Permian Completions Weekly

Everyone remembers the loading race. From 2016 to 2018 the median Midland Basin well went from 1,510 to 1,854 lb per perforated foot — a 23% jump in two years, the era when every investor deck had a "bigger fracs" slide. Then the race just… stopped. Since 2018 the median has crept from 1,854 to 2,036 lb/ft — about 10% in seven years — and the 2026 wells joined so far sit at 2,103. Call it ~2,000 lb/ft: the design equilibrium the basin settled on once the cost-per-barrel math stopped rewarding more sand.

The sand bill still doubled — but sideways. Median perforated interval went from 7,657 ft (2016) to 10,432 ft (2025), and median sand per well from 12.0M to 24.6M lb. The growth was laterals, not loading. Per-well demand kept sand mines busy while the recipe barely moved.

Design still isn't monolithic: among 2025–26 completions, Pioneer's median well takes 31.0M lb (n = 583) against Diamondback's 20.1M (n = 492) — same basin, same benches, an 11-million-pound difference in what shows up on the pad.

The basis. Completion cost is now built per lateral foot: four foot-priced lines — pumping spread and stimulation service, wireline/plugs/pumpdown, chemicals and additives, and site, rentals, supervision, fuel and other — at $225 per completed foot combined; two pass-throughs priced on the job's own measured volumes — proppant at $0.016/lb on the disclosed sand, and water split by the disclosed produced/recycled share; and three flat per-well lines — drillout, flowback and early production, and tubing, wellhead and surface facilities. The sand-throughput day term is gone entirely.

Where the lateral comes from, and what is wrong with it. The 10,362 ft is a perforated interval, not a drilled lateral, and it is not this issue's wells. It is the median perf span of the lagged cohort: Texas wells completed 6 to 12 months before this issue (n = 1,608, 82% TX coverage). It has to be, because RRC perforation records land 4 to 8 months after the frac — the direct join on this issue's own settled jobs is 4.0%, and on a 28-day window it is essentially zero. A current-window lateral does not exist. Worse, the few recent joins are biased large — prompt filers are big-job operators — so using them would overstate the median. New Mexico has no perforation source at all: dbf900 is a Texas RRC file, only 7 of 15,652 NM disclosures ever join, and NM is a third of every window. The foot priced here is therefore a Texas perforated foot applied basin-wide. For scale, the median permitted lateral this issue is 11,123 ft, 7% longer — the expected heel/toe standoff between drilled and perforated.

The sanity check that should have caught this. The old number was $10.2M on a ~10,362 ft well. Divide it by the lateral: ≈$984 per foot for the completion alone — more than the all-in drilling-and-completion cost per foot of every public Permian operator, which runs about $675–$1,035/ft (Permian Resources at $675–685/ft; Matador at $835–855/ft; Civitas Delaware at $880/ft). A completion-only number cannot exceed an all-in number. The rebuilt figure lands at $410/ft, implying roughly $710/ft all-in once about $300/ft of drilling is added — inside the operator band rather than above it. Standing rule: check any well cost per lateral foot against operator guidance before it ships.

Pricing the median understates the basin. The lateral distribution is bimodal, not a bell: on the well-covered cohort 38% of wells fall in the 10,000–11,000 ft (two-mile) band and 14% in the 15,000–16,000 ft (three-mile) band — roughly one job in seven is a three-miler. A median-only model prices those wells light. Holding these rates and moving only the lateral: a p25 10,018 ft well costs $4.18M and a p75 13,305 ft well costs $4.92M, against $4.25M at the median. The basin run-rate above is therefore a floor, not a midpoint.

Quantities. The median sand, water and produced-water share are this issue's own settled window (job end 42–126 days back), the same population the old model priced. The retained FracFocus snapshot of this vintage reproduces the originally published cost to the digit, so the correction is a change of method, not of data. The settled pace is the original published 308.3.

The per-operator table has been withdrawn, not corrected. It was generated by the same defective model. Rebuilding it needs per-operator lateral, sand and water for this window, and the retained inputs do not support that at operator granularity. Publishing a corrected basin figure beside an uncorrected operator table would be worse than dropping it.