Oil and gas production is a system, like any other. It ends with product out of the ground, but it starts long before, weeks and often months before that with the simple process of clearing the pad. Then the drill rig comes in to actually poke the hole in the ground. After that, the well is cased and cemented. At this point, the well is a DUC (Drilled and Uncompleted). The frac crew then comes onsite, and performs the fracing in stages - perforating one section, injecting, then installing plugs to isolate previously frac'd sections. Once that is complete, the plugs are drilled out (or dissolved) and then the well can be finally "tied-in-line" (TIL).
And like any system, slack between those steps can be used to make sure each step can run efficiently. Some DUC inventory is necessary to ensure rigs and frac crews can operate independently. Some Delayed TILs (DTILs) can be used to match completed wells to market conditions. But the downside for an observer is that these buffers can result in delays to changes in production from earlier steps, or in many cases, the noise and variation in earlier steps can be used to smooth out the result. In addition, since each step operates semi-independently, it is possible for changes to happen out of "sequence" - i.e. a change in rigs and DUCs doesn't necessarily have to translate to a change in production.
Traditionally, rig count was the only reliably visible, near-real-time measure in this system. With our frac crew, completions and DTIL data, that is no longer the case, and that brings us to ...
The recent rig count decline has been getting all the press, as the only broadly available number, and the decline is certainly concerning - but as discussed above, there is enough slack in the system in the DUC inventory that rig count can stay down for a while and never impact production. (See this article for our DUC methodology.)
And as you can see below - Permian operators have been drawing down that slack in the system - DUCs - for quite a while. (The slope of the DUC count is really a function of the rig to crew ratio, as noted.)
But a decline in frac crew counts actually preceded the decline in rigs by a few months in West TX. After a brief rise, it has been declining, steadily since February. The dynamic in Permian-NM has been the reverse - with rigs declining since Feb, and crews just starring to decline since April.
And in Permian, declining crew counts are definitely more concerning than in the NE, since we've neither seen evidence of nor heard of much DTILs in Permian. Which means that reduced crew counts tend to lead directly to reduced completions, all other things being equal. And that's what it appears we are seeing:
Though completions have dipped before (see Dec 2024), West TX appears to be in a sustained downswing. It's probably too early to tell if Permian NM is. There has been a sharp decline there - but it was from a higher level, and it has slightly rebounded.
Given the lower level of completions and the lack of DTIL inventory in the Permian, this should tend to result in reduced production relatively quickly. But the question is how much and how soon? For that we can turn to another feature of Hyperion, Production Studio. In our case, we ran three different scenarios, which happened to use the three different modes of Production Studio. We used the same inputs for all three cases - 10%/year improvement in IP Rate, start date of 1 Jan 2025 (after the end of recent Permian production data), same type curves, and 30 days from completion end to TIL.
The three cases were:
(For all of these, you should ignore the volatility of completion count - the mechanistic nature of these simulations tends to shove the completion counts around a bit to match production exactly.)
The key takeaway to these is that the difference between growth and flat is only about 15% at this point. Now, taking the flat crews scenario to production, and comparing with the LTF scenario gives us this:
(Note the zoomed in scale). Specifically, if crews stay this low, production might be declining already.
To be clear we are NOT forecasting that Permian oil production has peaked. But what is clear is that we are very near that cusp, and that real time monitoring of Permian completions is critical at this time: state data is long out of date and modelling is usually useless during these dynamic periods.