What clean utility bill data needs before you can validate it

Validation gets the attention, but it only works if the bill has already been turned into clean, structured data. A PDF is not something you can recompute against a tariff. The quiet work of extraction and normalization is what makes the rest possible.

Extraction: every field, any layout

Utilities format bills differently, and the same utility changes its layout over time. Extraction has to pull the account, meter, service period, usage, demand and each individual charge reliably across all of them. Miss a rider line and the recomputation is wrong before it starts.

Normalization: one shape for everything

• Units and periods aligned so bills are comparable across accounts and months.
• Meters and accounts identified consistently, even when the utility relabels them.
• Charges mapped to a common structure so the same rider is the same field everywhere.

Then, and only then, validation

With bills as structured, normalized data and the tariffs on hand, recomputing and comparing is straightforward. The reason validation feels hard is almost always that the data underneath was never cleaned. Get the extraction and normalization right and checking bills becomes routine.