We rebuilt every producing horizontal oil well in Converse County, Wyoming from the state's own records, fitted each one, and asked a single question: at what point does another pound of sand stop paying for itself. The answer is not where the industry's behaviour suggests it believes it is.
Converse County holds 1,459 producing horizontal and directional oil wells, out of 2,863 across the Powder River Basin and 4,023 statewide. We pulled the monthly production history of every one of them, together with 531,535 formation tops, 149,933 perforation records and 75,335 completion treatments covering the whole state.
All of it is public. It comes from the Wyoming Oil and Gas Conservation Commission, and every number below can be checked against the same filings by anybody who wants to. No client data was used and none of this is anyone's confidential information.
Of those wells, 1,371 had at least eighteen producing months. We fitted 1,368 of them and refused 174 outright, either because the hyperbolic exponent pinned against its bound or because the fit quality was too poor to mean anything. Refusing a fit is not a failure. Reporting one that should have been refused is.
Rate per producing day, not per calendar month. A well that was down for eleven days did not decline. It was off. Dividing by the month confuses the two, and downtime then gets forecast forward as reservoir behaviour.
The flowback month dropped. The first partial month is choked and cleaning up. It is not on the depletion trend that governs everything after it.
Fitted on the logarithm of rate. Production spans two orders of magnitude across a well's life. A fit on raw rate is dominated by the first six months and effectively ignores the tail, which is exactly where the reserve lives.
A terminal decline switch. The curve runs hyperbolic until its own decline flattens to six percent a year, then exponential to the economic limit. Without that switch, a well fitted at an exponent of 1.4 books sixty percent more oil than it will ever deliver, and at 1.8 it books almost nine times too much. That correction alone is larger than most of the arguments people have about type curves.
Across the wells that passed, the median hyperbolic exponent is 1.05, the median secant decline 76.2 percent, and the median fit quality 0.886 on log rate.
Median estimated recovery per well, and the same figure divided by perforated length. Wells with under a thousand feet of perforation are excluded from the second column, because a vertical completion returns a ratio in the millions and that is arithmetic rather than geology.
| Formation | Wells | Median recovery, bbl | Per 1,000 ft |
|---|---|---|---|
| Niobrara | 436 | 414,225 | 46,269 |
| Turner | 255 | 325,713 | 50,604 |
| Parkman | 109 | 339,260 | 43,451 |
| Sussex | 101 | 223,342 | 41,721 |
| Frontier | 92 | 273,860 | 36,624 |
| Teapot | 60 | 325,103 | 32,816 |
On raw recovery the Niobrara wins and it is not close. Normalised for how much rock each well was actually completed across, the Turner is ahead of it. The Niobrara's advantage is substantially that its wells are longer, which is a decision somebody made rather than a property of the rock.
| Year | Median proppant per well, lb | Median stages |
|---|---|---|
| 2012 | 848,205 | 11 |
| 2016 | 1,393,619 | 14 |
| 2019 | 4,921,560 | 19 |
| 2022 | 12,044,437 | 38 |
| 2025 | 24,531,509 | 41 |
This is the single most important number in the study, and not because of what it says about fracs. It is important because it means any comparison of wells across vintages is measuring the calendar unless it explicitly controls for design. A performance map built without that control is a map of when people drilled, wearing a geologist's coat.
Wells split at the median proppant per foot, compared only against others completed in the same year. That restriction costs sample size and is the whole reason the answer is trustworthy.
| Completion year | Wells | Extra lb per ft | Extra bbl per ft | Pounds per extra barrel |
|---|---|---|---|---|
| 2012 | 27 | 428 | 14.5 | 30 |
| 2013 | 56 | 312 | −1.2 | no gain |
| 2020 | 31 | 1,065 | −0.7 | no gain |
| 2021 | 41 | 742 | 3.0 | 245 |
| 2022 | 96 | 1,625 | 11.6 | 140 |
| 2023 | 99 | 1,198 | 8.6 | 140 |
| 2024 | 115 | 1,736 | 17.4 | 100 |
Weighted across cohorts, an extra barrel costs about 129 pounds of extra proppant. Recovery scales with roughly the quarter power of intensity, so the response is real, consistent, and nowhere close to proportional.
Two cohorts bought nothing at all. In 2013 and 2020 the heavier half recovered marginally less per foot than the lighter half, despite pumping substantially more sand.
A ten thousand foot Niobrara lateral, priced through a monthly cash flow with capital spent before first sales rather than alongside it. Oil at $70 less a $6 differential, sand at nine cents a pound delivered and pumped, three barrels of water per barrel of oil, discounted at ten percent. Change any of those and the answer moves.
| Step in intensity | Extra capital | Gain in present value | Return per dollar of sand |
|---|---|---|---|
| 1.00x to 1.25x | $281,250 | +$224,687 | 0.80 |
| 1.25x to 1.50x | $281,250 | +$154,378 | 0.55 |
| 1.50x to 2.00x | $562,500 | +$168,357 | 0.30 |
| 2.00x to 2.50x | $562,500 | +$43,634 | 0.08 |
Every step still adds barrels, and every step still adds value, which is precisely why the decision keeps getting made. But the last one returns eight cents of present value per dollar of sand, which is inside the noise of execution risk, price and water handling. It is a coin flip dressed as a capital programme.
The base design does not clear a ten percent hurdle at these terms at all. It returns nine percent. The heavier completions are not an optimisation on a good well. They are what makes the well work.
Recent wells are forecasts, not observations. The 2022 to 2024 cohorts have short histories, so their recoveries lean on extrapolation, and the terminal decline switch moves such numbers materially.
Perforated interval stands in for lateral length. Neither the state nor the operator publishes the latter directly. Directional surveys would replace the proxy and are the obvious next improvement.
There is no pressure data in the public record. So there is no rate transient analysis and no material balance, which means no independent check on contacted volume. Decline analysis alone cannot tell you whether a well is draining what you think it is.
Costs are parameters, not filings. Well cost is not public. Everything in the decision table moves with the sand price, the differential, water handling and the discount rate, all of which are inputs a reader can and should change.
This compares medians within cohorts. It is not a regression, and it does not control for formation within a completion year.
Monthly production, formation tops, perforations and completion treatments all come from the Wyoming Oil and Gas Conservation Commission. Well locations and status come from its public map service. Nothing here required a subscription, a licence or a login.
That is deliberate, and it is worth saying plainly. The commercial data vendors prohibit building tools or models on their data. Any study that can actually be reproduced by its reader has to be built on the public record instead.
Ask us to run this on your asset, in your basin, against your own production history.
The Asset Health Check is the cheapest possible way to find out whether we are worth the larger engagement. Two weeks, fixed fee, and a written answer either way.