A meter testing program that samples every lot to the standard, keeps every test record complete and sends a failed lot straight to replacement
Every lot sampled to the standard; a failed lot goes straight to replacement.
See one case, screen by screen ↓Every regulated utility has to prove its meters register correctly. Most do it by statistical sample: the installed meters are declared as homogeneous lots by make, model and purchase year, a sample is sized from each lot under ANSI/ASQ Z1.9 and pulled every year, each meter is tested on a meter-shop test board, and the lot passes or fails on the result. Each test record has to hold the unit, its location, the date, the reason for the test and the as-found and as-left readings, and the whole programme is summarised in an annual report to the commission.
Today that work sits in four places. The test boards export a file each, the asset register holds the lot, the sample is drawn in a spreadsheet, and the removal work order lives somewhere else. A record missing its reason or its as-left reading is a finding at audit. Lots are judged once, when the annual report is assembled — months after the samples were tested, with less of the 12-month removal window left and sometimes after the manufacturer's warranty has run out. Meanwhile meters that read slow lose revenue quietly, and a failed lot is not a line in a report but thousands of exchanges to plan by route, stock to order and a commission notice to file within 30 days.
Estimated split for one programme year at a utility the size of Lumina, in working days, by hand and with the solution. The annual report figure starts from the 3–4 weeks of compliance analyst time Lumina estimates today.
Seven agents size, read, chart and draft every lot's testing; the metering engineer signs each decision.
Lumina Utilities has 1.24 million electricity and 680,000 water meters, declared in 48 lots for the 2026 sample programme. Lot KX2-2016 is 8,400 Kestrel KX-200 single-phase smart meters bought in 2016, across Harrowgate, Eastvale, Millbrook and Ashby Down. On Tuesday 13 October, test board TB-2 at the Eastvale meter shop has 14 new results. Here is what happens next, screen by screen, in the working solution.
Nadia owns the 2026 sample programme and signs every lot decision. The lot board shows every lot as a square sized to its meters in service, by make down and purchase year across: 34 passed, 1 on watch, 5 in test, 2 failed, 6 due — 5,070 of 6,050 samples tested. KX2-2016 is in test at 61 of 75, with a quality index of 1.70 against an acceptability constant k of 1.53.
“In test · 61 of 75 · close to k — Q so far 1.70 · best case if the rest read 0.00%: 1.95 · 14 results waiting on TB-2”
The plan reads as one line: 8,400 meters, general inspection level II, code letter L from the Z1.9 table, 75 samples, AQL 2.5%, k 1.53. The plan calculator recomputes the letter, the sample and k for any lot size from the approved Appendix B. The sample was a seeded random draw on 14 January, and 75 removal work orders were raised the same day.
Limit: the sample planner never redraws a sample once results exist, and the seed stays on the record.
Ben has tested 14 KX2-2016 samples this morning; the board file tb2-20261013-0740.csv is waiting. On the test bench the other boards are busy — customer tests on TB-1, a new Orrin S6 shipment on TB-3, water samples on TB-4. The test record checker has flagged 3 of 74 records today as not complete, and reference standard RS-3 is due for calibration in 6 days.
“RS-3 is due in 6 days. TB-3 stops taking tests on 19 Oct unless it is recalibrated.”
The test record checker reads the file: 14 rows, each serial matched to KX2-2016 and its removal work order, all 14 records complete, reference standard RS-2 calibrated 2 Sep. The acceptance analyst adds each result to the chart against the ±2% band and redraws the running quality index. Most read slow at light load, and at sample 68 the analyst shows the lot cannot reach k whatever the last seven read.
“Cannot pass at sample 68 — even if every remaining sample reads 0.00%, Q reaches only 1.49”
At 75 of 75 the lot fails: mean −0.90%, s 0.86, Q 1.29 against k 1.53. Six of the 75 meters are outside ±2%, all slow, and about 828 meters of the lot are estimated below −2%. KX2-2016 turns red on the board, and all 8,400 meters are flagged for replacement — Harrowgate 3,120, Eastvale 2,640, Millbrook 1,580, Ashby Down 1,060.
Limit: the acceptance analyst proposes pass or fail with the working shown. It never signs a lot decision.
The lot record holds the Z1.9 worksheet, every sample, the out-of-limit list and the agents' notes. Five of the six out-of-limit meters are in serial range KX16-04xxxx, production weeks 14–22 of 2016 — the weeks Kestrel bulletin KB-19-07 names for light-load drift. Light-load errors in the lot average −2.51% against −0.49% at full load, and delivery certificate KC-16-2207 shows that batch passed light load by only 0.4%.
“The lot was declared as one group on 14 Jan 2026 and cannot be split after sampling (Rule 7.40.5(c)). Serial range KX16-04 is proposed as its own lot in the 2027 plan.”
Nadia sees the lot, the plan, the result and the decision on one screen: Z1.9 level II, letter L, n 75, AQL 2.5%, k 1.53; Q 1.29, mean −0.90%, s 0.86, 6 of 75 outside ±2%; lot fails, remove within 12 months under Rule 7.40.6. She signs as metering engineer.
Meaning of signature: “I have reviewed the 75 test records and approve this lot decision”
Programme RP-2026-03 was drafted at 10:17: five waves by meter-reading route from November 2026 to March 2027, with the 2,310 meters in serial range KX16-04 first in each wave. The clocks sit on top — commission notice due 12 Nov, Kestrel warranty ends 30 Nov, removal window ends 13 Oct 2027, no customer refunds due because no meter read fast beyond +2%. Stock is 6,200 Orrin S6 on the shelf and 2,200 more on SH-2026-032, with 4 crews at 90 exchanges a crew-day.
Limit: the replacement planner never sends anything to the commission, a supplier or a customer — the notice, the warranty claim and the 8,400 letters wait for a named person.
The 2026 annual meter test report to the Westmark Public Service Commission is a draft built from 5,084 test records. Section 4, Failed groups and actions, now shows KX2-2016 — 8,400 meters, Q 1.29, failed 13 Oct, RP-2026-03 in five waves — beside the two lots that failed earlier in the year. Every section carries a note that opens the records behind it.
Checks before filing: “3 records incomplete on the test bench · 966 samples still to test before year end” — Priya signs and files.
The same lot, seen by the five people who own it — from the test board to the commission — and what their year looks like now.
Seven agents size, read, chart and draft every lot's testing; the metering engineer signs each decision.
Groups the installed meters into homogeneous lots by make, model, firmware family and purchase year from the asset register, and reports each lot with its size and districts.
Sizes each lot's sample to ANSI/ASQ Z1.9 level II from the approved table — code letter, sample size and acceptability constant — draws a seeded random sample and raises the removal work orders.
Reads each test-board file, matches every serial to its lot, sample number and removal work order, and checks unit, location, date, reason, as-found and as-left, technician and the reference standard's calibration within 120 days.
Computes the mean, standard deviation, quality indices against both limits and the estimated share outside limits as results arrive, works out the best quality index a lot could still reach, and proposes pass or fail.
Reads manufacturers' certificates of conformity — quantity, serial range, test results, firmware — compares them with the shipment manifest and the receiving tests, and links manufacturer bulletins to the lots they name.
Turns a failed lot into replacement waves by meter-reading route inside the removal window, checks stock and crews, drafts the commission notice, customer letters and any warranty claim, and works out refunds for fast meters.
Drafts the annual meter test report in the commission's section order and customer test letters in plain words with the refund worked out.
Signs every lot decision. The agents propose; a named person decides.
Ask about any lot, test record, customer test or the annual report — or say what to change. Every answer points to its records.
Why is KX2-2016 failing?
Light-load drift in one production run. With this morning's results from TB-2, 6 of 75 samples are outside ±2%, all slow; light-load errors average about −2.5% against −0.5% at full load. 5 of the 6 sit in serial range KX16-04, the weeks named in Kestrel bulletin KB-19-07. The delivery certificate shows that batch passed light load by only 0.4% in 2016.
Can we split KX2-2016 by serial range so the rest passes?
Not for 2026. Rule 7.40.5(c) freezes a group once its sample is drawn, and this lot was declared on 14 Jan. Splitting it now would look like choosing the result. Serial range KX16-04 is proposed as its own lot in the 2027 plan, which also gives the rest of the 2016 meters their own sample next year.
Which lots could fail next year?
Two to watch: OS4-2019 passed with Q 1.67, only 0.14 above k, and older Orrin S4 lots are trending down; BW15-2013 and BW15-2014 passed at 2.46, the lowest of the water lots. The lot board is filtered to the lots on watch.
Draft the commission notice for KX2-2016
The notice is drafted for Priya Raman: group KX2-2016, 8,400 meters, Q 1.29 against 1.53, six meters slow, removal in five waves by March 2027. It is due by 12 Nov, 30 days after the failure. Nothing is sent until you approve it.
13 screens from the working solution, on its sample data. Pick one to see it large.
The sample plan, the limits, the clocks and who decides are settings. Every change saves a new draft plan version that needs the metering engineer's approval.
| Setting | Default | Choose from |
|---|---|---|
| AQL | 2.5% | 2.5% · 1.5% · 4.0% |
| Inspection level | General II | General II · General III · Special S-4 |
| Variability method | Standard deviation, unknown | Standard deviation, unknown · Range method |
| Watch margin above k | 0.20 | 0.20 · 0.10 · 0.30 — a lot that passes this close to k goes on watch |
| Group by | Make, model, purchase year | Make, model, purchase year · Make, model, firmware, year |
| Electricity in service · water in service | ±2.0% WA · ±1.5% normal flow | ±2.0% WA · ±1.0% WA for electricity; ±1.5% · ±2.0% normal flow for water |
| Commission notice after a failure · failed-lot removal | 30 days · 12 months | 30 days · 14 days; 12 months · 18 months |
| Customer test clock · reference standard calibration | 10 working days · 120 days | 10 · 15 working days; 120 · 90 days |
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 arithmetic, inputs shown: $0.32M = KX2-2016 8,400 meters × 7,800 kWh × $0.16 × 0.90% ($94K) + OS4-2018 21,600 × 7,800 kWh × $0.16 × 0.71% ($191K) + BW15-2012 14,200 × 180 m³ × $1.90 × 0.64% ($31K), the demo utility's inputs; 150,700 = 1.24M electricity meters ÷ 15 years + 680,000 water meters ÷ 10 years; the working-days split is our estimate · “typical” = published figures (TESCO, Meter Testing 101, NC Meter School 2022, citing ANSI C12.1 — statistical plans need under 30% of the testing of a periodic plan). Standards and rules named as context: ANSI C12.1, ANSI/ASQ Z1.9; the acceptability constants shown are the demo utility's approved plan. Lumina Utilities, the Westmark Public Service Commission and its rules, the meter makers, the people, places and records named on this page are fictional — characters and sample data in the working solution.
Software that runs a utility's in-service meter testing program from the sample plan to the annual report to the commission. Meter Testing Compliance puts every installed meter into lots, sizes each lot's sample, reads every test-board file into a complete test record, charts each lot's acceptance as results arrive and drafts the replacement programme, notices, customer letters and the annual report. The metering engineer decides each lot, and regulatory compliance files.
ANSI C12.1 allows in-service testing by statistical sample, with meters divided into homogeneous groups. The solution sizes each lot's sample from your approved plan — in the sample case 8,400 meters at general inspection level II gives code letter L and 75 samples at AQL 2.5% — then computes the mean, standard deviation and quality index as results arrive and compares the quality index with the acceptability constant k. A lot passes when Q is at least k.
The acceptance analyst shows the point where the lot can no longer pass and proposes the failure with its working. Once the metering engineer signs, the replacement planner's draft programme opens: waves by meter-reading route inside the removal window, stock and crews, the commission notice, the warranty claim and the customer letters. The annual report's failed-groups section is redrafted from the same records.
Yes. In the sample utility 20 of the 48 lots are water meters — piston and ultrasonic — tested on a water flow bench against ±1.5% at normal flow, beside 28 electricity lots tested against ±2.0% weighted average registration. Limits, clocks and grouping are settings.
Each request runs on a 10-working-day clock, and the test record checker flags a record that is not complete — such as a meter adjusted and returned to service with no as-left reading. The report writer drafts the result letter in plain words with the refund worked out from the rule — in the sample case $38.19 for a meter found 2.61% fast — and customer relations approves and sends it.
Yes. The certificate reader compares the certificate of conformity with the manifest and the receiving tests — quantity, serial range, errors and firmware. In the sample shipment of 4,000 Orrin S6 meters, 3 serials on the board were not on the certificate, so the proposal is to accept 3,997 and hold 3. The agent never accepts a shipment itself.
Yes. The agents read, calculate, draft and propose; they never sign or send. The metering engineer signs every lot decision with its meaning recorded, regulatory compliance signs and files notices and the report, customer relations approves customer letters, and the meter shop supervisor releases records as complete.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your commission's testing rules, your approved sample plan, test-board file formats and last year's annual report — and it goes live once every quality gate has passed. We will walk you through it on your own lots first.
Every held bill released inside its cycle, each estimate with its method and source.
1-min demoEnergy Theft DetectionAn 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 demoWe'll run Meter Testing Compliance on a sample of your own lots, test-board files and last year's annual report.