60 water bills, one spreadsheet, and a culprit in the front yard

A Katy household's water bill doubled in 2024. In 2026 it doubled again. Then the utility district presented a 10-year plan with a third round of increases, one scenario including a flat $100 a month per connection.

The ratepayer did what we'd tell any owner to do before accepting a number: audit it. No inside information. Only documents any resident can get: five years of bills, the district's proposed budget, its published rate order, and the handouts from the board meeting.

Step 1: reconcile before you argue

AI transcribed 59 monthly bills. The one missing bill was reconstructed from the meter reads on either side. All 60 were rebuilt from the raw meter readings. A rate model reverse-engineered from the line items reproduced every water charge to the cent, 60 of 60. The meter chain had no gaps in five years.

So the district's arithmetic was clean. That finding shaped everything after it. Concede what the data proves and contest only what it doesn't. The amount the district needed was honest. How it planned to collect it was the question.

Step 2: build the measurement nobody asked for

The plan, as presented, answered one question: how much money is missing? It then adjusted rate tables until the gap closed. The prior question, what each service actually costs to deliver, was never asked. One afternoon with the district's own budget answered it:

The books close. That's a cost-of-service measurement, done from public records with a spreadsheet and AI.

Step 3: let the measurement expose the design

Three things surfaced once unit costs were on the table:

• Water was priced below cost. True cost: $6.95 per 1,000 gallons. Proposed price where most usage lives: $5.90–$7.90. Flat fees on everyone were subsidizing the heaviest irrigators.
• A third of the "sewer" charge was the trash bill. By the presentation's own note, the proposed $55 sewer fee was "designed to cover Garbage & Recycling." A cost filed under the wrong line can't be checked by the people paying it.
• The increases landed hardest on the smallest users, because they lived in fixed charges. In the 2024 restructure, a 3,000-gallon month went up 196%. A 25,000-gallon month went up 88%.

Step 4: counter-propose at the same bottom line

Criticism without an alternative is noise. Using the district's own distribution of 679 meters, the ratepayer calibrated a replacement structure:

• a low fixed base;
• water priced above its true cost;
• sewer near its measured cost;
• commercial accounts moved toward parity;
• a modest tax component.

It produces the same total revenue, within 0.2%, checked formula by formula in an open workbook.

Same money, opposite incentives. Under the district's lead scenario, a 7,000-gallon neighbor would see +96%; under the counter-proposal, +19%. Heavy users would pay above what their water costs to deliver.

The twist

The data also found the household's own problem. Starting in mid-December 2025, the sprinkler system kept running through the winter, three days a week, at about triple what the lawn needed. Every 2026 winter month used more water than every winter month of the four years before. There was no overlap at all.

So under the ratepayer's own counter-proposal, the ratepayer's own bill goes up more than almost anyone's. They proposed it anyway. An argument that costs its author money is hard to dismiss.

Why this is on a business blog

The same method works on a utility district, a plant, a fleet, or a front office:

1. Reconcile the record before you trust it or attack it.
2. Measure unit costs from primary documents, not summaries.
3. Let the measurement expose the design, not the other way around.
4. Change the design, not the total, when the total is right.

The AI did the tedious part: five years of bills transcribed, a rate model rebuilt, scenarios recalculated in minutes. A person decided what to ask, checked every figure against the source, and took the result to the board. That's the partnership we set up for owners. Your people know the business, the tools learn it from them, and the numbers get checked before anyone acts on them.

If your numbers were built backwards from a conclusion, we can rebuild them forwards. Request a 20-minute fit conversation.