Definition
The Field-Level Recovery Interference Index (FLRII) measures the ratio of the full-field simulator's EUR forecast to the sum of individually optimized per-well EUR forecasts.
Plain language.
The Measurement Gap
Lina is the asset manager for a tight oil operator with about 180 active producers across four contiguous sections in the Eagle Ford. The asset team uses an AI optimization platform for completion design: the model takes log data, offset well performance, and prior completion outcomes and recommends frac stage spacing, cluster spacing, and proppant loading per well. The model has been in production for two years and has informed the design of roughly 95 wells. Each individual well's actual production has tracked within 8% of the model's per-well EUR forecast on average, a credible track record.
In a routine reserves audit, the third-party engineer flags an inconsistency. The sum of per-well EUR forecasts for the asset's 180 wells totals 26.3 MMBO. The asset's full-field simulator, updated last quarter with current production data and pressure measurements, projects 21.4 MMBO of ultimate recovery for the asset. The gap is 4.9 MMBO (roughly 19%) and it has been growing across the last three reserves cycles. The wells are interfering with each other in ways the per-well model does not capture. Each well's individual forecast was honest. The portfolio-level total was a fiction.
Lina runs the trace. The per-well AI optimizer treats each well as if it draws from an undisturbed reservoir region. The completions it recommends pull harder on the formation than the spacing supports. Pressure communication between offsetting wells reduces each well's actual draw area below what the per-well model assumed. The wells produce well (within 8% of forecast) but the field as a whole produces less than the simple sum of the per-well forecasts. The operator has been booking reserves and approving development capital on the sum-of-wells number rather than the field-level number, and the gap has been compounding as the asset develops.
This is not a story about a bad model. The per-well model performed well on its terms: predicting per-well outcomes within tolerance. The structural problem is that per-well optimization optimizes the wrong thing when wells share a reservoir. The right thing to optimize is field-level recovery. Per-well optimization is a useful tool for completion design within a fixed development plan; it is not a substitute for field-level evaluation.
Legacy reservoir metrics cannot detect this directly. Per-well EUR is the metric the AI optimizer was tuned to improve, and on a per-well basis it is improving. Recovery factor at the asset level is computed retrospectively and is fundamentally lagging: it will eventually reveal the gap but only after multiple years of production. Decline rate analysis on individual wells reveals their performance but does not flag the interference effect: each well's decline can look normal while the field-level recovery is leaving barrels in the ground. The gap between the sum of per-well EURs and the field-level EUR is the right number to look at, but in most operations nobody is computing it as a real-time metric.
The measurement gap is precise. AI per-well optimization is structurally blind to inter-well interference, and current reporting practice does not surface the gap between the sum of per-well forecasts and the field-level forecast as a measurable, trackable number. The question FLRII proposes to answer is the one Lina could not ask quarter-by-quarter: is per-well optimization destroying total field value through interference?
Formula
The Field-Level Recovery Interference Index (FLRII) measures the ratio of the full-field simulator's EUR forecast to the sum of individually optimized per-well EUR forecasts.
Plain language. Take every well in the field and sum its individually optimized EUR: what the per-well AI says each well will produce in isolation. Now run the full-field simulator with the same wells in their actual configuration and let it tell you what the field will produce as a system. Divide the field-level number by the sum-of-wells number. That is FLRII. If wells did not interfere with each other at all, FLRII would be 1.0: the field is the sum of the wells. The further FLRII is below 1.0, the more interference is destroying value at the field level relative to what per-well optimization assumes.
Notation:
FLRII = EUR_field_model ÷ Σ EUR_per_well_optimized
Where:
- EUR_field_model = the field-level ultimate recovery from a full-field simulation that includes inter-well pressure communication and shared-reservoir effects (MSTB or MBOE)
- EUR_per_well_optimized = each well's individual AI-optimized EUR forecast under the assumption of undisturbed reservoir (same units)
- The summation runs across all wells in the field unit
- The two EUR figures must be reported to the same recovery horizon and on the same fluid basis (oil-only versus BOE, for instance) for the ratio to be meaningful
Both forecasts must be done with the same wells in the same configuration: same locations, completions, and operational policy. The only difference between the two numbers is whether inter-well effects are included.
Edge cases. Early field life with wide spacing produces FLRII near 1.0 because wells genuinely do not interfere when they are far apart and the reservoir pressure is undisturbed. The metric is not diagnostic in the early development phase: it becomes diagnostic as the field densifies. A facility computing FLRII = 0.98 in the first year of development should not interpret that as "no interference"; it should expect FLRII to drop as the field develops, and should compute it through the development cycle to see the trend.
A full-field simulator is required. The metric cannot be computed without a simulator that models inter-well effects. Most operators have a simulator but it is often out of date (see PMFDR discussion). Updating the simulator state quarterly is a prerequisite for FLRII to be diagnostic. Operators who do not maintain a current field-level simulator cannot compute FLRII and should treat that as the first investment, not as a reason to skip the metric.
Segment by recovery mechanism. Primary depletion fields, waterflood fields, gas injection fields, and tight oil fields all have different interference dynamics. A primary depletion field will see less interference at wider spacing than a waterflood field. A tight oil field with hydraulic frac stimulation has interference both from pressure depletion and from frac hits. A facility computing FLRII across mixed recovery mechanisms aggregates effects that are structurally different. Compute per recovery mechanism and report by segment.
Proposed formula, pending validation pilots.