, and who must approve, are settings. Every change is recorded with who made it.
| Setting | Default | Choose from |
|---|---|---|
| Straight line for gaps up to | 2 hours (8 intervals) | 2 hours (8 intervals) · 1 hour (4 intervals); longer gaps use similar day, average day or neighbouring meter, chosen by backtest on 8 matching days |
| Scale to the register | On | On · Off — when register reads exist on both sides of the gap |
| Storm mode | On | On · Off — stop estimating on feeders in a declared outage |
| Gaps over 4 hours on time-of-use accounts | On · the MDM analyst approves | On · Off |
| Multiplier or CT/VT ratio changes | On · the MDM analyst and Hugo Lindqvist approve | On · Off |
| Bills over $5,000: second check | Off | On · Off — when on, Hugo Lindqvist checks the bill before release |
| Consecutive estimated bills | 3 bills | 3 bills · 2 bills — then a read is required before the next bill |
| Field orders, approved by Darius Okoro | 80 a day | 80 a day · 120 a day — daily cap |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample utility data · “target” = the design goal, measured in the live solution · “estimated” = our estimate: $4.6M a year = 357 needless orders a week × 52 weeks × $250, on the demo utility's figures — replace them with your own; the hours split is also estimated · “typical” = published figures (Utility Analytics Institute, “AI in VEE”, Oracle Utilities, July 2025 — a machine-learning model cut false-positive VEE exceptions 63 % in high-usage test scenarios, vendor-authored; industry studies — a truck roll costs $250–$500). Lumina Utilities and the people, premises, meters and accounts named on this page are fictional — characters and sample data in the working solution.
VEE (validation, estimation and editing) is the rule book a meter data management system runs over every read. Reads that fail a test become VEE exceptions on the analysts' to-do list. VEE exception management is the work of explaining each one, fixing the read with the right estimate or edit, and deciding whether a meter needs a visit. VEE Exception Management does that with six agents, and the MDM analyst approves what reaches a bill.
When the overnight run finishes, the triage agent closes known false alarms under supervisor-approved standing rules and ranks the rest by bill held and cycle close. The evidence and root-cause agents gather the reads, register, events, outages and premise and name the cause. The estimation agent proposes the estimate, and the analyst accepts, swaps the method or edits it.
It tests every method the rule book allows for the gap length and tariff on the meter's own last 8 matching days, hiding the same hours, and picks the lowest error. In the sample case a similar Tuesday scored 2.8% against 11.6% for a straight line, which the rule book does not allow over 2 hours. The estimate is scaled to the register whenever both register reads exist.
The field-order agent recommends a visit only when the evidence shows the meter or installation is wrong or silent for 7 days or more. In the sample week, the rules alone would have raised 418 investigations; it recommended 61. At $250 a visit, 357 needless visits a week comes to an estimated $4.6M a year. Meter operations approves every visit.
Yes, through the weekly rule review. The rule-tuning agent finds the rules that raise the most exceptions with a known cause and tests a change on 8 weeks of reads, showing exceptions removed and true faults missed. The meter data supervisor approves or declines each change; the agent never changes a rule.
Yes. The sample utility reads 890,000 water meters every hour alongside its electricity meters, and the rule book includes water spike and night-flow tests. A night flow that never stops is treated as a probable customer-side leak: the customer gets a leak alert, with no edit and no field order.
Yes. Agents propose; a named analyst approves every estimate and edit that reaches a bill, a multiplier change needs a second approver, billing operations releases held bills and meter operations approves field orders. Original reads are never overwritten, and every estimate carries its method, source and approver.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your VEE rule book, approval thresholds and a sample of past exceptions and reads — and it goes live once every quality gate has passed. We will walk you through it on your own exception queue first.
An estimated $8.4M a year recovered, with 3 in 5 site visits finding theft.
1-min demoMeter Event Management18,214 alarms become 23 actions, and every safety alarm reaches a person in 5 minutes.
1-min demoSmart Water Meter Leak AlertsThe customer hears about a leak the next morning, with its cost a day, not on the bill.
1-min demoWe'll run VEE Exception Management on a sample of your own VEE exceptions and reads.