Where is the lost oil, and why?
36 horizontal Bone Spring and Wolfcamp wells in Eddy and Lea counties, New Mexico. Each well gets its own Arps decline curve. The gap between that curve and what the well actually made over the last 12 reported months is split into four causes and priced. Tap a well on the map or in the table.
Field map · color = recoverable value
Wells ranked by recoverable value
| # | Well | County | Formation | Expected bbl/d | Actual bbl/d | Gap by cause | Gap bbl/d | Recoverable $/mo |
|---|
Portfolio: expected to actual
Average daily oil over the last 12 reported months for all 36 wells. Each step down is one cause of lost production. The green step up is wells running ahead of their own curve, usually after a workover or lift change.
Morning report · 5 wells to act on today
The ranked table looks back 12 months. This report looks at what is still open in the last two reported months and what it costs to act, so the order is different on purpose.
Daily surveillance
Method, assumptions and limits
Data
Steps
- Monthly oil is turned into two rates: calendar-day rate (oil ÷ days in month) and producing-day rate (oil ÷ days the well reported producing).
- An Arps hyperbolic decline,
q(t) = qi / (1 + b·Di·t)^(1/b), is fitted to each well's producing-day rate in log space. The fit uses months from the early peak up to 12 months before the data cutoff, and only months with at least 90% uptime. b is bounded between 0.1 and 1.6. - The evaluation window is the last 12 reported months (Aug 2025 to Jul 2026). The expected rate is the fitted curve at full uptime.
- Gap = expected − actual, in bbl/d, built month by month and split into the four causes below. Months where a well beat its curve are tracked separately as "ahead of curve" and do not cancel another well's gap.
- Recoverable value = each cause's gap × the share an engineer can usually win back × oil price × 30.4 days.
The four causes, in plain words
Limits
- Monthly state data only. Real work uses daily SCADA, well tests and allocation; this demo shows the method, not a verdict on any well.
- Operators are not named. API numbers are public, so any well can be looked up in OCD records.
- Allocated monthly volumes can be revised, and the newest months are the least complete. August 2026 was left out because only 44 of 52 sampled wells had reported it at pull time.
- A well's own curve is the baseline. If a well was underperforming from day one, its own curve hides that. Comparing against offset-well type curves would catch it and is the next step.
- Wells flagged "ahead of curve" have probably had a workover, refrac or lift change. Their baseline should be refitted before their gap is trusted.
- Rules are deliberately simple and readable. The recoverable shares are planning assumptions, not measured outcomes.
Sample
Horizontal wells with active status, API numbers 30-015-44800 to 30-015-46000 (Eddy) and 30-025-44300 to 30-025-45800 (Lea), south of 32.45°N, between 104.1°W and 103.3°W. That gives wells first produced in 2018 to 2020. The first 18 per county in API order were kept, which gives a spread of formations and operators without picking for good or bad stories.