Analyze · The data
The sample dataset
Every case study in this section runs on one made-up dataset: 14 wells, ten months of 2026, and the tables an operations accountant works with every day. It's small enough to read and big enough to hide real-looking mistakes, because some were planted on purpose.
Everything here is invented. The wells, owners, operators, vendors, prices and amounts are made up for illustration. No employer or partner data is used anywhere on this site.
The tables
Every table carries well name + cost center, and every join uses both. That's the rule that keeps a cost or a barrel off a lookalike well.
| Table | What a row is | Rows | Columns |
|---|---|---|---|
| wells | The well master: one row per well name + cost center | well_name, cost_center, well_type, operator, field, status, status_effective, royalty_burden, company_wi | |
| decks | Each owner's working and net revenue interest by well | well_name, cost_center, owner, wi, nri | |
| afes | Authorizations for expenditure, operated and non-op | afe_no, well_name, cost_center, afe_type, gross_amount, status, approved | |
| invoices | AP invoice lines on operated wells, at 100% | invoice_id, vendor, invoice_no, invoice_date, month, well_name, cost_center, afe_no, cost_type, account, amount | |
| jib_lines | Each invoice line split to each owner by the JIB deck | month, well_name, cost_center, afe_no, cost_type, owner, wi, amount, invoice_id | |
| nonop_jib | Operators' JIBs received on non-operated wells (the company's net share) | month, operator, well_name, cost_center, afe_no, cost_type, gross_amount, company_wi, net_amount | |
| production | Monthly oil, gas and water by well | month, well_name, cost_center, oil_bbl, gas_mcf, water_bbl | |
| revenue | Revenue paid to each owner by well and month | month, well_name, cost_center, owner, nri, gross_value, net_revenue | |
| accruals | The company's net accrual by well and group, with the actual that came in | month, well_name, cost_center, cost_group, accrued_net, actual_net |
What's planted in it
These are the errors the case studies are built to find. Each mirrors something that happens in real review.
| Planted error | Where it shows up |
|---|---|
| Jones Unit 1 revenue for September and October booked to lookalike cost center CC-2210 | SQL: lookalike cost center and billed, not paid |
| Smith 2H Partner B dropped from the revenue deck from July (suspense) | SQL: billed, not paid |
| Baker 1 Partner A paid at NRI 0.1875 instead of 0.175 | SQL: NRI check |
| Davis 4H workover posted $40,000 over its $1.25M AFE | SQL: AFE spend and the dashboard |
| A duplicate invoice, an invoice coded to the wrong cost center, a charge on a closed AFE, a Capital line on a P&A AFE, and a charge after Clark 2 was sold | SQL: coding exceptions |
| Accruals that missed: AFE field estimates that lagged, Smith 3H accruing after it was shut in | SQL: accrual accuracy, the Excel review and the dashboard |
| An operator billing a well that isn't in the cross-reference | Power Query: unmapped wells |
Look inside
Showing the first 25 rows. Copy pastes the whole table into Excel or a text file. When the site is hosted on its own domain, each table is also available as a CSV file in analyze/data/csv/.
The data was generated with a short Python script from fixed settings, so it's the same every time. All of it is invented.