Drilling Optimization Deferred Quality Index (DODQI)

GMI-OG-010Oil and GasIntegration-Required

Definition

Drilling Optimization Deferred Quality Index, or DODQI, is the ratio of post-completion wellbore integrity events on AI-optimized wells to the same rate on conventionally drilled wells over a defined lookback window, paired with the economic ratio of excess integrity cost to drilling savings achieved. The KPI exists to answer one decision: is the AI drilling optimizer''s ROP improvement creating downstream wellbore integrity problems that cost more to fix than the drilling savings the program is booking.

The Measurement Gap

Marisol is the drilling engineering manager for an unconventional operator with three rigs running in the Eagle Ford. Eighteen months ago her company rolled out an AI drilling optimizer that adjusts weight-on-bit, RPM, and flow rate in real time against a torque-and-drag model. The vendor''s pilot report from the first six wells was clean: average 14% ROP improvement, $186,000 reduction in drilling AFE per well, no measurable change in stuck pipe rate or BHA failure rate. The operator scaled the optimizer to the whole program. Eight months in, the program lead presented to the executive team and the number on the slide was a $2.1 million annualized drilling-cost reduction across the fleet.

The wells started producing. Initial production curves looked normal. Then the integrity calls started coming in. The first one was a sustained casing pressure event on a Karnes County well at month 11: operator pulled it back, traced the leak to a poor cement bond above the production casing shoe. The cement remediation cost $340,000. Then a second well showed up at month 14 with casing wear that engineering attributed to high RPM sections in the curve. A third well at month 17 produced a workover ticket flagged as "anomalous wellbore integrity event: suspected drilling-induced." By the time Marisol pulled the population data, AI-optimized wells were producing 1.8 times the post-completion integrity events of the conventionally drilled wells from the prior twelve months. The integrity event cost was running $510,000 per AI-optimized well in excess of conventional baseline. The drilling savings the program had been booked on were $186,000 per well.

The program was now net negative on a per-well basis and the executives did not know it because the metric that captured drilling savings reported every month and the metric that captured integrity consequence did not exist. Drilling AFE variance was a real metric: it had a number, a target, and a green status. Integrity events were tracked but they were tracked as work orders against producing wells, not as deferred consequences of drilling decisions made eighteen months earlier. Nobody had built the bridge between the two.

Legacy drilling metrics cannot see this on their own. ROP improvement is a within-the-section number: it reports how fast the bit cut formation, not whether the casing string set in that section will leak in three years. Mechanical specific energy (MSE) reports drilling efficiency at the bit, not cement-bond outcomes. Non-productive time captures rig events while the rig is still on the well, not workovers that show up 400 days later. Drilling AFE variance reports cost against plan during the well''s construction; it does not get re-opened when an integrity event in month 18 retroactively makes the construction more expensive than the plan ever assumed. The vendor''s pilot accuracy was honest: it just did not extend the measurement window far enough to see the deferred consequence the technology had created.

What is missing is a metric that asks the question Marisol could not have asked at month six: are the wells the AI is helping us drill faster also wells we will spend more on later, and by how much? That question requires a population comparison against a conventional baseline, a 36-month outcome window, and an integrity event taxonomy that ties post-completion failures to construction quality rather than to reservoir or operational causes. None of that fits inside a real-time drilling dashboard. All of it has to be built across the boundary between drilling and well integrity, which is the boundary the AI was implicitly crossing without anyone measuring the cost.

Formula

Drilling Optimization Deferred Quality Index, or DODQI, is the ratio of post-completion wellbore integrity events on AI-optimized wells to the same rate on conventionally drilled wells over a defined lookback window, paired with the economic ratio of excess integrity cost to drilling savings achieved. The KPI exists to answer one decision: is the AI drilling optimizer''s ROP improvement creating downstream wellbore integrity problems that cost more to fix than the drilling savings the program is booking.

The structural form is two numbers, not one. The first is DODQI_rate, defined as integrity events per AI-optimized well in the lookback window divided by integrity events per conventionally drilled well over the same period of well life. The second is DODQI_economic, defined as the excess integrity cost on AI-optimized wells beyond conventional baseline divided by the drilling cost savings booked from AI optimization. The rate tells you whether the AI is associated with more integrity events. The economic ratio tells you whether the dollars net out. Both are needed because the rate can be elevated without the economics flipping if the integrity events are small, and the economics can flip without the rate being dramatic if the events are catastrophic.

The lookback window is 36 months from well spud completion. Three years is long enough to capture sustained casing pressure that develops slowly, cement remediation that follows the first thermal cycle, and the workover wave that emerges after about a year of production. It is short enough that program decisions made today are answerable inside a measurement window. Operators who want earlier signal can run a 12-month and a 24-month DODQI in parallel; the trajectory between them is informative even before the 36-month number is final.

The integrity event taxonomy is the part that determines whether the metric is honest. Events that count are sustained casing pressure (SCP) declarations, casing failure events, cement remediation work orders, unexpected workovers attributed to well construction quality, and wellbore restoration jobs traced to construction defects. Events that do not count are reservoir-driven failures, operational events from production handling, and integrity issues with explicit non-construction root cause. The attribution is engineering judgment, not algorithmic, and it has to be documented per event by the well integrity engineer reviewing the work order. Without a written attribution standard, DODQI will drift with whoever happens to fill in the root cause field.

Wells in the two populations have to be comparable. The minimum is twenty wells per population, drilled in the same formation, similar measured depth, similar lateral length, similar trajectory style, and within the same operational period. Comparing AI-optimized Eagle Ford curve sections to conventionally drilled Bakken laterals will produce a number, but the number will not be meaningful. The construction of the populations is part of the metric and it should be defended in writing.

Proposed formula, pending validation pilots.

At a Glance

Category
Subsurface Decision Quality
← Downhole Sensor Reliability Index (DSRI)Drilling Prediction Confidence Calibration (DPCC) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment