Customer satisfaction index reporting that explains every move in the score by name, with the customers behind it
Every move in the score explained by name, with the customers’ voices behind each line.
See one case, screen by screen ↓Regulators, governments and benchmark studies score utilities on customer satisfaction, and the scores carry money and reputation. Each month a performance analyst rebuilds the index by hand: happiness-meter votes, survey files, the regulator’s results and operational KPIs arrive in different shapes, are pasted into a workbook, and the method document’s exclusions are re-applied one more time.
Then the score moves and the CX director asks why. A 2-point fall is “probably the outages”, argued in a meeting a week later. Nobody separates a real fall in a service from a change in who voted, or checks whether a 0.2 move is inside the margin of error. Voice of Customer already holds the calls and complaints behind the move, but the score report does not link to them, so the board never hears a customer — and when a service owner fixes something, nobody measures whether the score came back.
Estimated split for one monthly index close and board paper, in working days, by hand and with the solution. The time after includes people reviewing and signing.
Six agents close the month, split every move into named drivers and write the paragraph; the CX director signs.
Lumina Utilities, an electricity and water utility, closes September 2026 on Tue 6 Oct at 07:12. Its Customer Happiness Index has fallen 2.1 points, the Q4 target agreed with the Coastal Region Utility Commission is 92.0%, and the board paper is due Thu 15 Oct. Here is the month, screen by screen, in the working solution.
Rania’s home is a bridge from August’s 91.4% to September’s 89.3%, every floating bar a named driver labelled with its service and service-quality lens, the Q4 target as a dashed line. Five problems took 2.8 points and two fixes gave back 0.6. The month close ran before she arrived: 50,518 votes read, 1,906 excluded by method v3.2, 48,612 valid, and the index matches the meter platform to 0.01.
“Down from 91.4% in August, on 48,612 valid votes — a real fall, outside the ±0.3 margin. 2.7 points below the Q4 target of 92.0%.”
The movement explainer splits the −2.1 by service and service-quality lens. Power Services × Timeliness holds −1.1 — late notices (−1.3) less faster energisation (+0.2) — half the fall in one cell. Seven drivers pass the test: each moved by more than its margin at 95%, on at least 200 votes. The change in which districts voted, −0.1, is shown on its own as mix, not hidden inside a service.
“Cells under 0.1 point are hatched and never named in the paper.”
Driver 1 of 7: Power Services › Power interruption, Timeliness. 2,612 votes, happiness 63.4% in September against 87.6% in August, margin ±0.1 — a real move. Cedar Vale is worst at 55.6% on 1,184 votes, then Palm Ridge at 63.0% and North Quarry at 69.1%. The Voice of Customer challenge is CH-7F21: “Planned-outage notices arrive after the outage has started”, with 214 voters who left words.
Owner Tomás Reyes, power operations: “Notice schedule back to 48 hours before — in place since Fri 2 Oct”
The notice log shows 9 of 14 planned interruptions in September with notice under 48 hours, 6,214 customers in all. On PI-0914-CV11, Cedar Vale feeder CV-11, 1,420 customers got the text 8 minutes after the supply went off. The cause was change NS-0903 on 3 Sep: notices went out when a job was released to the crew, on average 14 hours before the start. It was reversed on Fri 2 Oct.
Service standard GS-4: “We will tell you at least 48 hours before a planned interruption to your electricity or water supply.”
Rania presses Explain this move. The score assembler re-reads the figures, the movement explainer confirms −1.3 of −2.1, the voice linker picks 6 of the 214 voices with names masked, and the method checker checks 11 figures against method v3.2 and the notice log. The narrative writer then writes eight sentences, each carrying links to the figures or the customers it rests on.
“The crews were not the problem — customers praised the work, not the warning.”
One linked voice opens as one journey of three contacts: a call at 09:34 — “My father is on a home oxygen concentrator” — an Unhappy vote at 09:41, and the outage team’s callback the next day. The text had gone at 9:08; the power went at nine. The household was not on the priority services list and has now been added. Across the 9 late notices, 38 priority services customers were affected, shown as a count, never as names.
Customer: “Thank you. Just tell us earlier.”
The voices page shows the 16 voices the voice linker chose from the 1,032 with words, by source, district and driver: a bakery in Palm Ridge told at 22:10 for a 07:00 cut, a chat rated Neutral whose words are a complaint, and a Happy vote from Westgate despite the short notice — “Short notice, but the power came back half an hour early.” Names and account numbers are masked.
“Four hours is not notice.” — chat, Cedar Vale, rated Neutral; the reading of the words is Unhappy
With no further fixes, Q4 lands at 89.6%. The notice fix, in place since 2 October, adds 1.2 points to 90.8% — the early read is 85.9% happy after planned outages, on 412 votes. Readings used on the next bill (+0.5), passcodes through the second text gateway (+0.4) and callbacks booked in agents’ calendars (+0.3) reach 92.0%. The callback fix, owned by Aisha Bello, has not started.
“A scenario is not a commitment until its owner confirms the date.”
Rania sends the paragraph to section 2 of the September board paper, where Daniel sees it highlighted. Before it goes for signature the method checker re-runs: 14 of 14 figures match method v3.2, no cause is claimed inside the margin, and the board paper and the Q3 return carry the same figures. Daniel signs at 15:42; Fatima Rahman presents it to the board on Wed 21 Oct.
Method check: “14 of 14 figures match method v3.2.”
The quarterly service return to the Coastal Region Utility Commission shows Q3 at 90.6% on 140,697 valid votes, 0.6 points below Q2. Its explanation of movement, section C2, names the late notices, the 6,214 customers and the 38 on the priority services list, and its Q4 commitments list each fix with its owner. Daniel gives the second check; Imran signs and files.
Nothing leaves the team without both names: the CX director on the board paper, the head of regulatory reporting on the return.
The same September, seen by the five people who carry it — from the analyst who closes the month to the officer who presents it to the board.
Six agents close the month, split every move into named drivers and write the paragraph; the CX director signs.
Collects happiness-meter votes, survey files and regulator and benchmark results, applies the method document’s exclusions exactly and recomputes each index with its margin of error, then compares it with the meter platform’s own total.
Splits each month-on-month move into named drivers by service and service-quality lens, reports the change in who voted on its own as mix, and makes every cut sum exactly to the move.
Links each driver to its Voice of Customer challenge and picks up to six voices that show the problem in the customer’s own words, across sources and districts, including one that disagrees if one exists.
Forecasts the quarter and the gap to target from the monthly series, with one scenario per open fix, estimating each fix’s uplift from the driver it closes and the date it lands.
Explain this move: writes the board paragraph and the Commission narrative in the board paper’s plain style — the number, the cause, who was affected in their words, what is being done and where the quarter will land.
Traces every number in a draft to the month close, the bridge, the notice log or the forecast, and checks it against the method document, the margin of error and the published score.
Signs; regulatory reporting files. The agents propose; a named person decides.
Ask why the score moved, who was affected, or whether the quarter will make target — every answer cites the method, the figures or the voices behind it.
Why did the index fall 2.1 points in September?
Five problems took 2.8 points and two August fixes gave back 0.6; a change in who voted cost 0.1 and all other services added 0.2. The largest is late notices of planned power interruptions: −1.3, from 2,612 votes whose happiness fell from 87.6% to 63.4%. On 9 of 14 planned outages the notice went out under the 48 hours promised in service standard GS-4.
Which customers were hurt most by the late notices?
6,214 customers on 9 planned outages. Cedar Vale is worst: 1,184 votes at 55.6% happy, against 87.6% across planned outages in August. 38 are on the priority services list, and one household on home oxygen heard only after the supply was off.
Can we still reach 92.0% in Q4?
Only with four fixes. With the notice fix in place since 2 October, Q4 is forecast at 90.8%. Adding readings on the next bill (+0.5), passcodes on the second gateway (+0.4) and booked callbacks (+0.3) brings it to 92.0%. The callback fix has not started; Aisha Bello owns it.
Is the 0.4 portal drop real or just noise?
Real. 4,420 votes, happiness 81.0% against 85.4% in August; the margin on that bar is ±0.1 points at 95%. Only two bars sit inside their margin: who voted (−0.1) and all other services (+0.2), and the paper does not call either a cause.
13 screens from the working solution, on its sample data. Pick one to see it large.
How the index is computed, when a move must be explained, and who signs what leaves the team are settings, saved as a draft for approval.
| Setting | Default | Choose from |
|---|---|---|
| Index | Happy ÷ all valid votes | Defined by the method document, here v3.2 — a change needs two signatures and a re-run of the last 12 months |
| Sources counted | 4 sources | App, website, service centres, text after a contact |
| Excluded | 2 rules | Staff test votes; repeat votes from one customer on one request within 24 h |
| Words-only voices | Evidence, not counted | Emails, chats and calls without a vote are linked to drivers but never counted in the index |
| Explain any move of | 0.3 points | At index level, month on month — set in points |
| Name a driver only with | 200 votes | Votes in the month; margin of error at 95% confidence |
| Cuts shown | 3 cuts | Driver · Main Service (6) · service-quality lens (7) |
| Who signs what | Daniel Okafor · Imran Qureshi | Board paper: CX director · Commission return: head of regulatory reporting, second check by the CX director |
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 data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures: Ofwat PR24 final methodology via Oxera (C-MeX worth ±18% of annual allowed residential retail revenue); J.D. Power 2025 US Electric Utility Residential Customer Satisfaction Study (499/1,000) and its white paper on satisfaction and return on equity (a correlation, $200K–$400K a year per 10 points on $1B of equity). Lumina Utilities, the Coastal Region Utility Commission, the districts, customers and people named on this page are fictional — characters and sample data in the working solution.
Software that closes a utility’s customer satisfaction or happiness score each month, explains why it moved and prepares what goes to the board and the regulator. Customer Satisfaction Index Reporting adds six agents that recompute the index from its method, split the move into named drivers, link the customers’ voices, forecast the gap to target and draft the paper. The CX director and the head of regulatory reporting sign.
The movement explainer splits each month-on-month move into named drivers by service and service-quality lens, each driver’s share of votes times its change in happiness, and the cuts sum exactly to the move. The voice linker attaches the Voice of Customer challenge and the customers’ words behind each driver. Press Explain this move and the narrative writer writes the paragraph, every sentence linked.
Yes. A move is reported as a cause only when it is larger than its margin of error at 95% and the driver has at least 200 votes; anything else is shown hatched and never named. The change in which districts voted is reported on its own as mix. In the sample month, who voted cost 0.1 points and was not reported as a cause.
Each score is recomputed from its own method document. In the working solution Lumina reports against four: its own Customer Happiness Index, the Commission’s quarterly service survey with its reward and penalty lines, a half-yearly national service index and the government mystery shopper. It also shows how the Commission’s score has followed the utility’s own over eight quarters.
The forecaster forecasts the quarter from the monthly series, with one scenario per open fix. Switch a fix on or off to see its effect on the quarter. A scenario is never presented as a commitment until its owner confirms the date.
Names and account numbers are always masked on voices and on the paper, and the voice linker never shows them. The priority services list appears as a count — in the sample month, 38 customers among those told late of a planned outage — never as names.
Yes. The agents never publish a paper and never change a published score. The CX director signs the board paper; the head of regulatory reporting signs and files the regulator return after a second check by the CX director. A method change needs both signatures.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your index method documents, survey and vote exports, a recent board paper and regulator return — and it goes live once every quality gate has passed. We will walk you through it on a month of your own scores first.
Every voice read, and each recurring problem named with an owner and a fix.
1-min demoClosed-Loop Feedback ManagementNo unhappy customer left without a call back, and every recovery measured.
1-min demoComplaint Root Cause AnalysisEach complaint trend proved with system data, owned, and fixed at the cause.
1-min demoWe’ll run Customer Satisfaction Index Reporting on a recent month of your own votes and survey results.