Sideline Ledger

The challenge

A travel team’s money moves through a bank account, a payment app, and a pile of receipt photos on someone’s phone. The person keeping track is a volunteer parent, usually with a spreadsheet, usually at night. Two things go wrong. Card purchases get counted twice, once from the receipt and once from the statement. And money swept from the payment app into checking shows up as income a second time. The report still adds up, and it is still wrong, in front of a room of parents who paid in.

The treasurer needed the monthly work to shrink, and needed the numbers to be ones they could defend. The treasurer after them needed to be able to pick it up.

The approach

RIG settled the rules before writing code, then built the tool around them:

• One rule shapes everything: the bank is the truth and receipts are evidence. Every line of a posted statement is a transaction. A receipt is never money; it attaches to the bank line it proves. Transfers between the team’s own accounts move balances and are kept out of income and spending.
• AI does the reading, arithmetic decides whether to trust it: the treasurer drops receipt photos and statements into two folders and AI extracts them. A model asked for every transaction will sometimes return most of them, so the statement’s opening balance plus every parsed line must equal its closing balance. If it does not, the statement is held back and a person decides.
• The monthly grind, removed: uncategorized transactions are grouped by vendor, so the same gas station eight times is one decision. Each decision is remembered and applied to the next statement. A learned rule only ever fills a blank; it never overwrites a category a person set.
• Reports that keep their word: monthly and annual reports freeze their figures once issued, so a later correction cannot rewrite what parents were handed. Each one opens with what the books do not yet know: statements not posted, payments with no receipt, transactions with no category.
• Built for a volunteer, and the next one: it runs on the treasurer’s own computer with no account to create. Receipt reading works with the AI account most people already have, can use a model on the same machine, or can be switched off. A printable setup guide and a one-command backup are there for the handover.

The outcome

Sideline Ledger is delivered and in the hands of its first team’s treasurer. It was not started from zero: it reuses the financial core of RIG Suite, reworked for a team instead of a business, which is why it could be built quickly without cutting the checks that matter.

Those checks are the part we would point to. Seventy automated end-to-end tests cover the failures a spreadsheet hides: double counting, transfers posing as income, a receipt imported twice, a report that will not render. They caught six real defects during the build.

We have not published usage results yet. When there is a season of reports to measure, the numbers will go here.

Working with RIG

Ask about Sideline Ledger