The producer — a horizontal Three Forks lateral in McKenzie County.
An independent operator runs one well and reviews production once a week — the GOR climb starts within the band their spreadsheet plots, so the weekly review misses it entirely.
The operator in this note runs a single horizontal Three Forks lateral in McKenzie County, North Dakota — a two-mile lateral on a sixteen-stage completions job, on artificial lift, two months past first oil. Their existing process is the one most independent operators of their size actually have: an operations engineer downloads the WITSML snapshot Monday morning, opens the reservoir-engineering spreadsheet, and reviews the trend by eye. That works for the signals a daily review was designed to catch — water-cut creeping two points, casing pressure building against a fourteen-day baseline. It does not work for a signal that resolves itself across several weeks, because the GOR shifts within the noisier band the spreadsheet plots.
Day 3 onward — the GOR drift.
Gas-oil ratio climbs from roughly 1,800 scf/bbl at the baseline to roughly 2,350 scf/bbl over fourteen days — a monotonic rise of about three-and-a-half percent per day, well below the SCADA alarm, missed by the weekly plot.
On day three the GOR channel — the gas-oil ratio trace coming off the test separator, indexed against the WITSML clock the operator already runs — starts to drift. Across fourteen days it climbs from a baseline near 1,800 scf/bbl to roughly 2,350 scf/bbl, monotonically, at about three-and-a-half percent per day. The climb sits well below the SCADA “GOR high” alarm, which trips at 3,200 scf/bbl — the operator's existing trip still has a thousand scf/bbl of margin. The shape of the curve, on the spreadsheet plot the engineer eyeballs each Monday, looks like the scatter the well has always exhibited. The climb resolves itself before the next Monday review.
Day 3, inside twelve hours — always-on monitoring.
Orogen's per-curve rolling envelopes on the GOR channel fire a drift trip inside twelve hours of the climb starting — the same basin-noise-floor envelope story as the Permian walkthrough, but anchored to a different signal.
Orogen is bound to the same WITSML store the operator already runs. The GOR channel gets its own rolling envelope, built off the offset-well baseline window the operator's reservoir team has on file. On a Bakken Three Forks pad the basin noise floor on GOR sits around two hundred scf/bbl — the alarm envelope we tune toward is tighter than the SCADA trip and looser than the weekly eyeball. Twelve hours into the climb, the GOR trace crosses its rolling envelope. The signature is a monotonic rise, not the spike-and-relax pattern the basin noise floor produces; the agent reads the rise as a gas-coning or a near-wellbore breakthrough signal, writes the appropriate imperative copy, and pushes the alert with the raw WITSML trace window attached.
Week 2 — the Monday hand-off.
The flag lands in the Monday morning one-page PDF the fleet emits, with the raw trace window attached — engineering walks into the standup with the flagged event already triaged into the same brief the night foreman has already seen.
Two weeks into the pilot, the Monday hand-off lands as it has every Monday since the fleet came online: a one-page PDF in the operations lead's inbox before the standup, summarising the well's last seven days — production rate, water-cut delta, casing pressure, and the AI-written commentary the agent authored from the trace windows. The flagged GOR event from day three is already triaged into the PDF's single screen-ready read-out, with the raw cohere vector and the raw WITSML trace window behind it. Detection to briefed queue, in hours instead of weeks — and the next monthly review reads the well's GOR trend with a fourteen days of context, not a single Monday-morning eyeball.