The Tank Ledger is the public-facing record of every tote that's moved through PTR. We publish totals and per-tank entries at /tank-ledger, and we audit the thing every quarter. Our internal audit process is not fancy — it is spreadsheets and shoe leather — but it catches about three percent of entries that need correction each cycle.
The audit protocol
- Random sample of 32 tank IDs — enough for statistical meaningfulness across our typical quarterly volume of about 2,800 intakes.
- Physical verification of 8 of those tanks if they are still in-yard, or paperwork verification of the other 24 that have shipped.
- Interview with the intake or wash operator who logged each sampled entry, when they are available.
- Cross-check against the customer's pickup paperwork and our weighbridge records.
- Discrepancy log for anything that doesn't reconcile. One page per discrepancy.
What we find, on average
Across the last eight quarters, our average discrepancy rate on sampled entries is 3.1%. The most common errors:
- Previous-contents label transposed between two adjacent intake rows.
- Grade upgraded or downgraded without a matching wash-line record.
- Disposition (resold vs. granulated) mis-entered.
- Ownership transfer documented on the wrong date.
Almost never: fabricated entries. The one case we had of an operator making up a previous-contents field (Q3 2021) was corrected the week after the audit, and the operator in question now runs our wash line. He says the audit is what made him pay attention, which is the right reaction.
The ledger is a public artifact. If we can't defend each row when asked, we don't get to publish the total.
We publish the corrections. If you read the ledger and see a “corrected on” timestamp, that's us in the audit cycle. The fact that you can read the corrections is more important than the fact that the corrections exist.